diff --git a/run.log b/run.log index 5bb2eddbd1d60a9f8ea8178ce577f9298739dd00..09681ae18be2d9a64161121a9e3f267668de9e48 100644 --- a/run.log +++ b/run.log @@ -1,59115 +1,63024 @@ -[2025-11-09 17:59:18,021][mllm.models.large_language_model_local][INFO] - Initializing adapter 'agent_adapter': no initial weights provided or found; starting from scratch. -[2025-11-09 17:59:18,816][mllm.models.adapter_training_wrapper][INFO] - Adapter 'agent_adapter': initialized with fresh weights (no initial weights found). -[2025-11-09 17:59:18,823][mllm.models.large_language_model_local][INFO] - Initializing adapter 'critic_adapter': no initial weights provided or found; starting from scratch. -[2025-11-09 17:59:19,941][mllm.models.adapter_training_wrapper][INFO] - Adapter 'critic_adapter': initialized with fresh weights (no initial weights found). -[2025-11-09 18:01:31,249][__main__][INFO] - Starting iteration 0. -[2025-11-09 18:01:31,261][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 0 and human policies 1. -[2025-11-09 18:01:31,261][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:01:37,870][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls +[2026-03-25 21:06:08,177][mllm.models.large_language_model_local][INFO] - Initializing adapter 'agent_adapter': no initial weights provided or found; starting from scratch. +[2026-03-25 21:06:08,892][mllm.models.adapter_training_wrapper][INFO] - Adapter 'agent_adapter': initialized with fresh weights (no initial weights found). +[2026-03-25 21:06:08,899][mllm.models.large_language_model_local][INFO] - Initializing adapter 'critic_adapter': no initial weights provided or found; starting from scratch. +[2026-03-25 21:06:09,579][mllm.models.adapter_training_wrapper][INFO] - Adapter 'critic_adapter': initialized with fresh weights (no initial weights found). +[2026-03-25 21:08:20,469][__main__][INFO] - Starting iteration 0. +[2026-03-25 21:08:20,471][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 0 and human policies 1. +[2026-03-25 21:08:20,472][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:08:24,729][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls -Given the values assigned to the items, I aim to maximize my points by targeting the highest value items first. Since hats are the most valuable for me with a value of 10, I will propose keeping all 10 hats and none of the books or balls. This strategy aims to ensure that I receive the maximum value from the high-value item (hats) without risking a proportional distribution if the total proposed amount exceeds the available items. did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:01:45,237][__main__][INFO] - Number of regex retries in iteration 0: 1 -[2025-11-09 18:01:45,238][__main__][INFO] - agents played in iteration 0 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:01:57,819][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 37.45%, Block Peak % of device VRAM: 18.68%, ΔTime: 00:00:00 -[2025-11-09 18:01:57,846][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 37.45%, Block Peak % of device VRAM: 18.68%, ΔTime: 00:00:00 -[2025-11-09 18:01:57,871][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 37.45%, Block Peak % of device VRAM: 18.68%, ΔTime: 00:00:00 -[2025-11-09 18:01:57,894][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 37.45%, Block Peak % of device VRAM: 18.68%, ΔTime: 00:00:00 -[2025-11-09 18:01:57,895][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:01:57,895][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:01:58,523][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:01:59,252][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:01:59,749][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:02:00,243][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:02:00,733][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:02:01,221][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:02:01,712][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:02:02,212][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:02:02,707][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:02:03,194][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:02:03,682][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:02:04,171][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:02:04,660][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:02:05,147][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:02:05,635][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:02:06,126][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:02:06,613][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:02:07,101][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:02:07,592][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:02:08,079][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:02:08,567][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:02:09,056][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:02:09,545][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:02:10,034][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:02:10,521][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:02:11,010][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:02:11,504][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:02:11,993][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:02:12,481][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:02:12,972][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:02:13,463][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:02:13,954][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:02:14,444][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:02:14,932][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:02:15,419][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:02:15,906][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:02:16,404][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:02:16,894][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:02:17,381][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:02:17,869][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:02:18,359][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:02:18,847][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:02:19,339][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:02:19,826][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:02:20,317][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:02:20,804][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:02:21,290][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:02:21,803][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:02:22,291][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:02:22,778][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:02:23,281][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:02:23,769][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:02:24,256][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:02:24,744][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:02:25,230][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:02:25,726][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:02:26,214][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:02:26,700][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:02:27,191][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:02:27,682][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:02:28,170][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:02:28,659][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:02:29,151][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:02:29,643][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:02:30,138][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9585 tokens. -[2025-11-09 18:02:30,868][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 15.99%, Current % of VRAM taken: 53.44%, Block Peak % of device VRAM: 60.80%, ΔTime: 00:00:32 -[2025-11-09 18:02:31,523][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:02:31,524][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:02:31,526][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:02:32,597][__main__][INFO] - Iteration 1 took 1m 1s (22.79% Gen, 75.47% Train). Generation: 13s, Training: 46s. Estimated remaining time: 51h 3m 14s. Estimated total time: 51h 6m 50s. Time estimates for 10 more iterations: 10m 13s, 100 more iterations: 1h 42m 13s, 500 more iterations: 8h 31m 8s. -[2025-11-09 18:02:32,600][__main__][INFO] - Starting iteration 1. -[2025-11-09 18:02:33,039][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 0 and human policies 1. -[2025-11-09 18:02:33,039][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:02:33,826][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:02:45,852][__main__][INFO] - Number of regex retries in iteration 1: 1 -[2025-11-09 18:02:45,853][__main__][INFO] - agents played in iteration 1 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:02:46,758][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 46.48%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:02:46,780][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 46.48%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:02:46,802][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 46.48%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:02:46,824][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 46.48%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:02:46,825][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:02:46,825][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:02:47,477][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:02:48,312][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:02:48,792][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:02:49,296][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:02:49,785][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:02:50,272][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:02:50,763][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:02:51,250][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:02:51,739][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:02:52,229][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:02:52,720][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:02:53,222][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:02:53,711][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:02:54,199][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:02:54,694][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:02:55,183][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:02:55,671][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:02:56,162][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:02:56,651][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:02:57,140][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:02:57,628][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:02:58,119][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:02:58,609][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:02:59,099][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:02:59,587][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:03:00,080][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:03:00,568][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:03:01,057][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:03:01,545][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:03:02,035][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:03:02,526][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:03:03,017][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:03:03,509][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:03:04,002][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:03:04,491][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:03:05,014][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:03:05,503][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:03:05,991][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:03:06,493][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:03:06,981][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:03:07,469][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:03:07,957][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:03:08,445][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:03:08,939][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:03:09,431][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:03:09,919][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:03:10,425][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:03:10,913][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:03:11,401][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:03:11,889][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:03:12,381][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:03:12,871][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:03:13,361][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:03:13,851][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:03:14,342][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:03:14,830][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:03:15,317][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:03:15,804][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:03:16,292][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:03:16,779][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:03:17,267][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:03:17,756][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:03:18,246][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:03:18,733][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:03:19,224][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9651 tokens. -[2025-11-09 18:03:19,912][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 13.97%, Current % of VRAM taken: 55.46%, Block Peak % of device VRAM: 61.47%, ΔTime: 00:00:32 -[2025-11-09 18:03:20,647][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:03:20,649][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:03:20,651][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:03:21,639][__main__][INFO] - Iteration 2 took 48s (26.36% Gen, 71.60% Train). Generation: 12s, Training: 34s. Estimated remaining time: 40h 25m 39s. Estimated total time: 40h 30m 3s. Time estimates for 10 more iterations: 8m 6s, 100 more iterations: 1h 21m 0s, 500 more iterations: 6h 45m 0s. -[2025-11-09 18:03:21,642][__main__][INFO] - Starting iteration 2. -[2025-11-09 18:03:22,083][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 0 and human policies 1. -[2025-11-09 18:03:22,083][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:03:34,876][__main__][INFO] - Number of regex retries in iteration 2: 0 -[2025-11-09 18:03:34,877][__main__][INFO] - agents played in iteration 2 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:03:35,712][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:03:35,735][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:03:35,760][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:03:35,782][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:03:35,782][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:03:35,783][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:03:36,433][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:03:36,895][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:03:37,390][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:03:37,883][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:03:38,372][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:03:38,861][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:03:39,355][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:03:39,848][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:03:40,339][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:03:40,828][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:03:41,318][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:03:41,806][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:03:42,300][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:03:42,789][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:03:43,283][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:03:43,772][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:03:44,263][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:03:44,753][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:03:45,245][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:03:45,736][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:03:46,230][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:03:46,720][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:03:47,211][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:03:47,699][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:03:48,189][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:03:48,678][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:03:49,168][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:03:49,659][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:03:50,150][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:03:50,639][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:03:51,127][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:03:51,615][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:03:52,104][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:03:52,593][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:03:53,085][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:03:53,574][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:03:54,068][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:03:54,560][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:03:55,050][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:03:55,539][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:03:56,028][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:03:56,533][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:03:57,022][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:03:57,510][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:03:58,010][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:03:58,498][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:03:58,987][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:03:59,476][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:03:59,965][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:04:00,456][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:04:00,944][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:04:01,432][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:04:01,924][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:04:02,413][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:04:02,903][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:04:03,392][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:04:03,883][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:04:04,376][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:04:04,864][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:04:05,351][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:04:05,840][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:04:06,330][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:04:06,817][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:04:07,305][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:04:07,797][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9654 tokens. -[2025-11-09 18:04:08,501][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 13.92%, Current % of VRAM taken: 55.42%, Block Peak % of device VRAM: 61.51%, ΔTime: 00:00:32 -[2025-11-09 18:04:09,234][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:04:09,236][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:04:09,237][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:04:10,387][__main__][INFO] - Iteration 3 took 48s (26.48% Gen, 71.13% Train). Generation: 12s, Training: 34s. Estimated remaining time: 40h 9m 59s. Estimated total time: 40h 15m 13s. Time estimates for 10 more iterations: 8m 3s, 100 more iterations: 1h 20m 30s, 500 more iterations: 6h 42m 32s. -[2025-11-09 18:04:10,389][__main__][INFO] - Starting iteration 3. -[2025-11-09 18:04:10,822][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 0 and human policies 1. -[2025-11-09 18:04:10,823][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:04:22,716][__main__][INFO] - Number of regex retries in iteration 3: 0 -[2025-11-09 18:04:22,717][__main__][INFO] - agents played in iteration 3 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:04:23,614][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.48%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:04:23,642][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.48%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:04:23,668][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.48%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:04:23,692][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.48%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:04:23,692][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:04:23,693][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:04:24,409][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:04:24,862][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:04:25,365][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:04:25,857][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:04:26,349][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:04:26,854][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:04:27,344][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:04:27,833][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:04:28,339][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:04:28,828][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:04:29,324][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:04:29,814][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:04:30,303][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:04:30,794][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:04:31,285][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:04:31,775][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:04:32,270][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:04:32,761][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:04:33,256][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:04:33,750][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:04:34,244][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:04:34,745][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:04:35,234][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:04:35,723][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:04:36,226][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:04:36,713][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:04:37,202][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:04:37,694][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:04:38,184][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:04:38,670][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:04:39,160][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:04:39,648][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:04:40,158][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:04:40,647][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:04:41,136][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:04:41,625][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:04:42,114][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:04:42,603][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:04:43,091][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:04:43,581][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:04:44,070][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:04:44,564][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:04:45,055][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:04:45,563][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:04:46,058][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:04:46,556][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:04:47,048][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:04:47,539][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:04:48,039][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:04:48,529][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:04:49,021][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:04:49,511][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:04:50,000][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:04:50,490][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:04:50,981][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:04:51,472][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:04:51,963][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:04:52,452][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:04:52,942][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:04:53,431][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:04:53,921][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:04:54,412][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:04:54,901][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:04:55,390][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:04:55,882][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9684 tokens. -[2025-11-09 18:04:56,524][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 13.99%, Current % of VRAM taken: 55.49%, Block Peak % of device VRAM: 61.69%, ΔTime: 00:00:32 -[2025-11-09 18:04:57,267][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:04:57,269][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:04:57,271][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:04:58,215][__main__][INFO] - Iteration 4 took 47s (25.09% Gen, 72.91% Train). Generation: 11s, Training: 34s. Estimated remaining time: 39h 23m 37s. Estimated total time: 39h 29m 39s. Time estimates for 10 more iterations: 7m 53s, 100 more iterations: 1h 18m 59s, 500 more iterations: 6h 34m 56s. -[2025-11-09 18:04:58,217][__main__][INFO] - Starting iteration 4. -[2025-11-09 18:04:58,641][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 0 and human policies 1. -[2025-11-09 18:04:58,642][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:05:11,798][__main__][INFO] - Number of regex retries in iteration 4: 0 -[2025-11-09 18:05:11,798][__main__][INFO] - agents played in iteration 4 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:05:12,687][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:05:12,713][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:05:12,740][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:05:12,763][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:05:12,764][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:05:12,765][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:05:13,484][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:05:13,951][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:05:14,454][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:05:14,947][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:05:15,439][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:05:15,929][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:05:16,420][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:05:16,911][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:05:17,402][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:05:17,892][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:05:18,380][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:05:18,869][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:05:19,375][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:05:19,865][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:05:20,355][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:05:20,846][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:05:21,334][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:05:21,824][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:05:22,314][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:05:22,803][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:05:23,294][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:05:23,784][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:05:24,275][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:05:24,766][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:05:25,259][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:05:25,764][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:05:26,255][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:05:26,744][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:05:27,233][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:05:27,724][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:05:28,215][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:05:28,713][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:05:29,230][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:05:29,721][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:05:30,212][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:05:30,706][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:05:31,199][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:05:31,691][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:05:32,183][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:05:32,673][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:05:33,164][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:05:33,656][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:05:34,149][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:05:34,640][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:05:35,130][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:05:35,622][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:05:36,112][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:05:36,604][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:05:37,099][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:05:37,591][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:05:38,081][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:05:38,572][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:05:39,063][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:05:39,552][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:05:40,044][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:05:40,539][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:05:41,030][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:05:41,522][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:05:42,011][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:05:42,502][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:05:42,997][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:05:43,486][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:05:43,974][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:05:44,465][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:05:44,959][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9682 tokens. -[2025-11-09 18:05:45,694][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.01%, Current % of VRAM taken: 55.51%, Block Peak % of device VRAM: 61.79%, ΔTime: 00:00:32 -[2025-11-09 18:05:46,449][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:05:46,450][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:05:46,452][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:05:47,414][__main__][INFO] - Iteration 5 took 48s (26.97% Gen, 71.05% Train). Generation: 13s, Training: 34s. Estimated remaining time: 40h 31m 49s. Estimated total time: 40h 38m 39s. Time estimates for 10 more iterations: 8m 7s, 100 more iterations: 1h 21m 17s, 500 more iterations: 6h 46m 26s. -[2025-11-09 18:05:47,417][__main__][INFO] - Starting iteration 5. -[2025-11-09 18:05:47,883][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 0 and human policies 1. -[2025-11-09 18:05:47,884][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:05:59,949][__main__][INFO] - Number of regex retries in iteration 5: 0 -[2025-11-09 18:05:59,950][__main__][INFO] - agents played in iteration 5 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:06:00,909][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:06:00,933][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:06:00,958][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:06:00,980][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:06:00,980][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:06:00,982][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:06:01,725][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:06:02,183][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:06:02,768][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:06:03,273][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:06:03,767][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:06:04,263][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:06:04,761][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:06:05,253][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:06:05,750][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:06:06,255][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:06:06,744][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:06:07,238][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:06:07,728][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:06:08,218][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:06:08,711][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:06:09,203][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:06:09,698][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:06:10,194][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:06:10,685][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:06:11,175][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:06:11,673][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:06:12,163][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:06:12,655][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:06:13,146][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:06:13,637][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:06:14,132][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:06:14,623][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:06:15,112][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:06:15,603][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:06:16,093][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:06:16,583][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:06:17,077][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:06:17,567][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:06:18,059][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:06:18,548][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:06:19,039][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:06:19,542][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:06:20,032][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:06:20,521][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:06:21,012][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:06:21,501][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:06:21,989][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:06:22,479][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:06:22,967][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:06:23,468][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:06:23,956][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:06:24,444][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:06:24,938][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:06:25,427][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:06:25,916][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:06:26,404][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:06:26,896][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:06:27,388][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:06:27,879][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:06:28,368][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:06:28,862][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:06:29,354][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:06:29,843][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:06:30,331][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:06:30,820][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:06:31,310][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:06:31,800][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:06:32,289][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:06:32,781][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:06:33,269][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9599 tokens. -[2025-11-09 18:06:33,926][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 13.93%, Current % of VRAM taken: 55.42%, Block Peak % of device VRAM: 61.60%, ΔTime: 00:00:32 -[2025-11-09 18:06:34,650][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:06:34,652][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:06:34,653][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:06:36,243][__main__][INFO] - Iteration 6 took 48s (24.95% Gen, 71.76% Train). Generation: 12s, Training: 34s. Estimated remaining time: 40h 10m 20s. Estimated total time: 40h 18m 0s. Time estimates for 10 more iterations: 8m 3s, 100 more iterations: 1h 20m 36s, 500 more iterations: 6h 43m 0s. -[2025-11-09 18:06:36,245][__main__][INFO] - Starting iteration 6. -[2025-11-09 18:06:36,681][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 0 and human policies 1. -[2025-11-09 18:06:36,681][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:06:48,317][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:06:49,132][__main__][INFO] - Number of regex retries in iteration 6: 1 -[2025-11-09 18:06:49,132][__main__][INFO] - agents played in iteration 6 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:06:49,990][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.50%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:06:50,018][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.50%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:06:50,044][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.50%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:06:50,067][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.50%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:06:50,067][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:06:50,068][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:06:50,811][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:06:51,282][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:06:51,785][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:06:52,278][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:06:52,771][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:06:53,262][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:06:53,772][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:06:54,262][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:06:54,753][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:06:55,248][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:06:55,738][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:06:56,227][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:06:56,715][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:06:57,205][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:06:57,696][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:06:58,184][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:06:58,674][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:06:59,167][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:06:59,657][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:07:00,147][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:07:00,647][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:07:01,137][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:07:01,629][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:07:02,123][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:07:02,613][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:07:03,102][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:07:03,591][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:07:04,080][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:07:04,569][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:07:05,058][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:07:05,571][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:07:06,061][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:07:06,549][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:07:07,041][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:07:07,529][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:07:08,018][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:07:08,506][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:07:08,995][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:07:09,486][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:07:09,974][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:07:10,463][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:07:10,955][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:07:11,444][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:07:11,932][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:07:12,422][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:07:12,911][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:07:13,404][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:07:13,893][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:07:14,386][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:07:14,881][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:07:15,371][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:07:15,862][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:07:16,352][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:07:16,843][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:07:17,336][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:07:17,826][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:07:18,315][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:07:18,805][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:07:19,293][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:07:19,782][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:07:20,270][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:07:20,759][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:07:21,252][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:07:21,741][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:07:22,230][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9626 tokens. -[2025-11-09 18:07:22,861][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 13.92%, Current % of VRAM taken: 55.41%, Block Peak % of device VRAM: 61.55%, ΔTime: 00:00:32 -[2025-11-09 18:07:23,580][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:07:23,582][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:07:23,583][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:07:24,539][__main__][INFO] - Iteration 7 took 47s (26.02% Gen, 71.99% Train). Generation: 12s, Training: 34s. Estimated remaining time: 39h 44m 28s. Estimated total time: 39h 52m 56s. Time estimates for 10 more iterations: 7m 58s, 100 more iterations: 1h 19m 45s, 500 more iterations: 6h 38m 49s. -[2025-11-09 18:07:24,541][__main__][INFO] - Starting iteration 7. -[2025-11-09 18:07:24,993][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 0 and human policies 1. -[2025-11-09 18:07:24,994][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:07:25,853][mllm.models.large_language_model_local][WARNING] - Response Proposal: 5 hats, 5 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:07:38,110][__main__][INFO] - Number of regex retries in iteration 7: 1 -[2025-11-09 18:07:38,111][__main__][INFO] - agents played in iteration 7 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:07:38,918][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.48%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:07:38,946][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.48%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:07:38,974][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.48%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:07:38,997][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.48%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:07:38,998][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:07:38,998][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:07:39,748][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:07:40,202][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:07:40,703][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:07:41,198][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:07:41,690][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:07:42,182][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:07:42,678][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:07:43,169][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:07:43,671][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:07:44,162][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:07:44,652][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:07:45,147][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:07:45,637][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:07:46,129][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:07:46,620][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:07:47,115][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:07:47,609][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:07:48,103][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:07:48,595][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:07:49,087][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:07:49,579][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:07:50,071][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:07:50,569][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:07:51,064][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:07:51,564][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:07:52,056][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:07:52,547][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:07:53,042][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:07:53,533][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:07:54,025][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:07:54,517][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:07:55,009][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:07:55,507][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:07:56,020][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:07:56,514][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:07:57,006][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:07:57,499][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:07:57,991][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:07:58,482][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:07:58,974][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:07:59,482][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:07:59,972][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:08:00,463][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:08:00,964][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:08:01,454][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:08:01,944][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:08:02,434][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:08:02,925][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:08:03,417][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:08:03,910][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:08:04,402][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:08:04,891][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:08:05,384][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:08:05,880][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:08:06,371][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:08:06,866][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:08:07,364][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:08:07,854][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:08:08,344][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:08:08,835][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:08:09,326][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:08:09,817][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:08:10,307][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:08:10,799][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:08:11,305][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9605 tokens. -[2025-11-09 18:08:11,912][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 13.95%, Current % of VRAM taken: 55.44%, Block Peak % of device VRAM: 61.56%, ΔTime: 00:00:32 -[2025-11-09 18:08:12,664][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:08:12,666][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:08:12,667][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:08:13,634][__main__][INFO] - Iteration 8 took 48s (26.97% Gen, 71.04% Train). Generation: 13s, Training: 34s. Estimated remaining time: 40h 22m 48s. Estimated total time: 40h 32m 4s. Time estimates for 10 more iterations: 8m 6s, 100 more iterations: 1h 21m 4s, 500 more iterations: 6h 45m 20s. -[2025-11-09 18:08:13,636][__main__][INFO] - Starting iteration 8. -[2025-11-09 18:08:14,070][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 0 and human policies 1. -[2025-11-09 18:08:14,071][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:08:21,874][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:08:26,935][__main__][INFO] - Number of regex retries in iteration 8: 1 -[2025-11-09 18:08:26,937][__main__][INFO] - agents played in iteration 8 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:08:27,795][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:08:27,824][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:08:27,850][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:08:27,875][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:08:27,876][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:08:27,876][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:08:28,602][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:08:29,056][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:08:29,553][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:08:30,047][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:08:30,542][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:08:31,037][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:08:31,526][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:08:32,014][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:08:32,503][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:08:32,994][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:08:33,484][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:08:33,976][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:08:34,488][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:08:34,999][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:08:35,490][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:08:35,983][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:08:36,498][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:08:36,990][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:08:37,481][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:08:37,984][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:08:38,479][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:08:38,971][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:08:39,464][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:08:39,953][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:08:40,443][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:08:40,931][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:08:41,421][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:08:41,912][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:08:42,402][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:08:42,901][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:08:43,391][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:08:43,881][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:08:44,371][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:08:44,861][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:08:45,351][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:08:45,842][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:08:46,331][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:08:46,821][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:08:47,311][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:08:47,799][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:08:48,304][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:08:48,795][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:08:49,285][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:08:49,776][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:08:50,268][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:08:50,764][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:08:51,255][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:08:51,750][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:08:52,243][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:08:52,731][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:08:53,222][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:08:53,716][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:08:54,206][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:08:54,715][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:08:55,204][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:08:55,693][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:08:56,185][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:08:56,675][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:08:57,164][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:08:57,654][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:08:58,143][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:08:58,634][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:08:59,125][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:08:59,615][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:09:00,105][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9670 tokens. -[2025-11-09 18:09:00,798][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.03%, Current % of VRAM taken: 55.53%, Block Peak % of device VRAM: 61.65%, ΔTime: 00:00:32 -[2025-11-09 18:09:01,554][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:09:01,556][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:09:01,557][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:09:02,660][__main__][INFO] - Iteration 9 took 48s (26.48% Gen, 71.25% Train). Generation: 12s, Training: 34s. Estimated remaining time: 40h 19m 26s. Estimated total time: 40h 29m 31s. Time estimates for 10 more iterations: 8m 5s, 100 more iterations: 1h 20m 59s, 500 more iterations: 6h 44m 55s. -[2025-11-09 18:09:02,662][__main__][INFO] - Starting iteration 9. -[2025-11-09 18:09:03,133][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 0 and human policies 1. -[2025-11-09 18:09:03,133][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:09:04,784][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:09:15,978][__main__][INFO] - Number of regex retries in iteration 9: 1 -[2025-11-09 18:09:15,979][__main__][INFO] - agents played in iteration 9 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:09:16,862][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:09:16,888][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:09:16,914][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:09:16,937][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:09:16,937][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:09:16,939][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:09:17,610][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:09:18,066][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:09:18,570][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:09:19,065][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:09:19,562][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:09:20,055][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:09:20,548][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:09:21,041][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:09:21,532][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:09:22,024][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:09:22,515][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:09:23,007][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:09:23,516][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:09:24,007][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:09:24,497][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:09:24,993][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:09:25,484][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:09:25,974][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:09:26,464][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:09:26,957][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:09:27,451][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:09:27,941][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:09:28,431][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:09:28,929][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:09:29,418][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:09:29,907][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:09:30,397][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:09:30,886][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:09:31,375][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:09:31,866][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:09:32,357][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:09:32,848][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:09:33,337][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:09:33,831][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:09:34,323][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:09:34,814][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:09:35,314][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:09:35,804][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:09:36,296][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:09:36,805][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:09:37,296][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:09:37,787][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:09:38,277][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:09:38,768][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:09:39,264][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:09:39,756][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:09:40,249][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:09:40,743][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:09:41,235][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:09:41,727][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:09:42,218][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:09:42,709][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:09:43,218][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:09:43,710][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:09:44,206][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:09:44,701][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:09:45,196][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:09:45,688][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:09:46,178][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:09:46,666][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:09:47,158][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:09:47,647][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:09:48,138][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:09:48,627][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:09:49,116][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9660 tokens. -[2025-11-09 18:09:49,778][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 13.95%, Current % of VRAM taken: 55.44%, Block Peak % of device VRAM: 61.66%, ΔTime: 00:00:32 -[2025-11-09 18:09:50,505][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:09:50,506][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:09:50,508][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:09:51,573][__main__][INFO] - Iteration 10 took 48s (26.52% Gen, 71.28% Train). Generation: 12s, Training: 34s. Estimated remaining time: 40h 11m 6s. Estimated total time: 40h 22m 1s. Time estimates for 10 more iterations: 8m 4s, 100 more iterations: 1h 20m 44s, 500 more iterations: 6h 43m 40s. -[2025-11-09 18:09:51,575][__main__][INFO] - Starting iteration 10. -[2025-11-09 18:09:52,032][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 0 and human policies 1. -[2025-11-09 18:09:52,032][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:09:58,297][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls +Given the values, you have a much higher value for hats compared to Bob, making hats the most desirable item for you. Books and balls have significantly lower values for you compared to their values for Bob, so it's best to focus solely on hats to maximize your points. did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 21:08:33,881][__main__][INFO] - Number of regex retries in iteration 0: 1 +[2026-03-25 21:08:33,882][__main__][INFO] - agents played in iteration 0 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:08:44,482][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 37.45%, Block Peak % of device VRAM: 18.68%, ΔTime: 00:00:00 +[2026-03-25 21:08:44,506][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 37.45%, Block Peak % of device VRAM: 18.68%, ΔTime: 00:00:00 +[2026-03-25 21:08:44,528][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 37.45%, Block Peak % of device VRAM: 18.68%, ΔTime: 00:00:00 +[2026-03-25 21:08:44,550][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 37.45%, Block Peak % of device VRAM: 18.68%, ΔTime: 00:00:00 +[2026-03-25 21:08:44,551][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:08:44,551][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:08:45,112][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:08:46,219][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:08:46,719][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:08:47,223][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:08:47,719][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:08:48,213][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:08:48,706][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:08:49,200][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:08:49,694][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:08:50,187][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:08:50,680][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:08:51,172][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:08:51,663][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:08:52,155][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:08:52,648][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:08:53,140][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:08:53,631][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:08:54,124][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:08:54,619][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:08:55,110][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:08:55,603][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:08:56,093][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:08:56,582][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:08:57,073][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:08:57,565][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:08:58,057][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:08:58,552][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:08:59,045][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:08:59,541][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:09:00,032][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:09:00,526][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:09:01,017][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:09:01,511][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:09:02,004][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:09:02,498][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:09:02,988][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:09:03,480][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:09:03,970][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:09:04,461][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:09:04,952][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:09:05,442][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:09:05,933][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:09:06,429][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:09:06,921][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:09:07,414][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:09:07,907][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:09:08,399][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:09:08,890][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:09:09,382][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:09:09,873][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:09:10,364][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:09:10,856][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:09:11,346][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:09:11,836][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:09:12,327][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:09:12,818][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:09:13,311][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:09:13,804][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:09:14,296][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:09:14,788][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:09:15,281][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:09:15,772][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:09:16,265][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:09:16,755][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:09:17,245][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9576 tokens. +[2026-03-25 21:09:18,143][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 15.93%, Current % of VRAM taken: 53.38%, Block Peak % of device VRAM: 60.89%, ΔTime: 00:00:33 +[2026-03-25 21:09:18,776][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:09:18,779][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:09:18,780][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:09:19,440][__main__][INFO] - Iteration 1 took 58s (22.74% Gen, 76.14% Train). Generation: 13s, Training: 44s. Estimated remaining time: 49h 4m 48s. Estimated total time: 49h 8m 27s. Time estimates for 10 more iterations: 9m 49s, 100 more iterations: 1h 38m 16s, 500 more iterations: 8h 11m 24s. +[2026-03-25 21:09:19,443][__main__][INFO] - Starting iteration 1. +[2026-03-25 21:09:19,846][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 0 and human policies 1. +[2026-03-25 21:09:19,846][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:09:33,036][__main__][INFO] - Number of regex retries in iteration 1: 0 +[2026-03-25 21:09:33,037][__main__][INFO] - agents played in iteration 1 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:09:33,858][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 46.44%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:09:33,881][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 46.44%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:09:33,903][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 46.44%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:09:33,925][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 46.44%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:09:33,926][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:09:33,927][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:09:34,501][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:09:34,949][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:09:35,445][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:09:35,937][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:09:36,430][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:09:36,922][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:09:37,413][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:09:37,904][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:09:38,397][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:09:38,889][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:09:39,385][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:09:39,879][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:09:40,371][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:09:40,866][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:09:41,359][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:09:41,851][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:09:42,343][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:09:42,839][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:09:43,331][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:09:43,824][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:09:44,319][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:09:44,813][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:09:45,304][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:09:45,796][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:09:46,288][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:09:46,785][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:09:47,279][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:09:47,770][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:09:48,264][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:09:48,757][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:09:49,251][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:09:49,745][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:09:50,238][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:09:50,731][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:09:51,226][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:09:51,720][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:09:52,212][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:09:52,705][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:09:53,201][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:09:53,693][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:09:54,185][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:09:54,676][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:09:55,168][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:09:55,660][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:09:56,152][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:09:56,644][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:09:57,136][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:09:57,628][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:09:58,122][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:09:58,612][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:09:59,104][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:09:59,595][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:10:00,089][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:10:00,579][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:10:01,071][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:10:01,562][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:10:02,055][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:10:02,546][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:10:03,038][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:10:03,528][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:10:04,020][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:10:04,511][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:10:05,004][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:10:05,497][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:10:05,988][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9613 tokens. +[2026-03-25 21:10:06,677][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 13.95%, Current % of VRAM taken: 60.37%, Block Peak % of device VRAM: 61.44%, ΔTime: 00:00:32 +[2026-03-25 21:10:07,368][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:10:07,370][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:10:07,372][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:10:08,016][__main__][INFO] - Iteration 2 took 48s (27.38% Gen, 71.28% Train). Generation: 13s, Training: 34s. Estimated remaining time: 40h 4m 5s. Estimated total time: 40h 8m 33s. Time estimates for 10 more iterations: 8m 1s, 100 more iterations: 1h 20m 17s, 500 more iterations: 6h 41m 25s. +[2026-03-25 21:10:08,019][__main__][INFO] - Starting iteration 2. +[2026-03-25 21:10:08,420][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 0 and human policies 1. +[2026-03-25 21:10:08,421][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:10:17,193][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls -Given Bob's per-item values, he will be more interested in books and balls, while I have a balanced interest in hats, books, and balls. Since the hats are worth significantly more to me than to Bob, I should propose keeping all 10 hats. This way, I ensure I get the full value of the hats, which are the most valuable to me. I propose giving away all books and balls since they are not as valuable to me compared to the hats. did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:10:05,277][__main__][INFO] - Number of regex retries in iteration 10: 1 -[2025-11-09 18:10:05,278][__main__][INFO] - agents played in iteration 10 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:10:06,154][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.50%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:10:06,183][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.50%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:10:06,210][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.50%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:10:06,234][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.50%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:10:06,235][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:10:06,236][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:10:06,936][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:10:07,389][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:10:07,890][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:10:08,382][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:10:08,872][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:10:09,366][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:10:09,861][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:10:10,350][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:10:10,840][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:10:11,330][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:10:11,821][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:10:12,311][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:10:12,806][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:10:13,297][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:10:13,802][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:10:14,296][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:10:14,787][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:10:15,277][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:10:15,767][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:10:16,256][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:10:16,748][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:10:17,240][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:10:17,728][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:10:18,218][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:10:18,706][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:10:19,197][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:10:19,688][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:10:20,182][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:10:20,671][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:10:21,162][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:10:21,652][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:10:22,143][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:10:22,633][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:10:23,125][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:10:23,617][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:10:24,120][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:10:24,610][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:10:25,101][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:10:25,592][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:10:26,084][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:10:26,574][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:10:27,065][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:10:27,555][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:10:28,047][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:10:28,537][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:10:29,027][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:10:29,523][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:10:30,014][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:10:30,519][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:10:31,009][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:10:31,499][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:10:31,989][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:10:32,480][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:10:32,971][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:10:33,462][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:10:33,951][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:10:34,443][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:10:34,931][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:10:35,422][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:10:35,911][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:10:36,400][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:10:36,888][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:10:37,377][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:10:37,870][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:10:38,366][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9678 tokens. -[2025-11-09 18:10:39,024][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 13.88%, Current % of VRAM taken: 55.38%, Block Peak % of device VRAM: 61.48%, ΔTime: 00:00:32 -[2025-11-09 18:10:39,785][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:10:39,787][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:10:39,788][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:10:42,366][__main__][INFO] - Iteration 11 took 50s (26.32% Gen, 68.56% Train). Generation: 13s, Training: 34s. Estimated remaining time: 41h 45m 0s. Estimated total time: 41h 56m 46s. Time estimates for 10 more iterations: 8m 23s, 100 more iterations: 1h 23m 53s, 500 more iterations: 6h 59m 27s. -[2025-11-09 18:10:42,369][__main__][INFO] - Starting iteration 11. -[2025-11-09 18:10:42,867][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 1 and human policies 1. -[2025-11-09 18:10:42,868][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:10:45,948][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:10:56,448][__main__][INFO] - Number of regex retries in iteration 11: 1 -[2025-11-09 18:10:56,448][__main__][INFO] - agents played in iteration 11 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:10:57,347][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.46%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:10:57,375][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.46%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:10:57,402][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.46%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:10:57,426][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.46%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:10:57,426][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:10:57,427][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:10:58,104][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:10:58,560][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:10:59,063][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:10:59,568][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:11:00,063][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:11:00,554][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:11:01,049][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:11:01,545][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:11:02,038][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:11:02,534][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:11:03,026][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:11:03,530][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:11:04,023][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:11:04,515][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:11:05,015][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:11:05,504][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:11:05,996][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:11:06,487][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:11:06,978][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:11:07,475][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:11:07,967][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:11:08,459][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:11:08,967][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:11:09,461][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:11:09,952][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:11:10,445][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:11:10,938][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:11:11,448][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:11:11,945][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:11:12,439][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:11:12,937][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:11:13,429][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:11:13,921][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:11:14,413][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:11:14,904][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:11:15,399][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:11:15,889][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:11:16,382][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:11:16,875][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:11:17,366][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:11:17,865][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:11:18,356][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:11:18,847][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:11:19,355][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:11:19,844][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:11:20,334][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:11:20,831][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:11:21,320][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:11:21,811][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:11:22,302][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:11:22,793][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:11:23,286][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:11:23,777][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:11:24,268][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:11:24,759][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:11:25,250][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:11:25,753][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:11:26,243][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:11:26,732][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:11:27,223][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:11:27,712][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:11:28,202][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:11:28,691][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:11:29,182][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:11:29,681][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9694 tokens. -[2025-11-09 18:11:30,392][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 13.96%, Current % of VRAM taken: 55.45%, Block Peak % of device VRAM: 61.67%, ΔTime: 00:00:32 -[2025-11-09 18:11:31,128][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:11:31,129][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:11:31,131][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:11:32,082][__main__][INFO] - Iteration 12 took 49s (27.59% Gen, 70.47% Train). Generation: 13s, Training: 34s. Estimated remaining time: 40h 48m 10s. Estimated total time: 41h 0m 45s. Time estimates for 10 more iterations: 8m 12s, 100 more iterations: 1h 22m 1s, 500 more iterations: 6h 50m 7s. -[2025-11-09 18:11:32,090][__main__][INFO] - Starting iteration 12. -[2025-11-09 18:11:32,568][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 1 and human policies 1. -[2025-11-09 18:11:32,568][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:11:33,596][mllm.models.large_language_model_local][WARNING] - Response Proposal: 5 hats, 0 books, 5 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:11:45,150][__main__][INFO] - Number of regex retries in iteration 12: 1 -[2025-11-09 18:11:45,151][__main__][INFO] - agents played in iteration 12 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:11:45,944][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:11:45,967][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:11:45,989][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:11:46,011][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:11:46,011][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:11:46,012][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:11:46,631][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:11:47,086][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:11:47,587][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:11:48,082][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:11:48,576][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:11:49,070][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:11:49,572][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:11:50,065][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:11:50,555][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:11:51,051][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:11:51,542][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:11:52,032][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:11:52,527][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:11:53,019][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:11:53,512][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:11:54,003][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:11:54,493][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:11:54,985][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:11:55,477][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:11:55,974][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:11:56,465][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:11:56,955][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:11:57,464][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:11:57,955][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:11:58,445][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:11:58,948][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:11:59,440][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:11:59,930][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:12:00,422][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:12:00,913][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:12:01,408][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:12:01,898][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:12:02,389][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:12:02,879][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:12:03,370][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:12:03,871][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:12:04,360][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:12:04,849][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:12:05,339][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:12:05,828][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:12:06,318][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:12:06,808][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:12:07,297][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:12:07,790][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:12:08,280][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:12:08,769][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:12:09,261][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:12:09,753][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:12:10,242][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:12:10,730][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:12:11,219][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:12:11,708][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:12:12,197][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:12:12,688][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:12:13,177][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:12:13,667][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:12:14,155][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:12:14,645][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:12:15,135][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:12:15,626][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:12:16,115][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:12:16,605][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:12:17,097][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:12:17,586][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:12:18,078][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9674 tokens. -[2025-11-09 18:12:18,704][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 13.97%, Current % of VRAM taken: 55.47%, Block Peak % of device VRAM: 61.60%, ΔTime: 00:00:32 -[2025-11-09 18:12:19,428][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:12:19,430][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:12:19,431][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:12:20,331][__main__][INFO] - Iteration 13 took 47s (26.34% Gen, 71.77% Train). Generation: 12s, Training: 34s. Estimated remaining time: 39h 34m 47s. Estimated total time: 39h 48m 11s. Time estimates for 10 more iterations: 7m 57s, 100 more iterations: 1h 19m 36s, 500 more iterations: 6h 38m 1s. -[2025-11-09 18:12:20,333][__main__][INFO] - Starting iteration 13. -[2025-11-09 18:12:20,782][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 1 and human policies 1. -[2025-11-09 18:12:20,783][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:12:34,023][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:12:34,840][__main__][INFO] - Number of regex retries in iteration 13: 1 -[2025-11-09 18:12:34,840][__main__][INFO] - agents played in iteration 13 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:12:35,661][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.49%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:12:35,686][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.49%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:12:35,711][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.49%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:12:35,734][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.49%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:12:35,735][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:12:35,736][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:12:36,349][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:12:36,802][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:12:37,308][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:12:37,810][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:12:38,309][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:12:38,806][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:12:39,302][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:12:39,793][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:12:40,285][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:12:40,777][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:12:41,268][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:12:41,759][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:12:42,249][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:12:42,741][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:12:43,233][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:12:43,724][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:12:44,219][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:12:44,714][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:12:45,203][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:12:45,694][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:12:46,184][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:12:46,673][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:12:47,183][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:12:47,674][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:12:48,163][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:12:48,662][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:12:49,154][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:12:49,648][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:12:50,140][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:12:50,633][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:12:51,128][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:12:51,619][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:12:52,111][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:12:52,603][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:12:53,093][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:12:53,584][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:12:54,076][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:12:54,566][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:12:55,069][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:12:55,562][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:12:56,051][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:12:56,557][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:12:57,050][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:12:57,544][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:12:58,038][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:12:58,531][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:12:59,026][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:12:59,518][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:13:00,010][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:13:00,502][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:13:00,992][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:13:01,484][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:13:01,977][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:13:02,467][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:13:02,957][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:13:03,446][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:13:03,937][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:13:04,430][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:13:04,932][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:13:05,425][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:13:05,915][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:13:06,407][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:13:06,896][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:13:07,387][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:13:07,878][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9663 tokens. -[2025-11-09 18:13:08,508][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.04%, Current % of VRAM taken: 55.54%, Block Peak % of device VRAM: 61.83%, ΔTime: 00:00:32 -[2025-11-09 18:13:09,219][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:13:09,220][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:13:09,222][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:13:10,184][__main__][INFO] - Iteration 14 took 49s (28.45% Gen, 69.60% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 55m 52s. Estimated total time: 41h 10m 5s. Time estimates for 10 more iterations: 8m 14s, 100 more iterations: 1h 22m 20s, 500 more iterations: 6h 51m 40s. -[2025-11-09 18:13:10,186][__main__][INFO] - Starting iteration 14. -[2025-11-09 18:13:10,636][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 1 and human policies 1. -[2025-11-09 18:13:10,636][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:13:11,884][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:13:24,608][__main__][INFO] - Number of regex retries in iteration 14: 1 -[2025-11-09 18:13:24,609][__main__][INFO] - agents played in iteration 14 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:13:25,458][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:13:25,484][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:13:25,510][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:13:25,532][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:13:25,533][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:13:25,534][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:13:26,212][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:13:26,683][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:13:27,181][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:13:27,682][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:13:28,179][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:13:28,675][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:13:29,182][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:13:29,833][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:13:30,334][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:13:30,830][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:13:31,322][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:13:31,819][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:13:32,312][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:13:32,806][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:13:33,302][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:13:33,798][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:13:34,292][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:13:34,792][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:13:35,284][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:13:35,782][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:13:36,271][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:13:36,762][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:13:37,261][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:13:37,750][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:13:38,241][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:13:38,731][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:13:39,222][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:13:39,717][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:13:40,206][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:13:40,697][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:13:41,187][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:13:41,678][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:13:42,181][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:13:42,671][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:13:43,162][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:13:43,671][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:13:44,163][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:13:44,653][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:13:45,157][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:13:45,647][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:13:46,137][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:13:46,629][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:13:47,123][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:13:47,618][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:13:48,144][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:13:48,636][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:13:49,131][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:13:49,625][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:13:50,121][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:13:50,617][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:13:51,109][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:13:51,608][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:13:52,099][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:13:52,594][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:13:53,089][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:13:53,584][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:13:54,094][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:13:54,585][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:13:55,076][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:13:55,570][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:13:56,060][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:13:56,550][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:13:57,041][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:13:57,531][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:13:58,024][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9744 tokens. -[2025-11-09 18:13:58,638][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 13.97%, Current % of VRAM taken: 55.47%, Block Peak % of device VRAM: 61.83%, ΔTime: 00:00:32 -[2025-11-09 18:13:59,373][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:13:59,375][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:13:59,376][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:14:00,365][__main__][INFO] - Iteration 15 took 49s (28.10% Gen, 69.91% Train). Generation: 13s, Training: 34s. Estimated remaining time: 41h 11m 25s. Estimated total time: 41h 26m 28s. Time estimates for 10 more iterations: 8m 17s, 100 more iterations: 1h 22m 52s, 500 more iterations: 6h 54m 24s. -[2025-11-09 18:14:00,367][__main__][INFO] - Starting iteration 15. -[2025-11-09 18:14:00,829][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 1 and human policies 1. -[2025-11-09 18:14:00,830][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:14:12,516][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 20 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:14:15,285][__main__][INFO] - Number of regex retries in iteration 15: 1 -[2025-11-09 18:14:15,285][__main__][INFO] - agents played in iteration 15 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:14:16,125][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:14:16,150][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:14:16,173][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:14:16,196][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:14:16,196][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:14:16,197][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:14:16,831][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:14:17,283][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:14:17,790][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:14:18,291][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:14:18,800][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:14:19,294][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:14:19,787][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:14:20,298][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:14:20,790][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:14:21,285][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:14:21,777][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:14:22,267][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:14:22,760][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:14:23,255][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:14:23,748][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:14:24,243][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:14:24,735][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:14:25,226][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:14:25,715][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:14:26,206][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:14:26,703][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:14:27,194][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:14:27,686][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:14:28,181][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:14:28,671][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:14:29,163][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:14:29,652][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:14:30,145][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:14:30,637][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:14:31,127][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:14:31,618][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:14:32,110][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:14:32,600][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:14:33,092][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:14:33,589][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:14:34,080][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:14:34,577][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:14:35,068][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:14:35,560][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:14:36,053][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:14:36,545][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:14:37,035][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:14:37,528][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:14:38,019][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:14:38,511][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:14:39,002][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:14:39,495][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:14:40,002][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:14:40,494][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:14:40,986][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:14:41,483][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:14:41,974][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:14:42,465][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:14:42,955][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:14:43,446][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:14:43,942][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:14:44,433][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:14:44,923][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:14:45,413][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:14:45,904][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:14:46,393][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:14:46,883][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:14:47,373][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:14:47,865][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:14:48,357][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9721 tokens. -[2025-11-09 18:14:48,990][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 13.96%, Current % of VRAM taken: 55.46%, Block Peak % of device VRAM: 61.70%, ΔTime: 00:00:32 -[2025-11-09 18:14:49,699][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:14:49,700][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:14:49,702][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:14:50,659][__main__][INFO] - Iteration 16 took 49s (29.01% Gen, 69.07% Train). Generation: 14s, Training: 34s. Estimated remaining time: 41h 15m 37s. Estimated total time: 41h 31m 31s. Time estimates for 10 more iterations: 8m 18s, 100 more iterations: 1h 23m 3s, 500 more iterations: 6h 55m 15s. -[2025-11-09 18:14:50,661][__main__][INFO] - Starting iteration 16. -[2025-11-09 18:14:51,118][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 1 and human policies 1. -[2025-11-09 18:14:51,119][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:14:54,192][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:14:56,793][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls +Given the per-item values, the strategic approach is to maximize the quantity of items with higher personal value. Here, I have the highest value for hats (10), but Alice values books (10) and balls (10) more than I do. By proposing to take all 10 hats, I ensure I get the maximum possible value from the items I prefer the most, while minimizing the risk of items being allocated proportionally if my proposal exceeds the item quantity. Since I have the highest value for hats, I aim to secure as many hats as possible to maximize my points in this round. did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 21:10:22,130][__main__][INFO] - Number of regex retries in iteration 2: 1 +[2026-03-25 21:10:22,131][__main__][INFO] - agents played in iteration 2 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:10:22,938][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.42%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:10:22,964][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.42%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:10:22,988][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.42%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:10:23,011][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.42%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:10:23,012][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:10:23,013][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:10:23,624][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:10:24,075][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:10:24,570][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:10:25,062][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:10:25,554][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:10:26,048][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:10:26,542][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:10:27,033][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:10:27,524][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:10:28,015][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:10:28,505][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:10:28,998][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:10:29,489][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:10:29,980][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:10:30,471][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:10:30,962][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:10:31,453][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:10:31,947][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:10:32,442][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:10:32,936][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:10:33,432][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:10:33,925][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:10:34,437][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:10:34,934][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:10:35,429][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:10:35,922][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:10:36,416][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:10:36,909][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:10:37,401][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:10:37,892][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:10:38,383][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:10:38,874][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:10:39,366][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:10:39,859][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:10:40,349][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:10:40,839][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:10:41,330][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:10:41,822][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:10:42,315][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:10:42,806][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:10:43,298][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:10:43,789][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:10:44,282][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:10:44,774][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:10:45,266][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:10:45,759][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:10:46,251][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:10:46,743][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:10:47,236][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:10:47,731][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:10:48,224][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:10:48,717][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:10:49,210][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:10:49,702][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:10:50,192][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:10:50,684][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:10:51,176][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:10:51,668][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:10:52,159][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:10:52,650][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:10:53,142][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:10:53,633][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:10:54,124][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:10:54,614][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:10:55,105][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9620 tokens. +[2026-03-25 21:10:55,787][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 12.95%, Current % of VRAM taken: 59.38%, Block Peak % of device VRAM: 61.46%, ΔTime: 00:00:32 +[2026-03-25 21:10:56,479][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:10:56,481][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:10:56,482][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:10:57,111][__main__][INFO] - Iteration 3 took 48s (28.16% Gen, 70.55% Train). Generation: 13s, Training: 34s. Estimated remaining time: 40h 29m 17s. Estimated total time: 40h 34m 34s. Time estimates for 10 more iterations: 8m 6s, 100 more iterations: 1h 21m 9s, 500 more iterations: 6h 45m 45s. +[2026-03-25 21:10:57,114][__main__][INFO] - Starting iteration 3. +[2026-03-25 21:10:57,513][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 0 and human policies 1. +[2026-03-25 21:10:57,513][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:10:58,168][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 21:11:06,251][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 21:11:10,385][__main__][INFO] - Number of regex retries in iteration 3: 2 +[2026-03-25 21:11:10,386][__main__][INFO] - agents played in iteration 3 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:11:11,227][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 52.43%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:11:11,250][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 52.43%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:11:11,273][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 52.43%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:11:11,296][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 52.43%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:11:11,296][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:11:11,297][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:11:11,888][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:11:12,336][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:11:12,832][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:11:13,325][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:11:13,818][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:11:14,310][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:11:14,803][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:11:15,294][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:11:15,787][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:11:16,281][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:11:16,776][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:11:17,270][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:11:17,763][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:11:18,256][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:11:18,748][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:11:19,241][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:11:19,735][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:11:20,229][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:11:20,723][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:11:21,239][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:11:21,736][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:11:22,227][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:11:22,721][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:11:23,213][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:11:23,707][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:11:24,198][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:11:24,689][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:11:25,180][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:11:25,672][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:11:26,164][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:11:26,658][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:11:27,150][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:11:27,642][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:11:28,133][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:11:28,625][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:11:29,116][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:11:29,609][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:11:30,100][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:11:30,594][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:11:31,089][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:11:31,580][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:11:32,075][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:11:32,569][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:11:33,065][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:11:33,561][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:11:34,057][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:11:34,550][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:11:35,044][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:11:35,540][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:11:36,033][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:11:36,525][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:11:37,020][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:11:37,513][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:11:38,005][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:11:38,497][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:11:38,991][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:11:39,484][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:11:39,977][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:11:40,469][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:11:40,962][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:11:41,453][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:11:41,946][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:11:42,439][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:11:42,932][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:11:43,424][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9666 tokens. +[2026-03-25 21:11:44,099][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 13.91%, Current % of VRAM taken: 60.33%, Block Peak % of device VRAM: 61.69%, ΔTime: 00:00:32 +[2026-03-25 21:11:44,792][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:11:44,794][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:11:44,796][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:11:45,433][__main__][INFO] - Iteration 4 took 47s (26.86% Gen, 71.81% Train). Generation: 12s, Training: 34s. Estimated remaining time: 39h 49m 56s. Estimated total time: 39h 56m 1s. Time estimates for 10 more iterations: 7m 59s, 100 more iterations: 1h 19m 52s, 500 more iterations: 6h 39m 20s. +[2026-03-25 21:11:45,435][__main__][INFO] - Starting iteration 4. +[2026-03-25 21:11:45,833][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 0 and human policies 1. +[2026-03-25 21:11:45,834][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:11:58,507][__main__][INFO] - Number of regex retries in iteration 4: 0 +[2026-03-25 21:11:58,508][__main__][INFO] - agents played in iteration 4 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:11:59,316][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.41%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:11:59,339][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.41%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:11:59,362][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.41%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:11:59,385][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.41%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:11:59,385][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:11:59,386][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:11:59,980][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:12:00,433][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:12:00,930][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:12:01,426][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:12:01,923][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:12:02,419][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:12:02,913][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:12:03,405][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:12:03,897][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:12:04,388][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:12:04,879][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:12:05,369][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:12:05,863][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:12:06,357][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:12:06,852][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:12:07,343][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:12:07,835][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:12:08,328][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:12:08,823][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:12:09,319][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:12:09,812][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:12:10,305][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:12:10,799][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:12:11,291][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:12:11,787][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:12:12,280][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:12:12,776][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:12:13,273][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:12:13,768][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:12:14,261][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:12:14,755][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:12:15,247][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:12:15,741][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:12:16,233][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:12:16,728][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:12:17,224][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:12:17,720][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:12:18,214][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:12:18,708][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:12:19,203][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:12:19,699][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:12:20,193][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:12:20,686][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:12:21,180][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:12:21,673][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:12:22,166][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:12:22,658][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:12:23,150][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:12:23,643][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:12:24,136][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:12:24,628][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:12:25,122][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:12:25,617][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:12:26,109][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:12:26,604][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:12:27,096][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:12:27,588][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:12:28,082][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:12:28,573][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:12:29,067][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:12:29,561][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:12:30,053][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:12:30,546][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:12:31,039][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:12:31,531][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9683 tokens. +[2026-03-25 21:12:32,219][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 13.94%, Current % of VRAM taken: 60.36%, Block Peak % of device VRAM: 61.63%, ΔTime: 00:00:32 +[2026-03-25 21:12:32,913][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:12:32,916][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:12:32,917][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:12:33,552][__main__][INFO] - Iteration 5 took 47s (26.56% Gen, 72.11% Train). Generation: 12s, Training: 34s. Estimated remaining time: 39h 39m 3s. Estimated total time: 39h 45m 56s. Time estimates for 10 more iterations: 7m 57s, 100 more iterations: 1h 19m 31s, 500 more iterations: 6h 37m 39s. +[2026-03-25 21:12:33,554][__main__][INFO] - Starting iteration 5. +[2026-03-25 21:12:33,955][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 0 and human policies 1. +[2026-03-25 21:12:33,955][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:12:35,373][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 21:12:41,438][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls -Explanation: Given the values, hats are highly valuable to me and moderately valued by Bob. Books are moderately valuable to me and highly valued by Bob. Balls are least valued by both of us. By taking all the hats, I maximize my utility given my high valuation of them, while minimizing the risk of over-allocating in other less valuable categories. did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:15:05,806][__main__][INFO] - Number of regex retries in iteration 16: 2 -[2025-11-09 18:15:05,806][__main__][INFO] - agents played in iteration 16 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:15:06,598][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.50%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:15:06,625][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.50%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:15:06,650][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.50%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:15:06,673][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.50%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:15:06,674][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:15:06,674][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:15:07,366][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:15:07,825][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:15:08,327][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:15:08,827][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:15:09,317][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:15:09,807][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:15:10,298][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:15:10,787][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:15:11,280][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:15:11,774][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:15:12,266][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:15:12,762][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:15:13,254][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:15:13,746][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:15:14,245][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:15:14,737][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:15:15,227][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:15:15,718][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:15:16,210][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:15:16,704][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:15:17,193][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:15:17,685][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:15:18,176][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:15:18,666][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:15:19,156][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:15:19,648][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:15:20,141][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:15:20,634][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:15:21,125][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:15:21,616][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:15:22,129][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:15:22,619][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:15:23,109][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:15:23,604][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:15:24,093][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:15:24,583][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:15:25,073][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:15:25,565][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:15:26,066][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:15:26,554][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:15:27,046][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:15:27,541][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:15:28,035][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:15:28,527][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:15:29,019][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:15:29,509][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:15:30,015][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:15:30,505][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:15:30,995][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:15:31,494][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:15:31,985][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:15:32,480][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:15:32,984][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:15:33,474][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:15:33,968][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:15:34,459][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:15:34,951][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:15:35,444][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:15:35,934][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:15:36,428][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:15:36,919][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:15:37,408][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:15:37,913][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:15:38,404][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:15:38,893][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9609 tokens. -[2025-11-09 18:15:39,515][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 13.93%, Current % of VRAM taken: 55.43%, Block Peak % of device VRAM: 61.48%, ΔTime: 00:00:32 -[2025-11-09 18:15:40,229][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:15:40,231][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:15:40,232][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:15:41,162][__main__][INFO] - Iteration 17 took 50s (29.35% Gen, 68.79% Train). Generation: 14s, Training: 34s. Estimated remaining time: 41h 25m 29s. Estimated total time: 41h 42m 13s. Time estimates for 10 more iterations: 8m 20s, 100 more iterations: 1h 23m 24s, 500 more iterations: 6h 57m 2s. -[2025-11-09 18:15:41,164][__main__][INFO] - Starting iteration 17. -[2025-11-09 18:15:41,616][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 1 and human policies 1. -[2025-11-09 18:15:41,617][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:15:42,890][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:15:43,854][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:15:46,439][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:15:55,846][__main__][INFO] - Number of regex retries in iteration 17: 3 -[2025-11-09 18:15:55,847][__main__][INFO] - agents played in iteration 17 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:15:56,642][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.48%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:15:56,665][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.48%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:15:56,687][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.48%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:15:56,709][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.48%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:15:56,709][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:15:56,710][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:15:57,372][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:15:57,839][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:15:58,345][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:15:58,842][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:15:59,340][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:15:59,834][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:16:00,329][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:16:00,821][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:16:01,313][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:16:01,824][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:16:02,316][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:16:02,808][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:16:03,303][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:16:03,794][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:16:04,286][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:16:04,779][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:16:05,270][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:16:05,765][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:16:06,257][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:16:06,747][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:16:07,239][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:16:07,731][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:16:08,229][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:16:08,722][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:16:09,214][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:16:09,721][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:16:10,215][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:16:10,708][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:16:11,205][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:16:11,695][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:16:12,187][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:16:12,676][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:16:13,166][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:16:13,662][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:16:14,153][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:16:14,646][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:16:15,144][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:16:15,636][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:16:16,128][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:16:16,620][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:16:17,112][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:16:17,603][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:16:18,096][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:16:18,586][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:16:19,086][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:16:19,577][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:16:20,068][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:16:20,558][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:16:21,048][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:16:21,543][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:16:22,034][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:16:22,524][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:16:23,016][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:16:23,507][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:16:23,998][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:16:24,489][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:16:24,980][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:16:25,473][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:16:25,966][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:16:26,457][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:16:26,949][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:16:27,440][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:16:27,934][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:16:28,426][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:16:28,917][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9679 tokens. -[2025-11-09 18:16:29,539][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 13.98%, Current % of VRAM taken: 55.48%, Block Peak % of device VRAM: 61.68%, ΔTime: 00:00:32 -[2025-11-09 18:16:30,269][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:16:30,270][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:16:30,272][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:16:31,637][__main__][INFO] - Iteration 18 took 50s (28.45% Gen, 68.82% Train). Generation: 14s, Training: 34s. Estimated remaining time: 41h 23m 28s. Estimated total time: 41h 41m 3s. Time estimates for 10 more iterations: 8m 20s, 100 more iterations: 1h 23m 22s, 500 more iterations: 6h 56m 50s. -[2025-11-09 18:16:31,641][__main__][INFO] - Starting iteration 18. -[2025-11-09 18:16:32,103][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 1 and human policies 1. -[2025-11-09 18:16:32,104][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:16:46,150][__main__][INFO] - Number of regex retries in iteration 18: 0 -[2025-11-09 18:16:46,150][__main__][INFO] - agents played in iteration 18 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:16:47,143][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:16:47,172][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:16:47,199][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:16:47,224][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:16:47,224][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:16:47,226][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:16:47,947][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:16:48,405][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:16:48,905][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:16:49,410][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:16:49,905][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:16:50,397][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:16:50,888][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:16:51,380][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:16:51,871][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:16:52,365][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:16:52,859][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:16:53,353][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:16:53,845][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:16:54,337][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:16:54,829][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:16:55,323][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:16:55,832][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:16:56,324][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:16:56,814][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:16:57,311][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:16:57,805][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:16:58,297][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:16:58,788][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:16:59,280][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:16:59,772][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:17:00,263][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:17:00,755][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:17:01,248][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:17:01,740][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:17:02,235][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:17:02,726][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:17:03,216][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:17:03,725][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:17:04,221][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:17:04,714][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:17:05,205][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:17:05,700][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:17:06,200][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:17:06,692][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:17:07,188][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:17:07,680][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:17:08,171][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:17:08,663][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:17:09,152][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:17:09,645][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:17:10,137][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:17:10,627][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:17:11,116][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:17:11,607][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:17:12,098][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:17:12,594][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:17:13,095][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:17:13,587][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:17:14,085][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:17:14,577][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:17:15,070][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:17:15,566][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:17:16,058][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:17:16,552][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:17:17,045][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:17:17,539][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:17:18,040][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:17:18,532][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:17:19,025][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:17:19,522][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9686 tokens. -[2025-11-09 18:17:20,222][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.03%, Current % of VRAM taken: 55.52%, Block Peak % of device VRAM: 61.70%, ΔTime: 00:00:32 -[2025-11-09 18:17:20,980][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:17:20,981][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:17:20,983][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:17:21,972][__main__][INFO] - Iteration 19 took 49s (28.17% Gen, 69.85% Train). Generation: 14s, Training: 34s. Estimated remaining time: 41h 15m 3s. Estimated total time: 41h 33m 28s. Time estimates for 10 more iterations: 8m 18s, 100 more iterations: 1h 23m 6s, 500 more iterations: 6h 55m 34s. -[2025-11-09 18:17:21,974][__main__][INFO] - Starting iteration 19. -[2025-11-09 18:17:22,491][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 1 and human policies 1. -[2025-11-09 18:17:22,492][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:17:29,987][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls +Given the per-item values, you have a high value for hats (10) while Bob values books significantly higher (10). To maximize your points in this round, it's best to take all the hats since they are your highest-value items and you are certain to receive them if you claim all of them. BOB values books highly, so taking books might lead to a lower allocation if there is competition. Similarly, balls are low-value for both of you, so it's better to avoid claiming them to focus on higher-value items. did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 21:12:47,125][__main__][INFO] - Number of regex retries in iteration 5: 2 +[2026-03-25 21:12:47,126][__main__][INFO] - agents played in iteration 5 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:12:47,977][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.42%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:12:48,000][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.42%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:12:48,022][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.42%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:12:48,045][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.42%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:12:48,045][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:12:48,046][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:12:48,650][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:12:49,103][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:12:49,600][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:12:50,096][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:12:50,600][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:12:51,092][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:12:51,586][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:12:52,081][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:12:52,574][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:12:53,067][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:12:53,560][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:12:54,052][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:12:54,544][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:12:55,036][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:12:55,531][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:12:56,024][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:12:56,518][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:12:57,010][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:12:57,503][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:12:57,995][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:12:58,488][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:12:58,984][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:12:59,477][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:12:59,968][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:13:00,461][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:13:00,953][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:13:01,445][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:13:01,940][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:13:02,433][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:13:02,925][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:13:03,417][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:13:03,909][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:13:04,403][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:13:04,893][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:13:05,384][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:13:05,873][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:13:06,366][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:13:06,856][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:13:07,347][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:13:07,838][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:13:08,329][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:13:08,820][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:13:09,310][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:13:09,801][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:13:10,290][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:13:10,781][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:13:11,271][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:13:11,768][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:13:12,260][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:13:12,752][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:13:13,245][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:13:13,739][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:13:14,231][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:13:14,725][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:13:15,219][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:13:15,712][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:13:16,206][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:13:16,698][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:13:17,190][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:13:17,681][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:13:18,172][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:13:18,665][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:13:19,157][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:13:19,648][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:13:20,141][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9610 tokens. +[2026-03-25 21:13:20,826][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 13.95%, Current % of VRAM taken: 60.37%, Block Peak % of device VRAM: 61.88%, ΔTime: 00:00:32 +[2026-03-25 21:13:24,029][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:13:24,031][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:13:24,033][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:13:24,667][__main__][INFO] - Iteration 6 took 50s (25.97% Gen, 72.78% Train). Generation: 13s, Training: 36s. Estimated remaining time: 42h 7m 54s. Estimated total time: 42h 15m 38s. Time estimates for 10 more iterations: 8m 27s, 100 more iterations: 1h 24m 31s, 500 more iterations: 7h 2m 36s. +[2026-03-25 21:13:24,669][__main__][INFO] - Starting iteration 6. +[2026-03-25 21:13:25,069][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 0 and human policies 1. +[2026-03-25 21:13:25,070][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:13:37,668][__main__][INFO] - Number of regex retries in iteration 6: 0 +[2026-03-25 21:13:37,669][__main__][INFO] - agents played in iteration 6 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:13:38,497][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.42%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:13:38,520][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.42%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:13:38,543][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.42%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:13:38,565][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.42%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:13:38,566][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:13:38,567][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:13:39,156][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:13:39,607][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:13:40,106][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:13:40,601][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:13:41,096][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:13:41,588][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:13:42,081][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:13:42,576][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:13:43,070][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:13:43,563][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:13:44,057][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:13:44,549][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:13:45,042][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:13:45,535][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:13:46,028][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:13:46,520][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:13:47,017][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:13:47,510][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:13:48,004][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:13:48,499][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:13:48,991][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:13:49,484][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:13:49,977][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:13:50,470][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:13:50,965][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:13:51,459][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:13:51,953][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:13:52,446][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:13:52,940][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:13:53,432][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:13:53,925][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:13:54,419][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:13:54,912][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:13:55,406][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:13:55,901][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:13:56,393][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:13:56,887][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:13:57,379][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:13:57,872][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:13:58,363][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:13:58,855][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:13:59,347][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:13:59,839][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:14:00,332][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:14:00,824][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:14:01,321][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:14:01,815][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:14:02,309][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:14:02,802][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:14:03,295][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:14:03,790][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:14:04,283][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:14:04,776][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:14:05,271][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:14:05,765][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:14:06,261][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:14:06,754][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:14:07,246][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:14:07,737][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:14:08,229][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:14:08,722][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:14:09,213][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:14:09,705][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:14:10,196][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:14:10,689][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9655 tokens. +[2026-03-25 21:14:11,370][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 13.95%, Current % of VRAM taken: 60.38%, Block Peak % of device VRAM: 61.60%, ΔTime: 00:00:32 +[2026-03-25 21:14:12,072][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:14:12,074][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:14:12,075][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:14:12,706][__main__][INFO] - Iteration 7 took 47s (26.45% Gen, 72.23% Train). Generation: 12s, Training: 34s. Estimated remaining time: 39h 33m 18s. Estimated total time: 39h 41m 50s. Time estimates for 10 more iterations: 7m 56s, 100 more iterations: 1h 19m 23s, 500 more iterations: 6h 36m 58s. +[2026-03-25 21:14:12,708][__main__][INFO] - Starting iteration 7. +[2026-03-25 21:14:13,111][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 0 and human policies 1. +[2026-03-25 21:14:13,111][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:14:25,752][__main__][INFO] - Number of regex retries in iteration 7: 0 +[2026-03-25 21:14:25,753][__main__][INFO] - agents played in iteration 7 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:14:26,551][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.41%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:14:26,574][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.41%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:14:26,597][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.41%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:14:26,620][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.41%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:14:26,620][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:14:26,621][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:14:27,220][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:14:27,673][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:14:28,168][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:14:28,660][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:14:29,151][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:14:29,644][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:14:30,136][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:14:30,627][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:14:31,119][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:14:31,610][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:14:32,103][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:14:32,595][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:14:33,086][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:14:33,578][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:14:34,069][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:14:34,561][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:14:35,052][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:14:35,543][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:14:36,038][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:14:36,530][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:14:37,022][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:14:37,514][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:14:38,008][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:14:38,501][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:14:38,994][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:14:39,485][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:14:39,979][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:14:40,472][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:14:40,964][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:14:41,457][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:14:41,949][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:14:42,443][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:14:42,937][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:14:43,431][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:14:43,925][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:14:44,419][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:14:44,912][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:14:45,403][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:14:45,897][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:14:46,389][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:14:46,881][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:14:47,372][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:14:47,864][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:14:48,355][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:14:48,845][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:14:49,337][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:14:49,828][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:14:50,321][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:14:50,813][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:14:51,305][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:14:51,798][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:14:52,290][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:14:52,785][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:14:53,279][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:14:53,770][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:14:54,265][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:14:54,759][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:14:55,252][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:14:55,745][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:14:56,238][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:14:56,730][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:14:57,222][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:14:57,713][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:14:58,205][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:14:58,697][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9603 tokens. +[2026-03-25 21:14:59,335][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 13.93%, Current % of VRAM taken: 60.36%, Block Peak % of device VRAM: 61.48%, ΔTime: 00:00:32 +[2026-03-25 21:15:00,041][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:15:00,044][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:15:00,045][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:15:00,686][__main__][INFO] - Iteration 8 took 47s (26.57% Gen, 72.08% Train). Generation: 12s, Training: 34s. Estimated remaining time: 39h 29m 28s. Estimated total time: 39h 38m 48s. Time estimates for 10 more iterations: 7m 55s, 100 more iterations: 1h 19m 17s, 500 more iterations: 6h 36m 28s. +[2026-03-25 21:15:00,691][__main__][INFO] - Starting iteration 8. +[2026-03-25 21:15:01,091][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 0 and human policies 1. +[2026-03-25 21:15:01,092][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:15:13,448][__main__][INFO] - Number of regex retries in iteration 8: 0 +[2026-03-25 21:15:13,449][__main__][INFO] - agents played in iteration 8 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:15:14,257][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.40%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:15:14,280][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.40%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:15:14,303][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.40%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:15:14,326][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.40%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:15:14,327][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:15:14,327][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:15:14,958][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:15:15,408][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:15:15,905][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:15:16,399][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:15:16,890][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:15:17,381][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:15:17,872][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:15:18,363][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:15:18,854][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:15:19,349][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:15:19,844][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:15:20,337][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:15:20,830][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:15:21,326][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:15:21,821][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:15:22,314][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:15:22,806][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:15:23,302][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:15:23,795][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:15:24,303][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:15:24,799][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:15:25,291][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:15:25,785][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:15:26,277][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:15:26,771][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:15:27,263][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:15:27,756][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:15:28,247][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:15:28,738][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:15:29,230][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:15:29,723][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:15:30,217][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:15:30,709][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:15:31,202][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:15:31,694][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:15:32,187][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:15:32,680][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:15:33,173][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:15:33,665][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:15:34,159][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:15:34,652][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:15:35,144][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:15:35,638][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:15:36,134][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:15:36,627][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:15:37,120][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:15:37,613][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:15:38,107][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:15:38,601][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:15:39,093][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:15:39,586][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:15:40,078][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:15:40,571][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:15:41,064][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:15:41,556][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:15:42,048][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:15:42,541][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:15:43,034][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:15:43,527][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:15:44,023][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:15:44,518][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:15:45,012][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:15:45,505][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:15:45,999][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:15:46,492][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9668 tokens. +[2026-03-25 21:15:47,133][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 13.99%, Current % of VRAM taken: 60.41%, Block Peak % of device VRAM: 61.59%, ΔTime: 00:00:32 +[2026-03-25 21:15:47,833][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:15:47,835][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:15:47,837][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:15:48,462][__main__][INFO] - Iteration 9 took 47s (26.09% Gen, 72.59% Train). Generation: 12s, Training: 34s. Estimated remaining time: 39h 18m 24s. Estimated total time: 39h 28m 32s. Time estimates for 10 more iterations: 7m 53s, 100 more iterations: 1h 18m 57s, 500 more iterations: 6h 34m 45s. +[2026-03-25 21:15:48,464][__main__][INFO] - Starting iteration 9. +[2026-03-25 21:15:48,865][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 0 and human policies 1. +[2026-03-25 21:15:48,865][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:16:01,274][__main__][INFO] - Number of regex retries in iteration 9: 0 +[2026-03-25 21:16:01,275][__main__][INFO] - agents played in iteration 9 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:16:02,083][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.43%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:16:02,107][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.43%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:16:02,131][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.43%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:16:02,153][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.43%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:16:02,154][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:16:02,155][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:16:02,751][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:16:03,202][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:16:03,703][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:16:04,199][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:16:04,694][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:16:05,188][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:16:05,681][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:16:06,174][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:16:06,668][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:16:07,160][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:16:07,653][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:16:08,144][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:16:08,637][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:16:09,130][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:16:09,624][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:16:10,119][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:16:10,613][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:16:11,106][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:16:11,602][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:16:12,095][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:16:12,587][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:16:13,082][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:16:13,573][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:16:14,064][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:16:14,557][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:16:15,051][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:16:15,543][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:16:16,035][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:16:16,528][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:16:17,021][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:16:17,512][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:16:18,004][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:16:18,496][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:16:18,990][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:16:19,483][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:16:19,976][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:16:20,469][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:16:20,966][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:16:21,460][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:16:21,954][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:16:22,448][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:16:22,943][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:16:23,438][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:16:23,934][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:16:24,428][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:16:24,923][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:16:25,417][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:16:25,911][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:16:26,405][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:16:26,900][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:16:27,393][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:16:27,888][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:16:28,384][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:16:28,879][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:16:29,374][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:16:29,868][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:16:30,364][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:16:30,857][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:16:31,348][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:16:31,842][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:16:32,336][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:16:32,830][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:16:33,323][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:16:33,816][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:16:34,308][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9677 tokens. +[2026-03-25 21:16:34,930][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 13.95%, Current % of VRAM taken: 60.38%, Block Peak % of device VRAM: 61.68%, ΔTime: 00:00:32 +[2026-03-25 21:16:35,629][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:16:35,632][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:16:35,634][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:16:36,276][__main__][INFO] - Iteration 10 took 47s (26.17% Gen, 72.47% Train). Generation: 12s, Training: 34s. Estimated remaining time: 39h 19m 38s. Estimated total time: 39h 30m 34s. Time estimates for 10 more iterations: 7m 54s, 100 more iterations: 1h 19m 1s, 500 more iterations: 6h 35m 5s. +[2026-03-25 21:16:36,278][__main__][INFO] - Starting iteration 10. +[2026-03-25 21:16:36,678][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 0 and human policies 1. +[2026-03-25 21:16:36,679][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:16:37,987][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 21:16:49,254][__main__][INFO] - Number of regex retries in iteration 10: 1 +[2026-03-25 21:16:49,255][__main__][INFO] - agents played in iteration 10 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:16:50,051][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.44%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:16:50,075][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.44%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:16:50,098][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.44%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:16:50,121][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.44%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:16:50,122][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:16:50,122][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:16:50,722][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:16:51,172][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:16:51,669][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:16:52,162][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:16:52,655][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:16:53,146][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:16:53,637][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:16:54,130][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:16:54,623][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:16:55,117][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:16:55,611][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:16:56,104][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:16:56,597][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:16:57,090][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:16:57,586][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:16:58,080][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:16:58,574][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:16:59,065][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:16:59,557][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:17:00,048][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:17:00,541][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:17:01,033][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:17:01,525][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:17:02,019][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:17:02,511][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:17:03,004][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:17:03,495][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:17:03,986][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:17:04,479][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:17:04,971][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:17:05,464][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:17:05,956][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:17:06,448][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:17:06,940][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:17:07,433][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:17:07,926][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:17:08,420][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:17:08,912][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:17:09,406][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:17:09,899][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:17:10,391][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:17:10,882][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:17:11,374][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:17:11,865][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:17:12,359][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:17:12,874][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:17:13,365][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:17:13,858][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:17:14,350][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:17:14,843][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:17:15,336][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:17:15,829][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:17:16,323][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:17:16,816][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:17:17,309][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:17:17,802][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:17:18,295][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:17:18,785][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:17:19,277][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:17:19,768][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:17:20,261][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:17:20,752][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:17:21,244][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:17:21,741][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:17:22,234][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9649 tokens. +[2026-03-25 21:17:22,899][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 12.95%, Current % of VRAM taken: 59.38%, Block Peak % of device VRAM: 61.47%, ΔTime: 00:00:32 +[2026-03-25 21:17:23,608][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:17:23,610][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:17:23,612][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:17:24,832][__main__][INFO] - Iteration 11 took 48s (26.12% Gen, 71.35% Train). Generation: 12s, Training: 34s. Estimated remaining time: 39h 55m 57s. Estimated total time: 40h 7m 41s. Time estimates for 10 more iterations: 8m 1s, 100 more iterations: 1h 20m 15s, 500 more iterations: 6h 41m 16s. +[2026-03-25 21:17:24,834][__main__][INFO] - Starting iteration 11. +[2026-03-25 21:17:25,233][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 1 and human policies 1. +[2026-03-25 21:17:25,233][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:17:38,389][__main__][INFO] - Number of regex retries in iteration 11: 0 +[2026-03-25 21:17:38,390][__main__][INFO] - agents played in iteration 11 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:17:39,206][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 52.44%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:17:39,229][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 52.44%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:17:39,253][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 52.44%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:17:39,276][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 52.44%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:17:39,276][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:17:39,277][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:17:39,875][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:17:40,329][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:17:40,832][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:17:41,329][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:17:41,826][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:17:42,325][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:17:42,821][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:17:43,316][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:17:43,811][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:17:44,305][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:17:44,799][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:17:45,292][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:17:45,785][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:17:46,277][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:17:46,770][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:17:47,262][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:17:47,755][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:17:48,248][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:17:48,742][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:17:49,236][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:17:49,730][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:17:50,226][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:17:50,723][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:17:51,219][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:17:51,713][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:17:52,208][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:17:52,703][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:17:53,198][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:17:53,692][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:17:54,186][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:17:54,681][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:17:55,175][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:17:55,668][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:17:56,161][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:17:56,653][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:17:57,145][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:17:57,639][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:17:58,132][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:17:58,627][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:17:59,123][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:17:59,619][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:18:00,114][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:18:00,608][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:18:01,102][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:18:01,595][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:18:02,090][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:18:02,586][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:18:03,079][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:18:03,573][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:18:04,065][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:18:04,558][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:18:05,051][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:18:05,545][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:18:06,037][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:18:06,535][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:18:07,030][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:18:07,525][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:18:08,019][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:18:08,511][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:18:09,003][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:18:09,497][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:18:09,990][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:18:10,481][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:18:10,973][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:18:11,464][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9676 tokens. +[2026-03-25 21:18:12,147][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 13.90%, Current % of VRAM taken: 60.33%, Block Peak % of device VRAM: 61.59%, ΔTime: 00:00:32 +[2026-03-25 21:18:12,907][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:18:12,910][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:18:12,911][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:18:13,558][__main__][INFO] - Iteration 12 took 48s (27.23% Gen, 71.43% Train). Generation: 13s, Training: 34s. Estimated remaining time: 40h 3m 44s. Estimated total time: 40h 16m 17s. Time estimates for 10 more iterations: 8m 3s, 100 more iterations: 1h 20m 32s, 500 more iterations: 6h 42m 42s. +[2026-03-25 21:18:13,561][__main__][INFO] - Starting iteration 12. +[2026-03-25 21:18:13,961][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 1 and human policies 1. +[2026-03-25 21:18:13,962][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:18:15,049][mllm.models.large_language_model_local][WARNING] - Response Proposition: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 21:18:15,051][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 21:18:26,688][__main__][INFO] - Number of regex retries in iteration 12: 2 +[2026-03-25 21:18:26,690][__main__][INFO] - agents played in iteration 12 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:18:27,506][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.42%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:18:27,530][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.42%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:18:27,553][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.42%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:18:27,577][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.42%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:18:27,577][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:18:27,578][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:18:28,229][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:18:28,681][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:18:29,178][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:18:29,672][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:18:30,165][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:18:30,661][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:18:31,154][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:18:31,646][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:18:32,143][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:18:32,639][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:18:33,133][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:18:33,627][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:18:34,122][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:18:34,614][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:18:35,111][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:18:35,606][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:18:36,100][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:18:36,627][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:18:37,126][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:18:37,622][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:18:38,119][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:18:38,614][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:18:39,108][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:18:39,604][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:18:40,098][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:18:40,591][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:18:41,086][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:18:41,580][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:18:42,073][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:18:42,568][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:18:43,062][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:18:43,557][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:18:44,050][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:18:44,544][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:18:45,037][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:18:45,529][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:18:46,023][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:18:46,517][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:18:47,010][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:18:47,502][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:18:47,994][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:18:48,485][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:18:48,977][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:18:49,469][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:18:49,963][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:18:50,456][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:18:50,949][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:18:51,443][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:18:51,935][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:18:52,428][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:18:52,919][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:18:53,411][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:18:53,904][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:18:54,395][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:18:54,886][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:18:55,377][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:18:55,869][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:18:56,362][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:18:56,854][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:18:57,347][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:18:57,841][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:18:58,334][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:18:58,827][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:18:59,321][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:18:59,814][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9687 tokens. +[2026-03-25 21:19:00,494][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.00%, Current % of VRAM taken: 60.43%, Block Peak % of device VRAM: 61.71%, ΔTime: 00:00:32 +[2026-03-25 21:19:01,197][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:19:01,200][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:19:01,201][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:19:01,845][__main__][INFO] - Iteration 13 took 47s (26.58% Gen, 72.07% Train). Generation: 12s, Training: 34s. Estimated remaining time: 39h 40m 53s. Estimated total time: 39h 54m 14s. Time estimates for 10 more iterations: 7m 58s, 100 more iterations: 1h 19m 48s, 500 more iterations: 6h 39m 2s. +[2026-03-25 21:19:01,848][__main__][INFO] - Starting iteration 13. +[2026-03-25 21:19:02,247][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 1 and human policies 1. +[2026-03-25 21:19:02,248][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:19:15,823][__main__][INFO] - Number of regex retries in iteration 13: 0 +[2026-03-25 21:19:15,824][__main__][INFO] - agents played in iteration 13 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:19:16,635][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:19:16,659][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:19:16,683][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:19:16,707][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:19:16,708][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:19:16,709][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:19:17,331][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:19:17,781][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:19:18,283][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:19:18,779][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:19:19,272][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:19:19,767][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:19:20,264][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:19:20,760][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:19:21,254][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:19:21,747][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:19:22,241][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:19:22,735][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:19:23,230][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:19:23,727][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:19:24,224][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:19:24,718][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:19:25,212][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:19:25,705][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:19:26,198][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:19:26,689][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:19:27,182][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:19:27,675][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:19:28,167][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:19:28,660][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:19:29,155][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:19:29,650][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:19:30,143][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:19:30,637][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:19:31,131][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:19:31,624][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:19:32,120][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:19:32,614][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:19:33,106][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:19:33,599][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:19:34,091][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:19:34,583][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:19:35,073][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:19:35,565][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:19:36,056][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:19:36,551][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:19:37,043][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:19:37,540][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:19:38,034][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:19:38,528][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:19:39,023][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:19:39,519][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:19:40,014][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:19:40,508][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:19:41,003][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:19:41,496][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:19:41,989][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:19:42,484][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:19:42,977][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:19:43,471][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:19:43,965][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:19:44,457][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:19:44,952][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:19:45,446][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:19:45,943][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:19:46,439][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:19:46,934][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:19:47,428][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:19:47,924][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:19:48,421][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:19:48,915][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9686 tokens. +[2026-03-25 21:19:49,607][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.04%, Current % of VRAM taken: 60.47%, Block Peak % of device VRAM: 61.68%, ΔTime: 00:00:32 +[2026-03-25 21:19:50,312][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:19:50,314][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:19:50,316][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:19:50,956][__main__][INFO] - Iteration 14 took 48s (27.87% Gen, 70.81% Train). Generation: 13s, Training: 34s. Estimated remaining time: 40h 21m 18s. Estimated total time: 40h 35m 28s. Time estimates for 10 more iterations: 8m 7s, 100 more iterations: 1h 21m 10s, 500 more iterations: 6h 45m 54s. +[2026-03-25 21:19:50,959][__main__][INFO] - Starting iteration 14. +[2026-03-25 21:19:51,359][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 1 and human policies 1. +[2026-03-25 21:19:51,360][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:19:54,313][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 21:19:59,748][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls -Given the per-item values, you have a significantly higher value for hats compared to Bob, while Bob values books much more than you do. By proposing to take all the hats, you are maximizing your potential points from this round since you have a higher value for these items. Taking all books would not be optimal as they are valued very low by you and only slightly by Bob. Similarly, balls are of no significant interest to either of you, so proposing to take them would not be a good strategy. did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:17:36,869][__main__][INFO] - Number of regex retries in iteration 19: 1 -[2025-11-09 18:17:36,870][__main__][INFO] - agents played in iteration 19 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:17:37,903][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:17:37,931][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:17:37,959][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:17:37,982][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:17:37,983][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:17:37,984][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:17:38,644][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:17:39,115][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:17:39,614][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:17:40,110][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:17:40,604][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:17:41,098][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:17:41,597][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:17:42,087][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:17:42,578][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:17:43,074][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:17:43,567][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:17:44,090][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:17:44,586][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:17:45,079][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:17:45,573][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:17:46,067][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:17:46,564][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:17:47,055][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:17:47,546][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:17:48,043][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:17:48,533][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:17:49,024][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:17:49,515][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:17:50,005][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:17:50,494][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:17:50,986][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:17:51,476][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:17:51,969][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:17:52,459][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:17:52,949][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:17:53,441][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:17:53,933][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:17:54,424][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:17:54,912][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:17:55,402][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:17:55,893][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:17:56,383][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:17:56,874][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:17:57,366][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:17:57,856][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:17:58,346][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:17:58,838][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:17:59,330][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:17:59,836][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:18:00,330][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:18:00,823][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:18:01,322][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:18:01,812][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:18:02,305][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:18:02,794][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:18:03,285][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:18:03,780][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:18:04,270][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:18:04,764][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:18:05,252][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:18:05,744][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:18:06,247][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:18:06,738][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:18:07,230][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:18:07,739][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:18:08,231][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:18:08,721][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:18:09,224][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:18:09,715][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:18:10,221][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9780 tokens. -[2025-11-09 18:18:10,914][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.00%, Current % of VRAM taken: 55.50%, Block Peak % of device VRAM: 61.79%, ΔTime: 00:00:32 -[2025-11-09 18:18:11,671][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:18:11,673][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:18:11,674][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:18:12,600][__main__][INFO] - Iteration 20 took 50s (28.69% Gen, 69.46% Train). Generation: 14s, Training: 34s. Estimated remaining time: 41h 26m 12s. Estimated total time: 41h 45m 28s. Time estimates for 10 more iterations: 8m 21s, 100 more iterations: 1h 23m 30s, 500 more iterations: 6h 57m 34s. -[2025-11-09 18:18:12,603][__main__][INFO] - Starting iteration 20. -[2025-11-09 18:18:13,095][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 1 and human policies 1. -[2025-11-09 18:18:13,096][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:18:26,615][__main__][INFO] - Number of regex retries in iteration 20: 0 -[2025-11-09 18:18:26,615][__main__][INFO] - agents played in iteration 20 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:18:27,572][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:18:27,600][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:18:27,627][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:18:27,651][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:18:27,652][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:18:27,653][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:18:28,335][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:18:28,792][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:18:29,294][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:18:29,789][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:18:30,282][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:18:30,778][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:18:31,273][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:18:31,768][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:18:32,268][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:18:32,760][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:18:33,251][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:18:33,759][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:18:34,250][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:18:34,743][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:18:35,241][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:18:35,731][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:18:36,236][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:18:36,727][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:18:37,217][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:18:37,709][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:18:38,199][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:18:38,691][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:18:39,181][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:18:39,672][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:18:40,175][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:18:40,669][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:18:41,160][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:18:41,653][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:18:42,146][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:18:42,637][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:18:43,131][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:18:43,623][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:18:44,116][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:18:44,608][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:18:45,100][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:18:45,595][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:18:46,089][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:18:46,585][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:18:47,076][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:18:47,568][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:18:48,061][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:18:48,553][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:18:49,044][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:18:49,536][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:18:50,026][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:18:50,517][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:18:51,006][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:18:51,496][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:18:51,989][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:18:52,479][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:18:52,968][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:18:53,459][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:18:53,950][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:18:54,445][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:18:54,935][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:18:55,425][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:18:55,917][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:18:56,408][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:18:56,900][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:18:57,391][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:18:57,883][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:18:58,376][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:18:58,865][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:18:59,359][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:18:59,850][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9716 tokens. -[2025-11-09 18:19:00,527][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 13.97%, Current % of VRAM taken: 55.47%, Block Peak % of device VRAM: 61.69%, ΔTime: 00:00:32 -[2025-11-09 18:19:01,269][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:19:01,271][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:19:01,272][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:19:03,236][__main__][INFO] - Iteration 21 took 50s (26.96% Gen, 69.12% Train). Generation: 13s, Training: 34s. Estimated remaining time: 41h 26m 58s. Estimated total time: 41h 47m 5s. Time estimates for 10 more iterations: 8m 21s, 100 more iterations: 1h 23m 34s, 500 more iterations: 6h 57m 50s. -[2025-11-09 18:19:03,238][__main__][INFO] - Starting iteration 21. -[2025-11-09 18:19:03,755][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 2 and human policies 1. -[2025-11-09 18:19:03,756][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:19:17,298][__main__][INFO] - Number of regex retries in iteration 21: 0 -[2025-11-09 18:19:17,298][__main__][INFO] - agents played in iteration 21 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:19:18,092][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:19:18,120][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:19:18,148][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:19:18,173][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:19:18,173][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:19:18,174][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:19:18,857][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:19:19,313][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:19:19,817][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:19:20,317][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:19:20,812][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:19:21,304][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:19:21,797][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:19:22,288][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:19:22,781][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:19:23,290][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:19:23,783][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:19:24,278][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:19:24,769][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:19:25,262][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:19:25,754][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:19:26,246][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:19:26,738][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:19:27,234][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:19:27,726][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:19:28,215][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:19:28,705][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:19:29,201][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:19:29,694][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:19:30,186][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:19:30,677][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:19:31,169][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:19:31,659][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:19:32,150][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:19:32,656][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:19:33,147][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:19:33,638][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:19:34,131][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:19:34,621][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:19:35,113][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:19:35,603][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:19:36,093][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:19:36,584][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:19:37,075][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:19:37,577][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:19:38,066][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:19:38,556][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:19:39,051][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:19:39,543][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:19:40,034][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:19:40,526][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:19:41,018][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:19:41,509][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:19:42,002][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:19:42,493][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:19:42,984][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:19:43,472][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:19:43,963][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:19:44,471][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:19:44,961][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:19:45,450][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:19:45,940][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:19:46,432][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:19:46,924][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:19:47,414][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:19:47,904][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:19:48,395][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:19:48,885][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:19:49,375][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:19:49,868][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:19:50,360][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9722 tokens. -[2025-11-09 18:19:51,039][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 13.98%, Current % of VRAM taken: 55.47%, Block Peak % of device VRAM: 61.61%, ΔTime: 00:00:32 -[2025-11-09 18:19:51,771][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:19:51,773][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:19:51,776][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:19:52,788][__main__][INFO] - Iteration 22 took 49s (27.62% Gen, 70.31% Train). Generation: 13s, Training: 34s. Estimated remaining time: 40h 30m 46s. Estimated total time: 40h 51m 41s. Time estimates for 10 more iterations: 8m 10s, 100 more iterations: 1h 21m 43s, 500 more iterations: 6h 48m 36s. -[2025-11-09 18:19:52,790][__main__][INFO] - Starting iteration 22. -[2025-11-09 18:19:53,258][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 2 and human policies 1. -[2025-11-09 18:19:53,258][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:20:07,078][__main__][INFO] - Number of regex retries in iteration 22: 0 -[2025-11-09 18:20:07,078][__main__][INFO] - agents played in iteration 22 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:20:07,872][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:20:07,908][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:20:07,940][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:20:07,966][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:20:07,967][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:20:07,968][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:20:08,629][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:20:09,081][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:20:09,605][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:20:10,102][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:20:10,602][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:20:11,093][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:20:11,589][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:20:12,081][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:20:12,576][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:20:13,073][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:20:13,572][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:20:14,067][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:20:14,563][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:20:15,056][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:20:15,549][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:20:16,045][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:20:16,536][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:20:17,028][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:20:17,520][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:20:18,011][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:20:18,501][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:20:18,992][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:20:19,482][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:20:19,995][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:20:20,489][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:20:20,981][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:20:21,488][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:20:21,979][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:20:22,471][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:20:22,963][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:20:23,455][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:20:23,945][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:20:24,435][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:20:24,926][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:20:25,419][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:20:25,911][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:20:26,404][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:20:26,897][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:20:27,389][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:20:27,894][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:20:28,386][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:20:28,878][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:20:29,377][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:20:29,870][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:20:30,370][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:20:30,864][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:20:31,357][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:20:31,851][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:20:32,344][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:20:32,836][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:20:33,327][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:20:33,819][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:20:34,310][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:20:34,802][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:20:35,294][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:20:35,786][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:20:36,278][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:20:36,770][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:20:37,262][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:20:37,755][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:20:38,246][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:20:38,737][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:20:39,251][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:20:39,758][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:20:40,258][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9787 tokens. -[2025-11-09 18:20:40,937][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 55.56%, Block Peak % of device VRAM: 61.76%, ΔTime: 00:00:32 -[2025-11-09 18:20:41,670][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:20:41,672][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:20:41,675][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:20:42,612][__main__][INFO] - Iteration 23 took 49s (28.00% Gen, 70.10% Train). Generation: 13s, Training: 34s. Estimated remaining time: 40h 45m 59s. Estimated total time: 41h 7m 44s. Time estimates for 10 more iterations: 8m 13s, 100 more iterations: 1h 22m 15s, 500 more iterations: 6h 51m 17s. -[2025-11-09 18:20:42,614][__main__][INFO] - Starting iteration 23. -[2025-11-09 18:20:43,084][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 2 and human policies 1. -[2025-11-09 18:20:43,084][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:20:51,803][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:20:56,915][__main__][INFO] - Number of regex retries in iteration 23: 1 -[2025-11-09 18:20:56,916][__main__][INFO] - agents played in iteration 23 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:20:57,732][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:20:57,761][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:20:57,788][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:20:57,821][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:20:57,822][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:20:57,823][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:20:58,481][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:20:58,937][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:20:59,439][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:20:59,935][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:21:00,429][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:21:00,923][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:21:01,416][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:21:01,912][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:21:02,406][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:21:02,895][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:21:03,387][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:21:03,878][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:21:04,371][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:21:04,875][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:21:05,368][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:21:05,860][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:21:06,355][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:21:06,847][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:21:07,343][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:21:07,834][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:21:08,326][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:21:08,819][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:21:09,314][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:21:09,815][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:21:10,307][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:21:10,798][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:21:11,293][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:21:11,785][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:21:12,281][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:21:12,773][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:21:13,268][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:21:13,761][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:21:14,251][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:21:14,746][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:21:15,238][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:21:15,729][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:21:16,221][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:21:16,712][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:21:17,204][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:21:17,708][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:21:18,199][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:21:18,690][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:21:19,183][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:21:19,673][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:21:20,165][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:21:20,656][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:21:21,149][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:21:21,644][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:21:22,137][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:21:22,629][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:21:23,121][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:21:23,613][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:21:24,113][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:21:24,604][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:21:25,095][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:21:25,591][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:21:26,083][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:21:26,575][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:21:27,069][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:21:27,563][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:21:28,057][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:21:28,550][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:21:29,046][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:21:29,543][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:21:30,038][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9805 tokens. -[2025-11-09 18:21:31,047][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.00%, Current % of VRAM taken: 55.50%, Block Peak % of device VRAM: 61.82%, ΔTime: 00:00:32 -[2025-11-09 18:21:31,791][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:21:31,793][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:21:31,795][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:21:32,832][__main__][INFO] - Iteration 24 took 49s (27.80% Gen, 70.11% Train). Generation: 13s, Training: 34s. Estimated remaining time: 41h 4m 53s. Estimated total time: 41h 27m 29s. Time estimates for 10 more iterations: 8m 17s, 100 more iterations: 1h 22m 54s, 500 more iterations: 6h 54m 34s. -[2025-11-09 18:21:32,835][__main__][INFO] - Starting iteration 24. -[2025-11-09 18:21:33,312][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 2 and human policies 1. -[2025-11-09 18:21:33,313][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:21:46,823][__main__][INFO] - Number of regex retries in iteration 24: 0 -[2025-11-09 18:21:46,824][__main__][INFO] - agents played in iteration 24 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:21:47,613][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.50%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:21:47,641][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.50%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:21:47,667][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.50%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:21:47,691][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.50%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:21:47,692][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:21:47,692][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:21:48,358][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:21:48,815][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:21:49,313][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:21:49,806][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:21:50,300][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:21:50,790][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:21:51,284][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:21:51,775][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:21:52,266][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:21:52,759][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:21:53,251][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:21:53,745][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:21:54,267][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:21:54,766][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:21:55,259][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:21:55,751][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:21:56,244][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:21:56,736][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:21:57,228][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:21:57,719][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:21:58,211][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:21:58,701][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:21:59,203][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:21:59,693][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:22:00,184][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:22:00,674][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:22:01,164][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:22:01,654][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:22:02,145][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:22:02,636][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:22:03,129][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:22:03,619][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:22:04,109][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:22:04,616][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:22:05,107][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:22:05,597][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:22:06,092][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:22:06,586][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:22:07,076][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:22:07,568][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:22:08,060][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:22:08,553][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:22:09,046][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:22:09,539][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:22:10,033][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:22:10,527][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:22:11,019][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:22:11,510][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:22:12,004][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:22:12,513][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:22:13,004][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:22:13,495][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:22:13,990][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:22:14,481][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:22:14,972][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:22:15,471][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:22:15,964][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:22:16,458][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:22:16,950][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:22:17,444][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:22:17,944][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:22:18,441][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:22:18,936][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:22:19,432][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:22:19,928][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9764 tokens. -[2025-11-09 18:22:20,589][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.01%, Current % of VRAM taken: 55.51%, Block Peak % of device VRAM: 61.83%, ΔTime: 00:00:32 -[2025-11-09 18:22:21,309][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:22:21,313][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:22:21,315][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:22:22,283][__main__][INFO] - Iteration 25 took 48s (27.59% Gen, 70.43% Train). Generation: 13s, Training: 34s. Estimated remaining time: 40h 25m 8s. Estimated total time: 40h 48m 33s. Time estimates for 10 more iterations: 8m 9s, 100 more iterations: 1h 21m 37s, 500 more iterations: 6h 48m 5s. -[2025-11-09 18:22:22,285][__main__][INFO] - Starting iteration 25. -[2025-11-09 18:22:22,744][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 2 and human policies 1. -[2025-11-09 18:22:22,745][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:22:23,931][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:22:36,585][__main__][INFO] - Number of regex retries in iteration 25: 1 -[2025-11-09 18:22:36,586][__main__][INFO] - agents played in iteration 25 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:22:37,367][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:22:37,389][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:22:37,414][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:22:37,436][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:22:37,436][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:22:37,437][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:22:38,076][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:22:38,529][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:22:39,028][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:22:39,525][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:22:40,021][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:22:40,518][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:22:41,010][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:22:41,519][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:22:42,010][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:22:42,501][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:22:42,989][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:22:43,478][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:22:43,969][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:22:44,459][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:22:44,948][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:22:45,438][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:22:45,927][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:22:46,417][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:22:46,913][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:22:47,405][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:22:47,900][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:22:48,391][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:22:48,885][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:22:49,378][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:22:49,869][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:22:50,360][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:22:50,850][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:22:51,340][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:22:51,833][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:22:52,323][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:22:52,812][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:22:53,317][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:22:53,807][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:22:54,297][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:22:54,797][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:22:55,287][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:22:55,777][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:22:56,267][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:22:56,757][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:22:57,248][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:22:57,737][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:22:58,227][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:22:58,718][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:22:59,208][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:22:59,698][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:23:00,188][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:23:00,677][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:23:01,169][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:23:01,662][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:23:02,152][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:23:02,643][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:23:03,133][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:23:03,631][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:23:04,121][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:23:04,611][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:23:05,106][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:23:05,596][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:23:06,088][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:23:06,577][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:23:07,068][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:23:07,564][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:23:08,056][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:23:08,550][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:23:09,045][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:23:09,539][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9713 tokens. -[2025-11-09 18:23:10,217][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.02%, Current % of VRAM taken: 55.52%, Block Peak % of device VRAM: 61.61%, ΔTime: 00:00:32 -[2025-11-09 18:23:10,947][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:23:10,949][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:23:10,952][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:23:12,088][__main__][INFO] - Iteration 26 took 49s (28.05% Gen, 69.65% Train). Generation: 13s, Training: 34s. Estimated remaining time: 40h 42m 57s. Estimated total time: 41h 7m 12s. Time estimates for 10 more iterations: 8m 13s, 100 more iterations: 1h 22m 14s, 500 more iterations: 6h 51m 12s. -[2025-11-09 18:23:12,091][__main__][INFO] - Starting iteration 26. -[2025-11-09 18:23:12,561][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 2 and human policies 1. -[2025-11-09 18:23:12,562][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:23:13,599][mllm.models.large_language_model_local][WARNING] - Response Proposal: 5 hats, 0 books, 5 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:23:18,902][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls +Given the per-item values: +- You value hats at 10 and books at 1. +- Bob values books at 10 and hats at 1. -Given Alice's per-item values, she would prioritize hats and books over balls. Since you have a much higher value for hats compared to Alice and books, it's strategic to take all the hats to maximize your points. Since the values for books and balls are less desirable for you, proposing to take all of the hats and none of the books or balls should be effective in maximizing your points for this round. did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:23:27,368][__main__][INFO] - Number of regex retries in iteration 26: 2 -[2025-11-09 18:23:27,369][__main__][INFO] - agents played in iteration 26 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:23:28,216][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:23:28,245][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:23:28,272][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:23:28,296][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:23:28,296][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:23:28,297][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:23:28,979][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:23:29,430][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:23:29,928][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:23:30,425][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:23:30,916][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:23:31,412][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:23:31,903][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:23:32,395][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:23:32,892][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:23:33,384][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:23:33,876][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:23:34,369][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:23:34,862][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:23:35,354][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:23:35,846][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:23:36,337][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:23:36,847][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:23:37,345][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:23:37,843][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:23:38,336][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:23:38,829][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:23:39,322][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:23:39,814][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:23:40,307][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:23:40,855][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:23:41,347][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:23:41,840][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:23:42,333][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:23:42,825][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:23:43,342][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:23:43,834][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:23:44,327][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:23:44,827][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:23:45,317][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:23:45,808][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:23:46,298][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:23:46,788][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:23:47,279][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:23:47,769][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:23:48,259][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:23:48,748][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:23:49,238][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:23:49,737][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:23:50,229][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:23:50,723][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:23:51,213][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:23:51,706][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:23:52,198][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:23:52,690][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:23:53,182][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:23:53,673][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:23:54,164][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:23:54,654][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:23:55,144][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:23:55,634][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:23:56,124][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:23:56,620][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:23:57,111][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:23:57,610][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:23:58,104][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:23:58,601][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:23:59,100][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:23:59,596][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:24:00,090][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:24:00,587][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9793 tokens. -[2025-11-09 18:24:01,287][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.03%, Current % of VRAM taken: 55.53%, Block Peak % of device VRAM: 61.94%, ΔTime: 00:00:32 -[2025-11-09 18:24:02,030][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:24:02,031][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:24:02,033][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:24:03,332][__main__][INFO] - Iteration 27 took 50s (29.17% Gen, 68.27% Train). Generation: 14s, Training: 34s. Estimated remaining time: 41h 53m 26s. Estimated total time: 42h 18m 33s. Time estimates for 10 more iterations: 8m 27s, 100 more iterations: 1h 24m 37s, 500 more iterations: 7h 3m 5s. -[2025-11-09 18:24:03,334][__main__][INFO] - Starting iteration 27. -[2025-11-09 18:24:03,825][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 2 and human policies 1. -[2025-11-09 18:24:03,826][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:24:17,588][__main__][INFO] - Number of regex retries in iteration 27: 0 -[2025-11-09 18:24:17,589][__main__][INFO] - agents played in iteration 27 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:24:18,409][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:24:18,437][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:24:18,463][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:24:18,486][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:24:18,487][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:24:18,488][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:24:19,148][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:24:19,602][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:24:20,110][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:24:20,605][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:24:21,099][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:24:21,594][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:24:22,087][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:24:22,583][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:24:23,075][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:24:23,568][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:24:24,060][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:24:24,552][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:24:25,050][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:24:25,545][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:24:26,043][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:24:26,541][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:24:27,034][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:24:27,526][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:24:28,016][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:24:28,507][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:24:28,997][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:24:29,489][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:24:29,982][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:24:30,475][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:24:30,966][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:24:31,493][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:24:31,987][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:24:32,480][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:24:32,976][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:24:33,471][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:24:33,969][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:24:34,467][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:24:34,962][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:24:35,457][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:24:35,947][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:24:36,439][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:24:36,941][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:24:37,433][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:24:37,926][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:24:38,419][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:24:38,910][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:24:39,405][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:24:39,897][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:24:40,389][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:24:40,881][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:24:41,372][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:24:41,865][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:24:42,357][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:24:42,849][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:24:43,342][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:24:43,834][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:24:44,326][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:24:44,831][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:24:45,323][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:24:45,812][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:24:46,304][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:24:46,795][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:24:47,285][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:24:47,776][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:24:48,269][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:24:48,763][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:24:49,255][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:24:49,749][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:24:50,242][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:24:50,737][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9890 tokens. -[2025-11-09 18:24:51,465][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 13.96%, Current % of VRAM taken: 55.46%, Block Peak % of device VRAM: 61.83%, ΔTime: 00:00:32 -[2025-11-09 18:24:52,205][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:24:52,207][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:24:52,212][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:24:53,148][__main__][INFO] - Iteration 28 took 49s (27.90% Gen, 70.20% Train). Generation: 13s, Training: 34s. Estimated remaining time: 40h 40m 15s. Estimated total time: 41h 6m 11s. Time estimates for 10 more iterations: 8m 13s, 100 more iterations: 1h 22m 12s, 500 more iterations: 6h 51m 1s. -[2025-11-09 18:24:53,151][__main__][INFO] - Starting iteration 28. -[2025-11-09 18:24:53,660][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 2 and human policies 1. -[2025-11-09 18:24:53,661][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:25:07,598][__main__][INFO] - Number of regex retries in iteration 28: 0 -[2025-11-09 18:25:07,599][__main__][INFO] - agents played in iteration 28 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:25:08,426][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:25:08,451][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:25:08,478][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:25:08,501][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:25:08,502][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:25:08,502][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:25:09,168][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:25:09,622][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:25:10,125][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:25:10,618][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:25:11,112][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:25:11,605][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:25:12,097][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:25:12,588][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:25:13,081][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:25:13,575][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:25:14,069][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:25:14,570][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:25:15,064][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:25:15,556][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:25:16,066][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:25:16,559][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:25:17,052][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:25:17,548][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:25:18,040][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:25:18,538][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:25:19,028][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:25:19,518][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:25:20,011][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:25:20,502][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:25:20,992][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:25:21,485][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:25:21,979][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:25:22,473][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:25:22,966][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:25:23,458][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:25:23,951][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:25:24,443][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:25:24,935][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:25:25,428][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:25:25,919][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:25:26,412][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:25:26,905][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:25:27,397][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:25:27,891][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:25:28,383][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:25:28,877][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:25:29,370][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:25:29,864][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:25:30,357][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:25:30,849][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:25:31,342][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:25:31,833][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:25:32,324][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:25:32,815][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:25:33,307][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:25:33,797][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:25:34,289][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:25:34,780][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:25:35,272][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:25:35,769][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:25:36,265][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:25:36,758][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:25:37,253][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:25:37,751][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:25:38,243][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:25:38,736][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:25:39,231][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:25:39,727][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:25:40,223][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:25:40,719][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9845 tokens. -[2025-11-09 18:25:41,424][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 13.98%, Current % of VRAM taken: 55.48%, Block Peak % of device VRAM: 61.71%, ΔTime: 00:00:32 -[2025-11-09 18:25:42,147][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:25:42,149][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:25:42,150][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:25:43,098][__main__][INFO] - Iteration 29 took 49s (28.19% Gen, 69.89% Train). Generation: 13s, Training: 34s. Estimated remaining time: 40h 45m 9s. Estimated total time: 41h 11m 55s. Time estimates for 10 more iterations: 8m 14s, 100 more iterations: 1h 22m 23s, 500 more iterations: 6h 51m 59s. -[2025-11-09 18:25:43,100][__main__][INFO] - Starting iteration 29. -[2025-11-09 18:25:43,595][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 2 and human policies 1. -[2025-11-09 18:25:43,595][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:25:54,544][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 20 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:25:56,517][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:25:57,692][__main__][INFO] - Number of regex retries in iteration 29: 2 -[2025-11-09 18:25:57,692][__main__][INFO] - agents played in iteration 29 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:25:58,544][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.50%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:25:58,566][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.50%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:25:58,588][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.50%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:25:58,610][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.50%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:25:58,611][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:25:58,612][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:25:59,287][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:25:59,739][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:26:00,235][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:26:00,736][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:26:01,229][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:26:01,721][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:26:02,216][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:26:02,710][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:26:03,203][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:26:03,695][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:26:04,187][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:26:04,680][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:26:05,173][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:26:05,668][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:26:06,163][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:26:06,655][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:26:07,149][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:26:07,647][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:26:08,141][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:26:08,632][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:26:09,125][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:26:09,618][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:26:10,114][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:26:10,607][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:26:11,105][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:26:11,597][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:26:12,090][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:26:12,591][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:26:13,083][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:26:13,578][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:26:14,070][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:26:14,563][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:26:15,060][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:26:15,553][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:26:16,047][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:26:16,540][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:26:17,031][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:26:17,523][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:26:18,015][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:26:18,505][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:26:19,000][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:26:19,491][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:26:19,985][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:26:20,478][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:26:20,972][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:26:21,475][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:26:21,966][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:26:22,458][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:26:22,965][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:26:23,458][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:26:23,949][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:26:24,442][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:26:24,932][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:26:25,425][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:26:25,915][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:26:26,407][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:26:26,902][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:26:27,393][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:26:27,886][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:26:28,386][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:26:28,878][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:26:29,377][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:26:29,873][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:26:30,371][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:26:30,874][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9934 tokens. -[2025-11-09 18:26:31,594][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 55.56%, Block Peak % of device VRAM: 62.05%, ΔTime: 00:00:32 -[2025-11-09 18:26:32,333][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:26:32,335][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:26:32,337][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:26:33,238][__main__][INFO] - Iteration 30 took 49s (28.40% Gen, 69.79% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 54m 33s. Estimated total time: 41h 22m 9s. Time estimates for 10 more iterations: 8m 16s, 100 more iterations: 1h 22m 44s, 500 more iterations: 6h 53m 41s. -[2025-11-09 18:26:33,240][__main__][INFO] - Starting iteration 30. -[2025-11-09 18:26:33,787][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 2 and human policies 1. -[2025-11-09 18:26:33,788][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:26:41,644][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls +By proposing to take all 10 hats, I maximize my share since I have a much higher value for hats compared to Bob. Since I have no value for books, taking any would be a loss. Similarly, balls have the least value for you, so they should be left for Bob. This proposal ensures I get the maximum value possible from the items that I highly value. did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 21:20:05,842][__main__][INFO] - Number of regex retries in iteration 14: 2 +[2026-03-25 21:20:05,844][__main__][INFO] - agents played in iteration 14 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:20:06,637][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:20:06,660][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:20:06,684][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:20:06,707][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:20:06,707][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:20:06,708][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:20:07,321][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:20:07,772][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:20:08,271][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:20:08,768][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:20:09,264][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:20:09,758][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:20:10,252][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:20:10,747][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:20:11,241][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:20:11,734][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:20:12,227][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:20:12,722][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:20:13,215][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:20:13,708][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:20:14,203][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:20:14,695][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:20:15,189][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:20:15,683][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:20:16,175][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:20:16,670][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:20:17,164][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:20:17,659][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:20:18,154][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:20:18,647][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:20:19,142][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:20:19,637][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:20:20,130][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:20:20,623][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:20:21,116][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:20:21,608][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:20:22,100][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:20:22,593][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:20:23,088][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:20:23,581][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:20:24,072][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:20:24,565][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:20:25,055][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:20:25,547][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:20:26,041][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:20:26,534][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:20:27,027][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:20:27,525][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:20:28,043][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:20:28,537][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:20:29,032][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:20:29,527][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:20:30,025][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:20:30,519][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:20:31,023][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:20:31,518][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:20:32,014][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:20:32,509][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:20:33,004][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:20:33,499][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:20:33,992][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:20:34,486][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:20:34,981][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:20:35,474][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:20:35,965][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:20:36,457][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:20:36,948][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:20:37,441][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:20:37,932][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:20:38,426][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:20:38,917][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9759 tokens. +[2026-03-25 21:20:39,614][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 13.97%, Current % of VRAM taken: 60.40%, Block Peak % of device VRAM: 61.89%, ΔTime: 00:00:32 +[2026-03-25 21:20:40,322][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:20:40,324][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:20:40,325][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:20:40,965][__main__][INFO] - Iteration 15 took 49s (29.20% Gen, 69.51% Train). Generation: 14s, Training: 34s. Estimated remaining time: 41h 5m 17s. Estimated total time: 41h 20m 18s. Time estimates for 10 more iterations: 8m 16s, 100 more iterations: 1h 22m 40s, 500 more iterations: 6h 53m 23s. +[2026-03-25 21:20:40,967][__main__][INFO] - Starting iteration 15. +[2026-03-25 21:20:41,367][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 1 and human policies 1. +[2026-03-25 21:20:41,367][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:20:54,831][__main__][INFO] - Number of regex retries in iteration 15: 0 +[2026-03-25 21:20:54,833][__main__][INFO] - agents played in iteration 15 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:20:55,644][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:20:55,668][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:20:55,692][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:20:55,715][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:20:55,716][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:20:55,716][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:20:56,364][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:20:56,819][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:20:57,318][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:20:57,814][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:20:58,311][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:20:58,805][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:20:59,302][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:20:59,797][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:21:00,291][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:21:00,790][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:21:01,284][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:21:01,782][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:21:02,278][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:21:02,771][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:21:03,268][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:21:03,767][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:21:04,262][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:21:04,758][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:21:05,254][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:21:05,747][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:21:06,243][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:21:06,741][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:21:07,238][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:21:07,732][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:21:08,227][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:21:08,720][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:21:09,213][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:21:09,705][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:21:10,198][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:21:10,689][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:21:11,181][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:21:11,672][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:21:12,166][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:21:12,661][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:21:13,154][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:21:13,649][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:21:14,147][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:21:14,641][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:21:15,136][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:21:15,631][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:21:16,124][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:21:16,620][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:21:17,113][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:21:17,606][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:21:18,101][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:21:18,595][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:21:19,087][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:21:19,581][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:21:20,076][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:21:20,570][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:21:21,066][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:21:21,563][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:21:22,059][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:21:22,552][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:21:23,048][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:21:23,543][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:21:24,037][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:21:24,530][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:21:25,022][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:21:25,514][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:21:26,007][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:21:26,501][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:21:26,995][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:21:27,488][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:21:27,981][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9732 tokens. +[2026-03-25 21:21:28,659][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 13.96%, Current % of VRAM taken: 60.39%, Block Peak % of device VRAM: 61.67%, ΔTime: 00:00:32 +[2026-03-25 21:21:29,375][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:21:29,378][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:21:29,379][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:21:30,016][__main__][INFO] - Iteration 16 took 48s (27.68% Gen, 71.01% Train). Generation: 13s, Training: 34s. Estimated remaining time: 40h 16m 41s. Estimated total time: 40h 32m 31s. Time estimates for 10 more iterations: 8m 6s, 100 more iterations: 1h 21m 5s, 500 more iterations: 6h 45m 25s. +[2026-03-25 21:21:30,019][__main__][INFO] - Starting iteration 16. +[2026-03-25 21:21:30,420][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 1 and human policies 1. +[2026-03-25 21:21:30,420][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:21:43,616][__main__][INFO] - Number of regex retries in iteration 16: 0 +[2026-03-25 21:21:43,617][__main__][INFO] - agents played in iteration 16 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:21:44,434][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.43%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:21:44,458][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.43%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:21:44,481][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.43%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:21:44,504][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.43%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:21:44,504][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:21:44,505][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:21:45,162][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:21:45,612][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:21:46,114][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:21:46,608][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:21:47,104][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:21:47,597][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:21:48,090][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:21:48,584][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:21:49,080][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:21:49,578][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:21:50,074][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:21:50,571][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:21:51,067][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:21:51,561][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:21:52,055][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:21:52,551][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:21:53,047][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:21:53,543][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:21:54,036][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:21:54,532][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:21:55,027][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:21:55,523][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:21:56,017][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:21:56,511][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:21:57,006][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:21:57,501][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:21:57,994][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:21:58,489][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:21:58,984][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:21:59,478][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:21:59,972][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:22:00,466][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:22:00,963][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:22:01,456][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:22:01,951][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:22:02,445][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:22:02,938][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:22:03,433][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:22:03,926][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:22:04,422][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:22:04,916][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:22:05,408][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:22:05,901][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:22:06,394][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:22:06,889][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:22:07,384][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:22:07,879][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:22:08,371][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:22:08,865][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:22:09,382][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:22:09,881][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:22:10,376][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:22:10,868][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:22:11,363][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:22:11,856][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:22:12,349][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:22:12,843][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:22:13,336][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:22:13,830][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:22:14,325][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:22:14,819][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:22:15,312][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:22:15,806][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:22:16,302][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:22:16,795][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9647 tokens. +[2026-03-25 21:22:17,473][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 13.96%, Current % of VRAM taken: 60.38%, Block Peak % of device VRAM: 61.52%, ΔTime: 00:00:32 +[2026-03-25 21:22:18,175][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:22:18,179][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:22:18,180][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:22:18,823][__main__][INFO] - Iteration 17 took 48s (27.26% Gen, 71.41% Train). Generation: 13s, Training: 34s. Estimated remaining time: 40h 3m 32s. Estimated total time: 40h 20m 11s. Time estimates for 10 more iterations: 8m 4s, 100 more iterations: 1h 20m 40s, 500 more iterations: 6h 43m 21s. +[2026-03-25 21:22:18,825][__main__][INFO] - Starting iteration 17. +[2026-03-25 21:22:19,224][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 1 and human policies 1. +[2026-03-25 21:22:19,225][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:22:32,544][__main__][INFO] - Number of regex retries in iteration 17: 0 +[2026-03-25 21:22:32,545][__main__][INFO] - agents played in iteration 17 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:22:33,364][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.43%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:22:33,387][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.43%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:22:33,410][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.43%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:22:33,433][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.43%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:22:33,434][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:22:33,434][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:22:34,098][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:22:34,550][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:22:35,050][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:22:35,542][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:22:36,036][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:22:36,532][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:22:37,028][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:22:37,522][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:22:38,016][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:22:38,509][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:22:39,003][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:22:39,496][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:22:39,990][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:22:40,483][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:22:40,987][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:22:41,482][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:22:41,979][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:22:42,473][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:22:42,970][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:22:43,467][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:22:43,963][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:22:44,460][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:22:44,957][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:22:45,451][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:22:45,947][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:22:46,442][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:22:46,936][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:22:47,429][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:22:47,926][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:22:48,420][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:22:48,912][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:22:49,407][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:22:49,903][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:22:50,397][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:22:50,891][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:22:51,385][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:22:51,882][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:22:52,377][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:22:52,871][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:22:53,366][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:22:53,861][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:22:54,357][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:22:54,851][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:22:55,347][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:22:55,840][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:22:56,334][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:22:56,830][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:22:57,326][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:22:57,820][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:22:58,312][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:22:58,804][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:22:59,296][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:22:59,789][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:23:00,282][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:23:00,776][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:23:01,268][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:23:01,761][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:23:02,257][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:23:02,753][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:23:03,247][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:23:03,741][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:23:04,234][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:23:04,728][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:23:05,222][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:23:05,716][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9676 tokens. +[2026-03-25 21:23:06,413][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 13.99%, Current % of VRAM taken: 60.42%, Block Peak % of device VRAM: 61.65%, ΔTime: 00:00:32 +[2026-03-25 21:23:07,117][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:23:07,119][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:23:07,121][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:23:07,767][__main__][INFO] - Iteration 18 took 48s (27.44% Gen, 71.23% Train). Generation: 13s, Training: 34s. Estimated remaining time: 40h 9m 42s. Estimated total time: 40h 27m 10s. Time estimates for 10 more iterations: 8m 5s, 100 more iterations: 1h 20m 54s, 500 more iterations: 6h 44m 31s. +[2026-03-25 21:23:07,770][__main__][INFO] - Starting iteration 18. +[2026-03-25 21:23:08,173][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 1 and human policies 1. +[2026-03-25 21:23:08,174][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:23:21,759][__main__][INFO] - Number of regex retries in iteration 18: 0 +[2026-03-25 21:23:21,760][__main__][INFO] - agents played in iteration 18 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:23:22,562][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.43%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:23:22,586][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.43%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:23:22,610][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.43%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:23:22,633][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.43%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:23:22,634][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:23:22,634][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:23:23,287][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:23:23,742][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:23:24,241][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:23:24,738][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:23:25,233][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:23:25,727][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:23:26,222][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:23:26,718][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:23:27,212][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:23:27,708][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:23:28,203][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:23:28,701][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:23:29,199][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:23:29,693][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:23:30,189][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:23:30,684][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:23:31,178][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:23:31,671][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:23:32,168][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:23:32,660][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:23:33,154][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:23:33,647][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:23:34,142][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:23:34,637][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:23:35,131][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:23:35,625][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:23:36,121][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:23:36,615][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:23:37,109][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:23:37,604][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:23:38,097][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:23:38,590][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:23:39,085][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:23:39,578][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:23:40,072][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:23:40,566][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:23:41,057][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:23:41,551][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:23:42,045][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:23:42,541][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:23:43,034][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:23:43,528][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:23:44,020][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:23:44,512][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:23:45,004][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:23:45,496][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:23:45,988][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:23:46,481][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:23:46,974][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:23:47,467][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:23:47,962][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:23:48,455][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:23:48,950][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:23:49,444][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:23:49,937][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:23:50,431][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:23:50,926][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:23:51,419][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:23:51,912][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:23:52,405][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:23:52,901][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:23:53,395][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:23:53,888][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:23:54,382][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:23:54,875][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9696 tokens. +[2026-03-25 21:23:55,541][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.00%, Current % of VRAM taken: 60.42%, Block Peak % of device VRAM: 61.71%, ΔTime: 00:00:32 +[2026-03-25 21:23:56,245][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:23:56,247][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:23:56,249][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:23:56,891][__main__][INFO] - Iteration 19 took 48s (27.89% Gen, 70.79% Train). Generation: 13s, Training: 34s. Estimated remaining time: 40h 17m 38s. Estimated total time: 40h 35m 54s. Time estimates for 10 more iterations: 8m 7s, 100 more iterations: 1h 21m 11s, 500 more iterations: 6h 45m 59s. +[2026-03-25 21:23:56,893][__main__][INFO] - Starting iteration 19. +[2026-03-25 21:23:57,293][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 1 and human policies 1. +[2026-03-25 21:23:57,293][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:24:11,128][__main__][INFO] - Number of regex retries in iteration 19: 0 +[2026-03-25 21:24:11,129][__main__][INFO] - agents played in iteration 19 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:24:11,921][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.45%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:24:11,944][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.45%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:24:11,967][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.45%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:24:11,990][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.45%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:24:11,991][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:24:11,992][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:24:12,600][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:24:13,053][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:24:13,550][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:24:14,047][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:24:14,544][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:24:15,038][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:24:15,531][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:24:16,027][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:24:16,521][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:24:17,024][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:24:17,520][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:24:18,015][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:24:18,513][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:24:19,009][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:24:19,506][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:24:20,003][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:24:20,503][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:24:20,996][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:24:21,487][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:24:21,980][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:24:22,474][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:24:22,966][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:24:23,458][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:24:23,952][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:24:24,446][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:24:24,940][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:24:25,434][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:24:25,927][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:24:26,423][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:24:26,917][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:24:27,410][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:24:27,905][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:24:28,397][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:24:28,890][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:24:29,383][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:24:29,878][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:24:30,371][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:24:30,865][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:24:31,358][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:24:31,851][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:24:32,346][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:24:32,840][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:24:33,334][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:24:33,828][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:24:34,326][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:24:34,819][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:24:35,314][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:24:35,808][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:24:36,305][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:24:36,798][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:24:37,293][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:24:37,786][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:24:38,279][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:24:38,772][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:24:39,265][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:24:39,758][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:24:40,251][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:24:40,746][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:24:41,239][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:24:41,733][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:24:42,228][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:24:42,725][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:24:43,218][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:24:43,713][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:24:44,207][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9743 tokens. +[2026-03-25 21:24:44,892][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.02%, Current % of VRAM taken: 60.44%, Block Peak % of device VRAM: 61.81%, ΔTime: 00:00:32 +[2026-03-25 21:24:45,595][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:24:45,597][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:24:45,600][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:24:46,239][__main__][INFO] - Iteration 20 took 48s (28.27% Gen, 70.43% Train). Generation: 13s, Training: 34s. Estimated remaining time: 40h 28m 13s. Estimated total time: 40h 47m 19s. Time estimates for 10 more iterations: 8m 9s, 100 more iterations: 1h 21m 34s, 500 more iterations: 6h 47m 53s. +[2026-03-25 21:24:46,241][__main__][INFO] - Starting iteration 20. +[2026-03-25 21:24:46,643][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 1 and human policies 1. +[2026-03-25 21:24:46,644][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:24:47,728][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 21:24:52,345][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls -Given the per-item values, I should focus on items where my value is higher relative to Bob's value. For hats, my value (10) is significantly higher than Bob's (1). For books, Bob values them more (10) than I do (10), so I should avoid proposing to keep any books. Balls have the same value for both of us (1), making it a less attractive option. Therefore, I propose keeping all 10 hats to maximize my points. did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:26:48,677][__main__][INFO] - Number of regex retries in iteration 30: 1 -[2025-11-09 18:26:48,678][__main__][INFO] - agents played in iteration 30 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:26:49,495][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:26:49,523][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:26:49,549][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:26:49,572][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:26:49,573][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:26:49,575][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:26:50,236][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:26:50,698][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:26:51,196][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:26:51,689][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:26:52,183][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:26:52,674][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:26:53,172][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:26:53,668][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:26:54,160][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:26:54,669][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:26:55,166][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:26:55,659][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:26:56,155][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:26:56,649][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:26:57,165][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:26:57,664][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:26:58,157][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:26:58,661][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:26:59,157][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:26:59,651][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:27:00,146][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:27:00,640][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:27:01,133][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:27:01,625][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:27:02,119][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:27:02,611][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:27:03,104][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:27:03,597][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:27:04,088][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:27:04,581][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:27:05,073][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:27:05,567][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:27:06,060][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:27:06,553][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:27:07,046][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:27:07,539][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:27:08,031][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:27:08,525][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:27:09,019][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:27:09,511][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:27:10,005][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:27:10,496][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:27:10,987][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:27:11,477][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:27:11,967][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:27:12,458][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:27:12,947][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:27:13,438][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:27:13,930][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:27:14,424][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:27:14,921][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:27:15,415][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:27:15,909][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:27:16,400][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:27:16,892][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:27:17,384][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:27:17,877][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:27:18,370][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:27:18,864][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:27:19,359][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:27:19,865][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:27:20,364][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:27:20,871][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:27:21,371][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:27:21,877][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9909 tokens. -[2025-11-09 18:27:22,568][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 55.68%, Block Peak % of device VRAM: 62.12%, ΔTime: 00:00:32 -[2025-11-09 18:27:23,364][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:27:23,366][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:27:23,367][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:27:25,179][__main__][INFO] - Iteration 31 took 51s (28.97% Gen, 67.50% Train). Generation: 14s, Training: 34s. Estimated remaining time: 42h 21m 8s. Estimated total time: 42h 49m 36s. Time estimates for 10 more iterations: 8m 33s, 100 more iterations: 1h 25m 39s, 500 more iterations: 7h 8m 16s. -[2025-11-09 18:27:25,181][__main__][INFO] - Starting iteration 31. -[2025-11-09 18:27:25,647][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 3 and human policies 1. -[2025-11-09 18:27:25,648][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:27:39,736][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls +Given the per-item values, you have a significantly higher value for hats compared to Bob, while Bob values books more than you do. This suggests that taking all the hats would be the most beneficial for you, as it maximizes your points based on your higher per-item value for hats. Books and balls do not offer as much value for you compared to the hats. did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 21:25:00,920][__main__][INFO] - Number of regex retries in iteration 20: 2 +[2026-03-25 21:25:00,921][__main__][INFO] - agents played in iteration 20 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:25:01,699][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.45%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:25:01,722][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.45%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:25:01,745][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.45%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:25:01,768][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.45%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:25:01,769][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:25:01,770][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:25:02,363][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:25:02,817][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:25:03,317][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:25:03,812][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:25:04,309][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:25:04,804][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:25:05,299][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:25:05,796][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:25:06,295][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:25:06,788][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:25:07,281][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:25:07,775][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:25:08,270][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:25:08,763][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:25:09,257][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:25:09,748][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:25:10,242][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:25:10,735][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:25:11,228][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:25:11,723][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:25:12,216][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:25:12,710][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:25:13,203][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:25:13,696][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:25:14,189][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:25:14,683][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:25:15,177][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:25:15,670][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:25:16,164][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:25:16,659][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:25:17,153][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:25:17,647][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:25:18,140][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:25:18,636][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:25:19,130][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:25:19,624][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:25:20,120][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:25:20,613][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:25:21,106][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:25:21,601][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:25:22,094][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:25:22,586][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:25:23,079][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:25:23,572][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:25:24,064][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:25:24,557][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:25:25,049][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:25:25,542][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:25:26,034][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:25:26,529][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:25:27,023][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:25:27,515][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:25:28,008][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:25:28,502][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:25:28,994][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:25:29,488][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:25:29,981][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:25:30,472][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:25:31,056][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:25:31,551][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:25:32,046][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:25:32,540][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:25:33,034][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:25:33,526][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:25:34,022][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9666 tokens. +[2026-03-25 21:25:34,706][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 13.96%, Current % of VRAM taken: 60.38%, Block Peak % of device VRAM: 61.82%, ΔTime: 00:00:32 +[2026-03-25 21:25:35,424][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:25:35,426][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:25:35,428][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:25:36,655][__main__][INFO] - Iteration 21 took 50s (28.55% Gen, 69.00% Train). Generation: 14s, Training: 34s. Estimated remaining time: 41h 20m 42s. Estimated total time: 41h 40m 38s. Time estimates for 10 more iterations: 8m 20s, 100 more iterations: 1h 23m 21s, 500 more iterations: 6h 56m 46s. +[2026-03-25 21:25:36,658][__main__][INFO] - Starting iteration 21. +[2026-03-25 21:25:37,058][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 2 and human policies 1. +[2026-03-25 21:25:37,059][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:25:47,988][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 21:25:50,527][__main__][INFO] - Number of regex retries in iteration 21: 1 +[2026-03-25 21:25:50,528][__main__][INFO] - agents played in iteration 21 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:25:51,374][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.44%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:25:51,397][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.44%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:25:51,420][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.44%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:25:51,443][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.44%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:25:51,444][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:25:51,444][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:25:52,047][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:25:52,499][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:25:52,997][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:25:53,491][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:25:53,984][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:25:54,477][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:25:54,970][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:25:55,467][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:25:55,962][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:25:56,457][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:25:56,952][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:25:57,446][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:25:57,941][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:25:58,436][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:25:58,931][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:25:59,427][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:25:59,921][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:26:00,413][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:26:00,907][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:26:01,401][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:26:01,894][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:26:02,388][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:26:02,881][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:26:03,376][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:26:03,870][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:26:04,364][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:26:04,858][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:26:05,352][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:26:05,847][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:26:06,341][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:26:06,834][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:26:07,329][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:26:07,826][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:26:08,317][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:26:08,808][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:26:09,298][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:26:09,788][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:26:10,279][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:26:10,770][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:26:11,262][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:26:11,754][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:26:12,247][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:26:12,741][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:26:13,234][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:26:13,728][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:26:14,223][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:26:14,717][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:26:15,211][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:26:15,705][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:26:16,197][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:26:16,690][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:26:17,183][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:26:17,674][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:26:18,167][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:26:18,658][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:26:19,150][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:26:19,643][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:26:20,137][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:26:20,630][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:26:21,124][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:26:21,617][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:26:22,112][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:26:22,607][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:26:23,100][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:26:23,593][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9725 tokens. +[2026-03-25 21:26:24,272][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 13.94%, Current % of VRAM taken: 60.37%, Block Peak % of device VRAM: 61.58%, ΔTime: 00:00:32 +[2026-03-25 21:26:24,977][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:26:24,979][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:26:24,980][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:26:25,618][__main__][INFO] - Iteration 22 took 48s (27.74% Gen, 70.95% Train). Generation: 13s, Training: 34s. Estimated remaining time: 40h 7m 16s. Estimated total time: 40h 28m 1s. Time estimates for 10 more iterations: 8m 5s, 100 more iterations: 1h 20m 56s, 500 more iterations: 6h 44m 40s. +[2026-03-25 21:26:25,621][__main__][INFO] - Starting iteration 22. +[2026-03-25 21:26:26,022][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 2 and human policies 1. +[2026-03-25 21:26:26,022][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:26:27,110][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 21:26:31,664][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls -Given the per-item values: -- You value hats at 10 and Bob values hats at 1. -- You value books at 10 but Bob values books at 10. -- You value balls at 1 and Bob values balls at 10. +This proposal is based on the assumption that Bob values books and balls much higher than hats, and I value hats and books higher than balls. By keeping all 10 hats, I ensure that I get the maximum quantity of the items I value most, while leaving all books and balls to Bob. This strategy aims to maximize my points for this round. did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 21:26:34,589][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 21:26:40,497][__main__][INFO] - Number of regex retries in iteration 22: 3 +[2026-03-25 21:26:40,498][__main__][INFO] - agents played in iteration 22 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:26:41,280][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.43%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:26:41,304][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.43%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:26:41,327][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.43%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:26:41,350][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.43%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:26:41,351][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:26:41,352][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:26:41,959][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:26:42,416][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:26:42,918][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:26:43,416][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:26:43,911][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:26:44,410][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:26:44,906][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:26:45,403][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:26:45,899][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:26:46,393][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:26:46,890][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:26:47,393][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:26:47,889][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:26:48,387][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:26:48,885][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:26:49,381][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:26:49,881][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:26:50,374][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:26:50,870][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:26:51,365][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:26:51,857][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:26:52,351][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:26:52,846][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:26:53,343][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:26:53,836][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:26:54,332][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:26:54,826][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:26:55,322][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:26:55,816][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:26:56,309][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:26:56,804][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:26:57,298][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:26:57,793][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:26:58,286][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:26:58,781][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:26:59,276][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:26:59,771][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:27:00,266][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:27:00,762][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:27:01,259][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:27:01,754][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:27:02,251][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:27:02,747][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:27:03,243][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:27:03,738][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:27:04,233][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:27:04,728][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:27:05,224][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:27:05,718][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:27:06,214][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:27:06,710][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:27:07,207][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:27:07,704][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:27:08,198][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:27:08,691][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:27:09,186][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:27:09,681][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:27:10,175][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:27:10,670][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:27:11,167][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:27:11,661][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:27:12,157][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:27:12,652][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:27:13,147][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:27:13,645][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9798 tokens. +[2026-03-25 21:27:14,324][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 60.56%, Block Peak % of device VRAM: 61.89%, ΔTime: 00:00:32 +[2026-03-25 21:27:15,041][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:27:15,043][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:27:15,044][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:27:15,676][__main__][INFO] - Iteration 23 took 49s (29.15% Gen, 69.57% Train). Generation: 14s, Training: 34s. Estimated remaining time: 41h 1m 10s. Estimated total time: 41h 22m 46s. Time estimates for 10 more iterations: 8m 16s, 100 more iterations: 1h 22m 45s, 500 more iterations: 6h 53m 47s. +[2026-03-25 21:27:15,679][__main__][INFO] - Starting iteration 23. +[2026-03-25 21:27:16,079][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 2 and human policies 1. +[2026-03-25 21:27:16,079][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:27:29,712][__main__][INFO] - Number of regex retries in iteration 23: 0 +[2026-03-25 21:27:29,713][__main__][INFO] - agents played in iteration 23 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:27:30,493][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:27:30,516][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:27:30,539][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:27:30,562][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:27:30,563][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:27:30,564][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:27:31,221][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:27:31,676][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:27:32,186][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:27:32,682][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:27:33,176][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:27:33,671][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:27:34,169][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:27:34,665][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:27:35,161][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:27:35,657][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:27:36,154][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:27:36,649][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:27:37,144][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:27:37,637][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:27:38,131][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:27:38,627][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:27:39,120][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:27:39,617][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:27:40,111][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:27:40,605][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:27:41,103][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:27:41,597][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:27:42,090][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:27:42,584][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:27:43,077][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:27:43,571][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:27:44,070][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:27:44,564][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:27:45,058][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:27:45,552][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:27:46,046][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:27:46,539][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:27:47,034][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:27:47,529][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:27:48,024][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:27:48,519][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:27:49,014][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:27:49,508][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:27:50,001][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:27:50,494][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:27:50,989][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:27:51,483][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:27:51,975][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:27:52,468][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:27:52,963][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:27:53,456][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:27:53,949][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:27:54,443][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:27:54,936][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:27:55,429][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:27:55,924][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:27:56,416][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:27:56,909][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:27:57,402][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:27:57,895][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:27:58,388][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:27:58,883][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:27:59,377][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:27:59,872][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:28:00,366][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:28:00,862][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:28:01,355][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:28:01,851][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:28:02,347][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:28:02,841][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9800 tokens. +[2026-03-25 21:28:03,519][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.02%, Current % of VRAM taken: 60.45%, Block Peak % of device VRAM: 61.94%, ΔTime: 00:00:32 +[2026-03-25 21:28:04,235][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:28:04,237][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:28:04,239][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:28:04,884][__main__][INFO] - Iteration 24 took 48s (27.93% Gen, 70.74% Train). Generation: 13s, Training: 34s. Estimated remaining time: 40h 17m 53s. Estimated total time: 40h 40m 17s. Time estimates for 10 more iterations: 8m 8s, 100 more iterations: 1h 21m 20s, 500 more iterations: 6h 46m 42s. +[2026-03-25 21:28:04,886][__main__][INFO] - Starting iteration 24. +[2026-03-25 21:28:05,285][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 2 and human policies 1. +[2026-03-25 21:28:05,286][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:28:06,435][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 21:28:19,253][__main__][INFO] - Number of regex retries in iteration 24: 1 +[2026-03-25 21:28:19,254][__main__][INFO] - agents played in iteration 24 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:28:20,031][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:28:20,055][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:28:20,079][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:28:20,103][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:28:20,103][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:28:20,104][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:28:20,779][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:28:21,233][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:28:21,730][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:28:22,225][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:28:22,719][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:28:23,213][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:28:23,709][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:28:24,204][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:28:24,697][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:28:25,190][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:28:25,688][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:28:26,188][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:28:26,712][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:28:27,207][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:28:27,709][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:28:28,205][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:28:28,706][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:28:29,200][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:28:29,694][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:28:30,187][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:28:30,682][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:28:31,178][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:28:31,672][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:28:32,167][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:28:32,661][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:28:33,156][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:28:33,650][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:28:34,147][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:28:34,644][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:28:35,139][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:28:35,635][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:28:36,131][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:28:36,629][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:28:37,126][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:28:37,622][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:28:38,116][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:28:38,609][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:28:39,104][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:28:39,598][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:28:40,093][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:28:40,588][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:28:41,083][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:28:41,578][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:28:42,069][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:28:42,566][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:28:43,060][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:28:43,554][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:28:44,048][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:28:44,543][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:28:45,038][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:28:45,532][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:28:46,026][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:28:46,519][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:28:47,014][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:28:47,507][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:28:48,005][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:28:48,500][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:28:48,995][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:28:49,490][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:28:49,985][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:28:50,482][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:28:50,978][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:28:51,473][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:28:51,968][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:28:52,464][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9776 tokens. +[2026-03-25 21:28:53,140][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.04%, Current % of VRAM taken: 60.46%, Block Peak % of device VRAM: 61.84%, ΔTime: 00:00:32 +[2026-03-25 21:28:56,995][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:28:56,997][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:28:56,999][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:28:57,640][__main__][INFO] - Iteration 25 took 52s (26.68% Gen, 72.09% Train). Generation: 13s, Training: 37s. Estimated remaining time: 43h 14m 28s. Estimated total time: 43h 37m 45s. Time estimates for 10 more iterations: 8m 43s, 100 more iterations: 1h 27m 15s, 500 more iterations: 7h 16m 17s. +[2026-03-25 21:28:57,642][__main__][INFO] - Starting iteration 25. +[2026-03-25 21:28:58,040][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 2 and human policies 1. +[2026-03-25 21:28:58,041][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:28:59,104][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 21:29:10,557][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 21:29:11,781][__main__][INFO] - Number of regex retries in iteration 25: 2 +[2026-03-25 21:29:11,782][__main__][INFO] - agents played in iteration 25 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:29:12,583][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:29:12,609][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:29:12,633][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:29:12,656][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:29:12,657][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:29:12,657][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:29:13,279][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:29:13,731][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:29:14,232][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:29:14,726][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:29:15,220][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:29:15,713][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:29:16,210][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:29:16,703][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:29:17,196][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:29:17,689][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:29:18,180][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:29:18,673][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:29:19,166][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:29:19,660][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:29:20,154][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:29:20,647][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:29:21,139][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:29:21,634][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:29:22,130][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:29:22,626][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:29:23,122][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:29:23,616][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:29:24,112][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:29:24,606][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:29:25,101][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:29:25,596][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:29:26,090][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:29:26,581][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:29:27,074][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:29:27,568][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:29:28,063][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:29:28,558][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:29:29,052][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:29:29,546][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:29:30,040][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:29:30,533][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:29:31,025][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:29:31,518][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:29:32,015][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:29:32,509][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:29:33,002][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:29:33,497][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:29:33,989][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:29:34,483][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:29:34,976][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:29:35,468][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:29:35,964][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:29:36,457][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:29:36,949][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:29:37,443][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:29:37,940][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:29:38,433][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:29:38,928][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:29:39,423][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:29:39,918][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:29:40,412][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:29:40,906][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:29:41,400][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:29:41,893][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:29:42,390][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:29:42,886][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:29:43,379][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:29:43,874][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:29:44,368][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:29:44,866][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9741 tokens. +[2026-03-25 21:29:45,550][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.02%, Current % of VRAM taken: 60.45%, Block Peak % of device VRAM: 61.63%, ΔTime: 00:00:32 +[2026-03-25 21:29:46,270][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:29:46,272][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:29:46,274][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:29:46,916][__main__][INFO] - Iteration 26 took 48s (28.11% Gen, 70.57% Train). Generation: 13s, Training: 34s. Estimated remaining time: 40h 19m 43s. Estimated total time: 40h 43m 50s. Time estimates for 10 more iterations: 8m 8s, 100 more iterations: 1h 21m 27s, 500 more iterations: 6h 47m 18s. +[2026-03-25 21:29:46,919][__main__][INFO] - Starting iteration 26. +[2026-03-25 21:29:47,323][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 2 and human policies 1. +[2026-03-25 21:29:47,323][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:29:51,267][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls + +Given the values, I see that the hats are highly valuable to me compared to their value to Bob. The books and balls have a similar low value for both of us. By keeping all the hats, I maximize my potential points from this round. did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 21:30:00,263][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls -If we propose to take all hats (10 hats, 0 books, 0 balls), we are maximizing the items you value the most, and since the total sum of your proposals (10 hats) is less than or equal to the total quantity (10 hats), you will receive exactly 10 hats. This will give you a contribution of \(10 \times 10 = 100\) points for this round because your per-item value for hats is 10. +Given the values: +- Hats: I value at 10, Bob values at 1. +- Books: I value at 10, Bob values at 10. +- Balls: I value at 1, Bob values at 10. -While Bob values books and balls more, this proposal plays to your strengths by taking the items you assess as most valuable, and it avoids contesting with Bob for the items he values highly, which could result in less favorable splits or even proportional sharing if my proposals exceed the available quantity. did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:27:42,246][__main__][INFO] - Number of regex retries in iteration 31: 1 -[2025-11-09 18:27:42,247][__main__][INFO] - agents played in iteration 31 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:27:43,013][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:27:43,035][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:27:43,057][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:27:43,079][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:27:43,079][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:27:43,080][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:27:43,673][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:27:44,122][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:27:44,623][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:27:45,115][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:27:45,609][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:27:46,104][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:27:46,600][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:27:47,111][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:27:47,604][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:27:48,096][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:27:48,591][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:27:49,084][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:27:49,578][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:27:50,070][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:27:50,564][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:27:51,057][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:27:51,553][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:27:52,050][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:27:52,545][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:27:53,038][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:27:53,539][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:27:54,031][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:27:54,527][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:27:55,023][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:27:55,514][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:27:56,008][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:27:56,505][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:27:56,999][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:27:57,518][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:27:58,010][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:27:58,505][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:27:58,997][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:27:59,490][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:27:59,986][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:28:00,475][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:28:00,966][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:28:01,456][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:28:01,949][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:28:02,439][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:28:02,931][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:28:03,422][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:28:03,913][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:28:04,404][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:28:04,895][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:28:05,386][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:28:05,878][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:28:06,369][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:28:06,862][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:28:07,352][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:28:07,845][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:28:08,335][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:28:08,826][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:28:09,318][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:28:09,814][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:28:10,306][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:28:10,798][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:28:11,293][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:28:11,787][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:28:12,280][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:28:12,776][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:28:13,272][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:28:13,768][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:28:14,266][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:28:14,760][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:28:15,255][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9869 tokens. -[2025-11-09 18:28:15,959][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 13.99%, Current % of VRAM taken: 55.49%, Block Peak % of device VRAM: 61.87%, ΔTime: 00:00:32 -[2025-11-09 18:28:16,663][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:28:16,665][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:28:16,666][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:28:17,551][__main__][INFO] - Iteration 32 took 51s (31.98% Gen, 66.31% Train). Generation: 16s, Training: 34s. Estimated remaining time: 42h 45m 52s. Estimated total time: 43h 15m 12s. Time estimates for 10 more iterations: 8m 39s, 100 more iterations: 1h 26m 30s, 500 more iterations: 7h 12m 32s. -[2025-11-09 18:28:17,553][__main__][INFO] - Starting iteration 32. -[2025-11-09 18:28:18,018][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 3 and human policies 1. -[2025-11-09 18:28:18,018][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:28:21,145][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:28:32,052][__main__][INFO] - Number of regex retries in iteration 32: 1 -[2025-11-09 18:28:32,052][__main__][INFO] - agents played in iteration 32 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:28:32,821][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:28:32,848][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:28:32,874][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:28:32,897][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:28:32,897][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:28:32,898][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:28:33,492][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:28:33,945][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:28:34,445][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:28:34,940][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:28:35,441][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:28:35,938][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:28:36,432][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:28:36,927][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:28:37,421][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:28:37,915][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:28:38,411][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:28:38,905][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:28:39,416][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:28:39,909][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:28:40,406][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:28:40,898][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:28:41,390][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:28:41,883][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:28:42,378][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:28:42,870][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:28:43,366][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:28:43,864][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:28:44,359][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:28:44,853][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:28:45,346][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:28:45,841][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:28:46,335][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:28:46,829][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:28:47,326][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:28:47,818][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:28:48,320][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:28:48,813][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:28:49,308][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:28:49,801][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:28:50,296][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:28:50,790][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:28:51,285][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:28:51,785][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:28:52,282][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:28:52,776][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:28:53,269][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:28:53,763][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:28:54,257][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:28:54,750][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:28:55,245][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:28:55,738][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:28:56,236][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:28:56,732][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:28:57,225][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:28:57,719][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:28:58,212][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:28:58,709][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:28:59,213][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:28:59,707][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:29:00,203][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:29:00,708][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:29:01,206][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:29:01,703][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:29:02,199][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:29:02,696][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:29:03,196][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:29:03,696][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:29:04,194][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:29:04,693][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:29:05,190][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9916 tokens. -[2025-11-09 18:29:05,902][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.04%, Current % of VRAM taken: 55.53%, Block Peak % of device VRAM: 61.90%, ΔTime: 00:00:32 -[2025-11-09 18:29:06,611][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:29:06,613][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:29:06,615][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:29:07,506][__main__][INFO] - Iteration 33 took 49s (28.36% Gen, 69.84% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 44m 16s. Estimated total time: 41h 14m 27s. Time estimates for 10 more iterations: 8m 14s, 100 more iterations: 1h 22m 28s, 500 more iterations: 6h 52m 24s. -[2025-11-09 18:29:07,508][__main__][INFO] - Starting iteration 33. -[2025-11-09 18:29:07,970][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 3 and human policies 1. -[2025-11-09 18:29:07,971][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:29:09,286][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:29:10,602][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:29:15,587][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls +Since I value hats most and Bob values them least, I can maximize my points by taking all the hats. This eliminates competition for the hats which are my highest value items. I will not propose to take any books as my value for them is equal to Bob's value, and I can secure all the hats. I will also not propose to take any balls as my value for them is much lower than Bob's value, and taking just the quantity I propose would be a waste since the remaining balls would be allocated to Bob based on his higher value. did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 21:30:02,678][__main__][INFO] - Number of regex retries in iteration 26: 2 +[2026-03-25 21:30:02,679][__main__][INFO] - agents played in iteration 26 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:30:03,471][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.44%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:30:03,495][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.44%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:30:03,518][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.44%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:30:03,541][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.44%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:30:03,542][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:30:03,542][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:30:04,146][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:30:04,599][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:30:05,102][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:30:05,603][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:30:06,102][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:30:06,602][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:30:07,097][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:30:07,593][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:30:08,090][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:30:08,586][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:30:09,086][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:30:09,582][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:30:10,077][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:30:10,572][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:30:11,067][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:30:11,562][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:30:12,058][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:30:12,554][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:30:13,049][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:30:13,546][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:30:14,045][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:30:14,542][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:30:15,039][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:30:15,547][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:30:16,043][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:30:16,539][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:30:17,035][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:30:17,533][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:30:18,030][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:30:18,527][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:30:19,025][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:30:19,526][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:30:20,024][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:30:20,518][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:30:21,010][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:30:21,504][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:30:21,996][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:30:22,491][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:30:22,985][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:30:23,477][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:30:23,970][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:30:24,467][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:30:24,962][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:30:25,456][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:30:25,950][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:30:26,444][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:30:26,939][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:30:27,433][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:30:27,929][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:30:28,424][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:30:28,920][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:30:29,415][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:30:29,911][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:30:30,406][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:30:30,903][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:30:31,398][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:30:31,894][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:30:32,390][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:30:32,885][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:30:33,380][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:30:33,875][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:30:34,371][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:30:34,866][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:30:35,364][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:30:35,863][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9820 tokens. +[2026-03-25 21:30:36,591][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.05%, Current % of VRAM taken: 60.47%, Block Peak % of device VRAM: 61.92%, ΔTime: 00:00:32 +[2026-03-25 21:30:37,312][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:30:37,315][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:30:37,319][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:30:37,954][__main__][INFO] - Iteration 27 took 50s (30.33% Gen, 68.41% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 46m 37s. Estimated total time: 42h 11m 35s. Time estimates for 10 more iterations: 8m 26s, 100 more iterations: 1h 24m 23s, 500 more iterations: 7h 1m 55s. +[2026-03-25 21:30:37,956][__main__][INFO] - Starting iteration 27. +[2026-03-25 21:30:38,359][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 2 and human policies 1. +[2026-03-25 21:30:38,360][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:30:40,302][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 21:30:49,340][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls -This proposal takes into account the significant value discrepancy between the items. Given the values, books and balls have a much higher value than hats. However, since Bob has a much higher value for books and balls, it would be more beneficial to keep the hats and offer nothing for the books and balls. This proposal aims to secure the maximum points from the hats, which have a lower value for you but a much higher value for you compared to the books and balls. did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:29:23,355][__main__][INFO] - Number of regex retries in iteration 33: 3 -[2025-11-09 18:29:23,356][__main__][INFO] - agents played in iteration 33 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:29:24,152][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:29:24,179][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:29:24,205][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:29:24,228][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:29:24,228][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:29:24,229][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:29:24,823][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:29:25,277][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:29:25,776][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:29:26,273][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:29:26,770][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:29:27,269][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:29:27,762][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:29:28,256][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:29:28,760][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:29:29,270][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:29:29,766][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:29:30,263][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:29:30,756][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:29:31,250][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:29:31,755][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:29:32,249][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:29:32,744][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:29:33,237][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:29:33,732][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:29:34,225][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:29:34,718][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:29:35,211][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:29:35,704][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:29:36,199][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:29:36,693][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:29:37,187][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:29:37,682][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:29:38,176][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:29:38,671][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:29:39,167][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:29:39,666][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:29:40,159][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:29:40,654][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:29:41,150][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:29:41,643][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:29:42,136][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:29:42,635][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:29:43,128][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:29:43,633][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:29:44,128][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:29:44,624][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:29:45,117][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:29:45,611][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:29:46,106][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:29:46,601][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:29:47,094][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:29:47,589][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:29:48,083][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:29:48,576][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:29:49,067][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:29:49,561][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:29:50,056][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:29:50,549][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:29:51,042][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:29:51,534][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:29:52,029][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:29:52,525][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:29:53,036][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:29:53,534][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:29:54,033][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:29:54,555][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:29:55,052][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:29:55,552][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:29:56,047][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:29:56,545][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9996 tokens. -[2025-11-09 18:29:57,230][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.02%, Current % of VRAM taken: 55.51%, Block Peak % of device VRAM: 61.97%, ΔTime: 00:00:32 -[2025-11-09 18:29:57,951][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:29:57,952][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:29:57,959][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:29:59,280][__main__][INFO] - Iteration 34 took 51s (29.98% Gen, 67.44% Train). Generation: 15s, Training: 34s. Estimated remaining time: 42h 14m 28s. Estimated total time: 42h 45m 31s. Time estimates for 10 more iterations: 8m 33s, 100 more iterations: 1h 25m 31s, 500 more iterations: 7h 7m 35s. -[2025-11-09 18:29:59,282][__main__][INFO] - Starting iteration 34. -[2025-11-09 18:29:59,772][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 3 and human policies 1. -[2025-11-09 18:29:59,773][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:30:11,056][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:30:14,176][__main__][INFO] - Number of regex retries in iteration 34: 1 -[2025-11-09 18:30:14,177][__main__][INFO] - agents played in iteration 34 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:30:14,963][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:30:14,986][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:30:15,008][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:30:15,029][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:30:15,030][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:30:15,031][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:30:15,630][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:30:16,079][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:30:16,580][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:30:17,075][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:30:17,569][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:30:18,063][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:30:18,557][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:30:19,050][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:30:19,559][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:30:20,053][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:30:20,546][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:30:21,038][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:30:21,531][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:30:22,041][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:30:22,537][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:30:23,029][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:30:23,533][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:30:24,026][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:30:24,517][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:30:25,010][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:30:25,505][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:30:26,001][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:30:26,493][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:30:26,986][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:30:27,482][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:30:27,976][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:30:28,468][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:30:28,963][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:30:29,462][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:30:29,968][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:30:30,464][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:30:30,957][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:30:31,452][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:30:31,946][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:30:32,437][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:30:32,930][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:30:33,423][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:30:33,914][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:30:34,406][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:30:34,900][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:30:35,392][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:30:35,890][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:30:36,386][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:30:36,880][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:30:37,378][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:30:37,871][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:30:38,368][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:30:38,865][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:30:39,364][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:30:39,855][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:30:40,348][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:30:40,845][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:30:41,336][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:30:41,829][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:30:42,323][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:30:42,816][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:30:43,315][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:30:43,812][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:30:44,311][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:30:44,814][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:30:45,312][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:30:45,809][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:30:46,307][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:30:46,804][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:30:47,306][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9950 tokens. -[2025-11-09 18:30:47,955][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.00%, Current % of VRAM taken: 55.50%, Block Peak % of device VRAM: 62.00%, ΔTime: 00:00:32 -[2025-11-09 18:30:48,730][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:30:48,732][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:30:48,740][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:30:49,697][__main__][INFO] - Iteration 35 took 49s (28.85% Gen, 69.23% Train). Generation: 14s, Training: 34s. Estimated remaining time: 41h 4m 23s. Estimated total time: 41h 36m 16s. Time estimates for 10 more iterations: 8m 19s, 100 more iterations: 1h 23m 12s, 500 more iterations: 6h 56m 2s. -[2025-11-09 18:30:49,699][__main__][INFO] - Starting iteration 35. -[2025-11-09 18:30:50,175][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 3 and human policies 1. -[2025-11-09 18:30:50,175][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:30:51,492][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:30:51,680][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:31:04,661][__main__][INFO] - Number of regex retries in iteration 35: 2 -[2025-11-09 18:31:04,662][__main__][INFO] - agents played in iteration 35 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:31:05,437][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:31:05,465][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:31:05,491][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:31:05,514][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:31:05,514][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:31:05,515][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:31:06,105][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:31:06,557][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:31:07,071][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:31:07,566][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:31:08,059][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:31:08,556][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:31:09,048][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:31:09,541][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:31:10,033][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:31:10,525][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:31:11,016][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:31:11,509][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:31:12,005][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:31:12,497][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:31:12,990][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:31:13,485][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:31:13,978][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:31:14,473][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:31:14,964][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:31:15,458][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:31:15,961][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:31:16,454][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:31:16,948][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:31:17,440][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:31:17,932][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:31:18,427][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:31:18,935][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:31:19,428][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:31:19,922][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:31:20,425][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:31:20,919][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:31:21,411][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:31:21,905][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:31:22,399][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:31:22,892][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:31:23,386][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:31:23,883][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:31:24,377][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:31:24,871][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:31:25,367][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:31:25,861][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:31:26,355][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:31:26,848][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:31:27,342][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:31:27,850][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:31:28,344][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:31:28,837][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:31:29,341][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:31:29,834][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:31:30,327][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:31:30,831][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:31:31,323][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:31:31,814][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:31:32,305][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:31:32,799][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:31:33,296][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:31:33,791][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:31:34,290][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:31:34,789][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:31:35,286][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:31:35,788][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:31:36,287][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:31:36,790][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:31:37,293][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:31:37,790][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10029 tokens. -[2025-11-09 18:31:38,450][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 55.60%, Block Peak % of device VRAM: 62.15%, ΔTime: 00:00:32 -[2025-11-09 18:31:39,181][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:31:39,182][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:31:39,184][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:31:40,261][__main__][INFO] - Iteration 36 took 50s (28.92% Gen, 68.92% Train). Generation: 14s, Training: 34s. Estimated remaining time: 41h 11m 40s. Estimated total time: 41h 44m 24s. Time estimates for 10 more iterations: 8m 20s, 100 more iterations: 1h 23m 28s, 500 more iterations: 6h 57m 24s. -[2025-11-09 18:31:40,263][__main__][INFO] - Starting iteration 36. -[2025-11-09 18:31:40,749][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 3 and human policies 1. -[2025-11-09 18:31:40,749][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:31:54,648][__main__][INFO] - Number of regex retries in iteration 36: 0 -[2025-11-09 18:31:54,648][__main__][INFO] - agents played in iteration 36 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:31:55,417][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:31:55,442][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:31:55,467][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:31:55,500][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:31:55,500][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:31:55,502][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:31:56,088][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:31:56,537][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:31:57,051][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:31:57,545][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:31:58,037][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:31:58,533][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:31:59,027][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:31:59,519][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:32:00,012][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:32:00,506][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:32:01,003][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:32:01,495][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:32:01,987][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:32:02,479][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:32:02,971][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:32:03,464][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:32:03,955][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:32:04,447][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:32:04,938][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:32:05,429][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:32:05,922][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:32:06,419][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:32:06,911][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:32:07,404][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:32:07,895][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:32:08,387][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:32:08,886][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:32:09,377][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:32:09,868][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:32:10,360][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:32:10,852][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:32:11,346][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:32:11,839][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:32:12,332][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:32:12,826][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:32:13,318][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:32:13,812][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:32:14,306][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:32:14,797][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:32:15,295][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:32:15,787][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:32:16,279][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:32:16,770][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:32:17,261][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:32:17,753][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:32:18,244][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:32:18,736][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:32:19,229][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:32:19,722][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:32:20,218][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:32:20,713][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:32:21,207][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:32:21,699][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:32:22,195][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:32:22,691][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:32:23,188][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:32:23,685][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:32:24,181][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:32:24,688][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:32:25,184][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:32:25,681][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:32:26,181][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:32:26,678][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:32:27,174][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:32:27,670][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9952 tokens. -[2025-11-09 18:32:28,381][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.03%, Current % of VRAM taken: 55.53%, Block Peak % of device VRAM: 61.88%, ΔTime: 00:00:32 -[2025-11-09 18:32:29,084][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:32:29,088][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:32:29,091][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:32:30,017][__main__][INFO] - Iteration 37 took 49s (28.21% Gen, 69.91% Train). Generation: 13s, Training: 34s. Estimated remaining time: 40h 29m 52s. Estimated total time: 41h 3m 25s. Time estimates for 10 more iterations: 8m 12s, 100 more iterations: 1h 22m 6s, 500 more iterations: 6h 50m 34s. -[2025-11-09 18:32:30,019][__main__][INFO] - Starting iteration 37. -[2025-11-09 18:32:30,482][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 3 and human policies 1. -[2025-11-09 18:32:30,484][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:32:31,770][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:32:44,669][__main__][INFO] - Number of regex retries in iteration 37: 1 -[2025-11-09 18:32:44,670][__main__][INFO] - agents played in iteration 37 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:32:45,455][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:32:45,480][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:32:45,505][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:32:45,527][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:32:45,528][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:32:45,529][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:32:46,117][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:32:46,567][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:32:47,065][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:32:47,559][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:32:48,052][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:32:48,546][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:32:49,036][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:32:49,530][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:32:50,021][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:32:50,517][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:32:51,010][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:32:51,505][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:32:52,020][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:32:52,514][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:32:53,006][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:32:53,498][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:32:53,990][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:32:54,483][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:32:54,979][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:32:55,472][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:32:55,979][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:32:56,471][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:32:56,963][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:32:57,468][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:32:57,961][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:32:58,458][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:32:58,952][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:32:59,446][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:32:59,944][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:33:00,444][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:33:00,936][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:33:01,428][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:33:01,926][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:33:02,421][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:33:02,914][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:33:03,411][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:33:03,906][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:33:04,401][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:33:04,893][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:33:05,391][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:33:05,889][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:33:06,391][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:33:06,884][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:33:07,376][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:33:07,869][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:33:08,361][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:33:08,852][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:33:09,345][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:33:09,837][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:33:10,329][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:33:10,822][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:33:11,315][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:33:11,809][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:33:12,304][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:33:12,794][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:33:13,288][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:33:13,786][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:33:14,307][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:33:14,802][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:33:15,297][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:33:15,793][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:33:16,289][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:33:16,797][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:33:17,294][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:33:17,792][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9942 tokens. -[2025-11-09 18:33:18,505][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.04%, Current % of VRAM taken: 55.54%, Block Peak % of device VRAM: 61.99%, ΔTime: 00:00:32 -[2025-11-09 18:33:19,252][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:33:19,254][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:33:19,256][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:33:20,142][__main__][INFO] - Iteration 38 took 49s (28.57% Gen, 69.64% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 48m 38s. Estimated total time: 41h 23m 2s. Time estimates for 10 more iterations: 8m 16s, 100 more iterations: 1h 22m 46s, 500 more iterations: 6h 53m 50s. -[2025-11-09 18:33:20,144][__main__][INFO] - Starting iteration 38. -[2025-11-09 18:33:20,612][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 3 and human policies 1. -[2025-11-09 18:33:20,613][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:33:34,793][__main__][INFO] - Number of regex retries in iteration 38: 0 -[2025-11-09 18:33:34,794][__main__][INFO] - agents played in iteration 38 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:33:35,565][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:33:35,594][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:33:35,621][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:33:35,644][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:33:35,645][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:33:35,645][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:33:36,255][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:33:36,707][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:33:37,210][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:33:37,705][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:33:38,198][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:33:38,697][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:33:39,193][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:33:39,689][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:33:40,182][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:33:40,675][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:33:41,169][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:33:41,665][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:33:42,162][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:33:42,656][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:33:43,151][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:33:43,646][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:33:44,139][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:33:44,631][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:33:45,123][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:33:45,614][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:33:46,107][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:33:46,600][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:33:47,094][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:33:47,589][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:33:48,085][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:33:48,577][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:33:49,068][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:33:49,567][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:33:50,059][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:33:50,551][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:33:51,046][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:33:51,538][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:33:52,032][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:33:52,527][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:33:53,020][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:33:53,511][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:33:54,005][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:33:54,508][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:33:55,003][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:33:55,494][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:33:55,987][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:33:56,478][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:33:56,971][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:33:57,472][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:33:57,966][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:33:58,458][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:33:58,949][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:33:59,442][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:33:59,936][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:34:00,430][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:34:00,923][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:34:01,413][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:34:01,905][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:34:02,397][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:34:02,889][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:34:03,386][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:34:03,892][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:34:04,388][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:34:04,887][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:34:05,392][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:34:05,890][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:34:06,390][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:34:06,887][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:34:07,385][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:34:07,888][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9929 tokens. -[2025-11-09 18:34:08,609][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 55.61%, Block Peak % of device VRAM: 61.81%, ΔTime: 00:00:32 -[2025-11-09 18:34:09,370][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:34:09,372][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:34:09,376][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:34:10,265][__main__][INFO] - Iteration 39 took 49s (28.56% Gen, 69.65% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 47m 28s. Estimated total time: 41h 22m 41s. Time estimates for 10 more iterations: 8m 16s, 100 more iterations: 1h 22m 45s, 500 more iterations: 6h 53m 46s. -[2025-11-09 18:34:10,267][__main__][INFO] - Starting iteration 39. -[2025-11-09 18:34:10,730][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 3 and human policies 1. -[2025-11-09 18:34:10,730][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:34:24,696][__main__][INFO] - Number of regex retries in iteration 39: 0 -[2025-11-09 18:34:24,697][__main__][INFO] - agents played in iteration 39 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:34:25,468][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:34:25,496][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:34:25,522][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:34:25,544][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:34:25,545][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:34:25,546][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:34:26,132][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:34:26,583][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:34:27,083][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:34:27,578][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:34:28,072][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:34:28,572][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:34:29,068][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:34:29,562][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:34:30,056][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:34:30,549][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:34:31,042][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:34:31,537][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:34:32,032][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:34:32,528][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:34:33,020][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:34:33,516][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:34:34,008][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:34:34,500][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:34:34,992][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:34:35,488][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:34:35,981][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:34:36,475][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:34:36,968][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:34:37,461][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:34:37,956][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:34:38,454][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:34:38,948][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:34:39,451][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:34:39,971][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:34:40,467][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:34:40,960][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:34:41,456][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:34:41,952][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:34:42,444][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:34:42,935][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:34:43,425][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:34:43,922][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:34:44,416][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:34:44,910][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:34:45,402][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:34:45,893][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:34:46,387][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:34:46,879][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:34:47,372][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:34:47,865][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:34:48,358][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:34:48,851][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:34:49,344][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:34:49,836][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:34:50,328][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:34:50,819][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:34:51,313][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:34:51,805][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:34:52,294][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:34:52,785][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:34:53,276][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:34:53,767][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:34:54,262][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:34:54,758][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:34:55,254][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:34:55,750][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:34:56,250][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:34:56,749][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:34:57,247][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:34:57,749][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9948 tokens. -[2025-11-09 18:34:58,435][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 55.59%, Block Peak % of device VRAM: 62.24%, ΔTime: 00:00:32 -[2025-11-09 18:34:59,157][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:34:59,163][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:34:59,164][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:35:00,126][__main__][INFO] - Iteration 40 took 49s (28.27% Gen, 69.78% Train). Generation: 13s, Training: 34s. Estimated remaining time: 40h 33m 48s. Estimated total time: 41h 9m 52s. Time estimates for 10 more iterations: 8m 13s, 100 more iterations: 1h 22m 19s, 500 more iterations: 6h 51m 38s. -[2025-11-09 18:35:00,128][__main__][INFO] - Starting iteration 40. -[2025-11-09 18:35:00,590][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 3 and human policies 1. -[2025-11-09 18:35:00,591][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:35:14,869][__main__][INFO] - Number of regex retries in iteration 40: 0 -[2025-11-09 18:35:14,870][__main__][INFO] - agents played in iteration 40 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:35:15,642][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:35:15,666][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:35:15,691][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:35:15,713][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:35:15,714][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:35:15,714][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:35:16,329][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:35:16,780][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:35:17,281][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:35:17,775][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:35:18,269][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:35:18,768][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:35:19,267][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:35:19,760][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:35:20,256][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:35:20,750][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:35:21,248][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:35:21,742][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:35:22,234][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:35:22,727][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:35:23,224][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:35:23,716][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:35:24,210][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:35:24,703][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:35:25,197][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:35:25,691][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:35:26,188][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:35:26,680][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:35:27,172][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:35:27,666][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:35:28,160][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:35:28,662][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:35:29,159][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:35:29,654][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:35:30,149][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:35:30,649][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:35:31,142][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:35:31,641][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:35:32,138][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:35:32,633][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:35:33,136][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:35:33,627][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:35:34,117][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:35:34,609][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:35:35,099][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:35:35,592][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:35:36,086][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:35:36,579][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:35:37,074][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:35:37,569][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:35:38,064][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:35:38,560][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:35:39,054][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:35:39,550][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:35:40,043][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:35:40,537][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:35:41,045][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:35:41,536][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:35:42,030][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:35:42,527][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:35:43,019][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:35:43,518][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:35:44,013][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:35:44,508][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:35:45,012][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:35:45,507][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:35:46,001][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:35:46,505][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:35:47,002][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:35:47,509][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:35:48,006][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9972 tokens. -[2025-11-09 18:35:48,727][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.03%, Current % of VRAM taken: 55.52%, Block Peak % of device VRAM: 61.99%, ΔTime: 00:00:32 -[2025-11-09 18:35:49,451][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:35:49,453][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:35:49,454][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:35:52,413][__main__][INFO] - Iteration 41 took 51s (27.55% Gen, 66.73% Train). Generation: 14s, Training: 34s. Estimated remaining time: 42h 34m 16s. Estimated total time: 43h 11m 11s. Time estimates for 10 more iterations: 8m 38s, 100 more iterations: 1h 26m 22s, 500 more iterations: 7h 11m 51s. -[2025-11-09 18:35:52,415][__main__][INFO] - Starting iteration 41. -[2025-11-09 18:35:52,879][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 4 and human policies 1. -[2025-11-09 18:35:52,880][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:35:54,770][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:36:07,277][__main__][INFO] - Number of regex retries in iteration 41: 1 -[2025-11-09 18:36:07,278][__main__][INFO] - agents played in iteration 41 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:36:08,042][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:36:08,070][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:36:08,096][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:36:08,119][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:36:08,119][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:36:08,120][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:36:08,716][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:36:09,168][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:36:09,670][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:36:10,170][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:36:10,666][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:36:11,168][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:36:11,666][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:36:12,161][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:36:12,654][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:36:13,152][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:36:13,654][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:36:14,149][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:36:14,641][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:36:15,147][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:36:15,640][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:36:16,130][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:36:16,623][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:36:17,119][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:36:17,612][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:36:18,108][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:36:18,605][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:36:19,098][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:36:19,593][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:36:20,087][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:36:20,579][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:36:21,072][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:36:21,566][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:36:22,062][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:36:22,554][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:36:23,051][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:36:23,545][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:36:24,036][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:36:24,529][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:36:25,024][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:36:25,521][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:36:26,014][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:36:26,508][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:36:27,001][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:36:27,495][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:36:27,990][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:36:28,485][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:36:28,978][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:36:29,471][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:36:29,963][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:36:30,459][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:36:30,951][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:36:31,446][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:36:31,938][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:36:32,429][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:36:32,921][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:36:33,419][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:36:33,911][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:36:34,409][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:36:34,903][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:36:35,402][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:36:35,901][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:36:36,398][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:36:36,892][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:36:37,391][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:36:37,884][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:36:38,380][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:36:38,879][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:36:39,375][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:36:39,872][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:36:40,369][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10084 tokens. -[2025-11-09 18:36:41,072][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.02%, Current % of VRAM taken: 55.51%, Block Peak % of device VRAM: 62.16%, ΔTime: 00:00:32 -[2025-11-09 18:36:41,767][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:36:41,768][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:36:41,770][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:36:42,646][__main__][INFO] - Iteration 42 took 49s (28.93% Gen, 69.31% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 50m 37s. Estimated total time: 41h 28m 23s. Time estimates for 10 more iterations: 8m 17s, 100 more iterations: 1h 22m 56s, 500 more iterations: 6h 54m 43s. -[2025-11-09 18:36:42,648][__main__][INFO] - Starting iteration 42. -[2025-11-09 18:36:43,133][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 4 and human policies 1. -[2025-11-09 18:36:43,133][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:36:57,653][__main__][INFO] - Number of regex retries in iteration 42: 0 -[2025-11-09 18:36:57,653][__main__][INFO] - agents played in iteration 42 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:36:58,438][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:36:58,466][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:36:58,491][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:36:58,513][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:36:58,514][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:36:58,515][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:36:59,112][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:36:59,563][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:37:00,067][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:37:00,560][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:37:01,054][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:37:01,549][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:37:02,046][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:37:02,545][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:37:03,037][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:37:03,533][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:37:04,028][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:37:04,524][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:37:05,017][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:37:05,508][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:37:06,003][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:37:06,499][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:37:06,993][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:37:07,486][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:37:07,979][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:37:08,473][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:37:08,963][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:37:09,454][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:37:09,947][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:37:10,440][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:37:10,935][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:37:11,432][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:37:11,927][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:37:12,421][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:37:12,914][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:37:13,407][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:37:13,901][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:37:14,398][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:37:14,891][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:37:15,390][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:37:15,887][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:37:16,383][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:37:16,873][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:37:17,368][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:37:17,869][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:37:18,361][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:37:18,856][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:37:19,352][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:37:19,845][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:37:20,338][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:37:20,833][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:37:21,327][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:37:21,822][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:37:22,314][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:37:22,806][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:37:23,296][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:37:23,787][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:37:24,280][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:37:24,767][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:37:25,257][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:37:25,747][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:37:26,238][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:37:26,729][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:37:27,236][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:37:27,733][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:37:28,228][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:37:28,730][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:37:29,229][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:37:29,728][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:37:30,228][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:37:30,727][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9996 tokens. -[2025-11-09 18:37:31,441][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 55.62%, Block Peak % of device VRAM: 61.96%, ΔTime: 00:00:32 -[2025-11-09 18:37:32,138][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:37:32,139][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:37:32,140][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:37:33,159][__main__][INFO] - Iteration 43 took 50s (29.02% Gen, 68.94% Train). Generation: 14s, Training: 34s. Estimated remaining time: 41h 2m 43s. Estimated total time: 41h 41m 19s. Time estimates for 10 more iterations: 8m 20s, 100 more iterations: 1h 23m 22s, 500 more iterations: 6h 56m 53s. -[2025-11-09 18:37:33,161][__main__][INFO] - Starting iteration 43. -[2025-11-09 18:37:33,635][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 4 and human policies 1. -[2025-11-09 18:37:33,635][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:37:48,273][__main__][INFO] - Number of regex retries in iteration 43: 0 -[2025-11-09 18:37:48,274][__main__][INFO] - agents played in iteration 43 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:37:49,051][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:37:49,074][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:37:49,109][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:37:49,131][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:37:49,132][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:37:49,133][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:37:49,720][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:37:50,170][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:37:50,684][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:37:51,181][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:37:51,674][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:37:52,169][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:37:52,666][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:37:53,173][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:37:53,666][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:37:54,165][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:37:54,660][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:37:55,153][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:37:55,648][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:37:56,142][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:37:56,636][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:37:57,129][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:37:57,622][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:37:58,114][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:37:58,607][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:37:59,100][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:37:59,598][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:38:00,091][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:38:00,586][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:38:01,079][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:38:01,572][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:38:02,065][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:38:02,562][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:38:03,056][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:38:03,548][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:38:04,039][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:38:04,530][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:38:05,023][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:38:05,516][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:38:06,014][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:38:06,509][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:38:07,000][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:38:07,495][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:38:07,988][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:38:08,482][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:38:08,986][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:38:09,480][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:38:09,976][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:38:10,471][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:38:10,965][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:38:11,458][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:38:11,952][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:38:12,443][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:38:12,933][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:38:13,425][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:38:13,916][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:38:14,409][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:38:14,901][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:38:15,393][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:38:15,886][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:38:16,377][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:38:16,869][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:38:17,360][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:38:17,882][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:38:18,378][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:38:18,876][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:38:19,376][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:38:19,873][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:38:20,376][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:38:20,875][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:38:21,374][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10062 tokens. -[2025-11-09 18:38:22,069][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.09%, Current % of VRAM taken: 55.59%, Block Peak % of device VRAM: 62.03%, ΔTime: 00:00:32 -[2025-11-09 18:38:22,801][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:38:22,802][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:38:22,804][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:38:23,683][__main__][INFO] - Iteration 44 took 50s (29.25% Gen, 68.99% Train). Generation: 14s, Training: 34s. Estimated remaining time: 41h 3m 0s. Estimated total time: 41h 42m 26s. Time estimates for 10 more iterations: 8m 20s, 100 more iterations: 1h 23m 24s, 500 more iterations: 6h 57m 4s. -[2025-11-09 18:38:23,685][__main__][INFO] - Starting iteration 44. -[2025-11-09 18:38:24,150][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 4 and human policies 1. -[2025-11-09 18:38:24,151][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:38:25,655][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:38:38,221][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:38:39,095][__main__][INFO] - Number of regex retries in iteration 44: 2 -[2025-11-09 18:38:39,096][__main__][INFO] - agents played in iteration 44 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:38:39,876][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:38:39,898][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:38:39,921][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:38:39,943][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:38:39,943][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:38:39,944][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:38:40,542][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:38:40,997][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:38:41,498][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:38:42,004][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:38:42,500][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:38:42,996][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:38:43,499][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:38:43,995][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:38:44,490][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:38:44,984][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:38:45,478][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:38:45,989][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:38:46,483][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:38:46,977][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:38:47,475][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:38:47,973][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:38:48,467][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:38:48,962][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:38:49,457][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:38:49,956][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:38:50,453][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:38:50,947][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:38:51,440][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:38:51,933][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:38:52,427][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:38:52,931][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:38:53,425][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:38:53,925][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:38:54,427][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:38:54,943][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:38:55,439][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:38:55,934][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:38:56,431][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:38:56,926][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:38:57,421][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:38:57,916][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:38:58,411][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:38:58,905][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:38:59,403][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:38:59,898][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:39:00,407][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:39:00,900][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:39:01,394][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:39:01,890][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:39:02,386][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:39:02,881][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:39:03,375][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:39:03,868][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:39:04,364][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:39:04,857][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:39:05,351][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:39:05,845][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:39:06,340][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:39:06,836][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:39:07,330][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:39:07,824][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:39:08,322][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:39:08,821][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:39:09,322][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:39:09,821][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:39:10,320][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:39:10,820][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:39:11,326][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:39:11,829][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:39:12,331][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10203 tokens. -[2025-11-09 18:39:13,015][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.18%, ΔTime: 00:00:32 -[2025-11-09 18:39:13,738][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:39:13,739][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:39:13,741][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:39:14,812][__main__][INFO] - Iteration 45 took 50s (29.50% Gen, 68.38% Train). Generation: 14s, Training: 34s. Estimated remaining time: 41h 32m 49s. Estimated total time: 42h 13m 7s. Time estimates for 10 more iterations: 8m 26s, 100 more iterations: 1h 24m 26s, 500 more iterations: 7h 2m 11s. -[2025-11-09 18:39:14,814][__main__][INFO] - Starting iteration 45. -[2025-11-09 18:39:15,276][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 4 and human policies 1. -[2025-11-09 18:39:15,276][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:39:30,160][__main__][INFO] - Number of regex retries in iteration 45: 0 -[2025-11-09 18:39:30,161][__main__][INFO] - agents played in iteration 45 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:39:30,944][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:39:30,971][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:39:30,997][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:39:31,020][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:39:31,021][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:39:31,022][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:39:31,619][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:39:32,091][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:39:32,590][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:39:33,085][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:39:33,584][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:39:34,078][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:39:34,580][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:39:35,082][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:39:35,576][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:39:36,074][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:39:36,569][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:39:37,066][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:39:37,561][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:39:38,056][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:39:38,551][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:39:39,048][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:39:39,541][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:39:40,043][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:39:40,540][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:39:41,036][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:39:41,538][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:39:42,035][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:39:42,534][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:39:43,030][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:39:43,528][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:39:44,020][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:39:44,514][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:39:45,008][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:39:45,503][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:39:46,000][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:39:46,515][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:39:47,008][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:39:47,502][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:39:47,997][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:39:48,491][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:39:48,993][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:39:49,487][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:39:49,980][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:39:50,476][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:39:50,969][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:39:51,466][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:39:51,961][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:39:52,455][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:39:52,950][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:39:53,443][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:39:53,938][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:39:54,432][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:39:54,924][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:39:55,419][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:39:55,915][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:39:56,408][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:39:56,911][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:39:57,407][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:39:57,905][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:39:58,401][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:39:58,899][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:39:59,400][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:39:59,903][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:40:00,398][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:40:00,896][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:40:01,392][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:40:01,892][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:40:02,410][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:40:02,910][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:40:03,410][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10265 tokens. -[2025-11-09 18:40:04,121][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 55.60%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 -[2025-11-09 18:40:04,834][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:40:04,835][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:40:04,837][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:40:05,713][__main__][INFO] - Iteration 46 took 50s (29.51% Gen, 68.75% Train). Generation: 14s, Training: 34s. Estimated remaining time: 41h 20m 44s. Estimated total time: 42h 1m 53s. Time estimates for 10 more iterations: 8m 24s, 100 more iterations: 1h 24m 3s, 500 more iterations: 7h 0m 18s. -[2025-11-09 18:40:05,715][__main__][INFO] - Starting iteration 46. -[2025-11-09 18:40:06,191][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 4 and human policies 1. -[2025-11-09 18:40:06,192][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:40:20,598][__main__][INFO] - Number of regex retries in iteration 46: 0 -[2025-11-09 18:40:20,599][__main__][INFO] - agents played in iteration 46 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:40:21,424][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:40:21,454][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:40:21,481][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:40:21,505][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:40:21,505][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:40:21,506][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:40:22,175][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:40:22,626][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:40:23,130][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:40:23,625][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:40:24,120][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:40:24,618][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:40:25,112][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:40:25,608][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:40:26,121][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:40:26,614][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:40:27,108][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:40:27,602][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:40:28,094][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:40:28,606][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:40:29,099][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:40:29,593][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:40:30,085][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:40:30,579][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:40:31,072][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:40:31,565][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:40:32,058][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:40:32,554][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:40:33,048][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:40:33,540][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:40:34,034][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:40:34,526][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:40:35,022][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:40:35,515][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:40:36,007][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:40:36,500][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:40:36,991][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:40:37,483][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:40:37,973][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:40:38,469][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:40:38,963][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:40:39,456][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:40:39,951][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:40:40,446][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:40:40,941][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:40:41,436][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:40:41,936][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:40:42,430][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:40:42,930][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:40:43,424][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:40:43,917][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:40:44,416][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:40:44,909][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:40:45,403][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:40:45,899][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:40:46,392][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:40:46,887][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:40:47,381][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:40:47,875][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:40:48,371][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:40:48,869][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:40:49,367][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:40:49,864][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:40:50,356][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:40:50,852][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:40:51,348][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:40:51,845][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:40:52,350][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:40:52,850][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:40:53,347][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:40:53,844][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9967 tokens. -[2025-11-09 18:40:54,534][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 55.56%, Block Peak % of device VRAM: 62.07%, ΔTime: 00:00:32 -[2025-11-09 18:40:55,272][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:40:55,274][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:40:55,276][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:40:56,178][__main__][INFO] - Iteration 47 took 49s (28.82% Gen, 69.37% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 57m 23s. Estimated total time: 41h 39m 22s. Time estimates for 10 more iterations: 8m 19s, 100 more iterations: 1h 23m 18s, 500 more iterations: 6h 56m 33s. -[2025-11-09 18:40:56,181][__main__][INFO] - Starting iteration 47. -[2025-11-09 18:40:56,655][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 4 and human policies 1. -[2025-11-09 18:40:56,655][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:41:10,869][__main__][INFO] - Number of regex retries in iteration 47: 0 -[2025-11-09 18:41:10,869][__main__][INFO] - agents played in iteration 47 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:41:11,665][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:41:11,690][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:41:11,715][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:41:11,737][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:41:11,738][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:41:11,738][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:41:12,391][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:41:12,846][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:41:13,349][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:41:13,843][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:41:14,338][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:41:14,836][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:41:15,336][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:41:15,829][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:41:16,327][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:41:16,821][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:41:17,317][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:41:17,813][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:41:18,306][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:41:18,801][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:41:19,295][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:41:19,791][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:41:20,287][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:41:20,783][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:41:21,289][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:41:21,788][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:41:22,284][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:41:22,781][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:41:23,278][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:41:23,776][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:41:24,272][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:41:24,766][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:41:25,260][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:41:25,755][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:41:26,252][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:41:26,747][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:41:27,244][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:41:27,739][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:41:28,232][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:41:28,726][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:41:29,220][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:41:29,715][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:41:30,221][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:41:30,714][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:41:31,207][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:41:31,717][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:41:32,210][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:41:32,704][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:41:33,202][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:41:33,696][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:41:34,190][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:41:34,694][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:41:35,187][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:41:35,681][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:41:36,173][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:41:36,671][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:41:37,168][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:41:37,663][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:41:38,155][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:41:38,647][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:41:39,140][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:41:39,639][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:41:40,135][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:41:40,630][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:41:41,130][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:41:41,630][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:41:42,128][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:41:42,622][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:41:43,118][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:41:43,628][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:41:44,123][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10193 tokens. -[2025-11-09 18:41:44,796][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 55.57%, Block Peak % of device VRAM: 62.09%, ΔTime: 00:00:32 -[2025-11-09 18:41:45,531][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:41:45,533][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:41:45,535][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:41:46,410][__main__][INFO] - Iteration 48 took 49s (28.57% Gen, 69.67% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 44m 56s. Estimated total time: 41h 27m 46s. Time estimates for 10 more iterations: 8m 17s, 100 more iterations: 1h 22m 55s, 500 more iterations: 6h 54m 37s. -[2025-11-09 18:41:46,412][__main__][INFO] - Starting iteration 48. -[2025-11-09 18:41:46,876][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 4 and human policies 1. -[2025-11-09 18:41:46,876][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:42:00,941][__main__][INFO] - Number of regex retries in iteration 48: 0 -[2025-11-09 18:42:00,941][__main__][INFO] - agents played in iteration 48 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:42:01,718][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:42:01,740][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:42:01,762][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:42:01,784][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:42:01,784][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:42:01,785][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:42:02,436][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:42:03,039][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:42:03,544][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:42:04,046][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:42:04,547][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:42:05,041][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:42:05,536][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:42:06,031][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:42:06,528][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:42:07,023][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:42:07,516][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:42:08,010][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:42:08,501][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:42:08,995][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:42:09,487][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:42:09,984][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:42:10,477][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:42:10,972][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:42:11,467][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:42:11,962][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:42:12,457][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:42:12,950][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:42:13,444][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:42:13,948][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:42:14,442][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:42:14,936][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:42:15,429][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:42:15,922][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:42:16,415][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:42:16,909][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:42:17,402][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:42:17,906][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:42:18,399][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:42:18,894][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:42:19,395][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:42:19,889][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:42:20,385][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:42:20,882][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:42:21,376][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:42:21,871][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:42:22,367][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:42:22,861][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:42:23,354][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:42:23,849][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:42:24,342][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:42:24,834][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:42:25,328][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:42:25,820][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:42:26,314][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:42:26,808][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:42:27,303][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:42:27,798][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:42:28,296][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:42:28,789][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:42:29,283][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:42:29,778][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:42:30,272][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:42:30,770][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:42:31,267][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:42:31,764][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:42:32,265][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:42:32,766][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:42:33,266][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:42:33,764][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:42:34,262][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10097 tokens. -[2025-11-09 18:42:34,950][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 55.58%, Block Peak % of device VRAM: 61.94%, ΔTime: 00:00:32 -[2025-11-09 18:42:35,666][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:42:35,668][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:42:35,671][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:42:36,609][__main__][INFO] - Iteration 49 took 49s (28.28% Gen, 69.83% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 43m 1s. Estimated total time: 41h 26m 40s. Time estimates for 10 more iterations: 8m 17s, 100 more iterations: 1h 22m 53s, 500 more iterations: 6h 54m 26s. -[2025-11-09 18:42:36,611][__main__][INFO] - Starting iteration 49. -[2025-11-09 18:42:37,076][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 4 and human policies 1. -[2025-11-09 18:42:37,077][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:42:39,354][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:42:51,968][__main__][INFO] - Number of regex retries in iteration 49: 1 -[2025-11-09 18:42:51,969][__main__][INFO] - agents played in iteration 49 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:42:52,808][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:42:52,834][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:42:52,858][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:42:52,881][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:42:52,881][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:42:52,882][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:42:53,540][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:42:53,996][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:42:54,498][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:42:54,996][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:42:55,492][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:42:55,993][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:42:56,490][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:42:56,988][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:42:57,487][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:42:57,983][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:42:58,481][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:42:58,977][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:42:59,476][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:42:59,975][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:43:00,467][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:43:00,961][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:43:01,457][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:43:01,949][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:43:02,450][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:43:02,943][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:43:03,437][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:43:03,928][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:43:04,421][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:43:04,916][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:43:05,413][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:43:05,907][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:43:06,405][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:43:06,897][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:43:07,390][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:43:07,886][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:43:08,379][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:43:08,874][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:43:09,369][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:43:09,862][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:43:10,362][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:43:10,858][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:43:11,350][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:43:11,843][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:43:12,335][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:43:12,837][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:43:13,330][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:43:13,823][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:43:14,315][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:43:14,809][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:43:15,303][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:43:15,798][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:43:16,291][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:43:16,788][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:43:17,280][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:43:17,776][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:43:18,271][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:43:18,767][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:43:19,260][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:43:19,756][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:43:20,251][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:43:20,751][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:43:21,247][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:43:21,743][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:43:22,237][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:43:22,731][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:43:23,228][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:43:23,730][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:43:24,225][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:43:24,726][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:43:25,221][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10176 tokens. -[2025-11-09 18:43:25,932][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 55.56%, Block Peak % of device VRAM: 62.12%, ΔTime: 00:00:32 -[2025-11-09 18:43:26,665][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:43:26,667][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:43:26,668][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:43:27,592][__main__][INFO] - Iteration 50 took 50s (29.48% Gen, 68.69% Train). Generation: 14s, Training: 34s. Estimated remaining time: 41h 21m 21s. Estimated total time: 42h 5m 52s. Time estimates for 10 more iterations: 8m 25s, 100 more iterations: 1h 24m 11s, 500 more iterations: 7h 0m 58s. -[2025-11-09 18:43:27,595][__main__][INFO] - Starting iteration 50. -[2025-11-09 18:43:28,069][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 4 and human policies 1. -[2025-11-09 18:43:28,070][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:43:29,514][mllm.models.large_language_model_local][WARNING] - Response Proposal: 5 hats, 5 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:43:36,289][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls +Given the scenario, I'll focus on the items that I value highly (hats and books) while considering Bob's valuations. Since I value hats significantly more than books and balls, and Bob values books more than hats, it makes sense to maximize my share of the hats. Books are equally valued between us, but since there are only 10 books and they are not as valuable to me as hats, it's better to allocate them proportionally to their actual value. -Given the values, it makes sense to prioritize the items where the difference in perception is the highest. Hats are valued at 10 by me and only 1 by Bob. Books are valued at 10 by both, but I value them less compared to their low value to Bob. Balls are valued at 1 by everyone, so they offer no advantage in this split. By proposing to take all the hats, I aim to exploit the significant disparity in Bob's valuation to maximize my points for this round. did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:43:43,422][__main__][INFO] - Number of regex retries in iteration 50: 2 -[2025-11-09 18:43:43,423][__main__][INFO] - agents played in iteration 50 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:43:44,220][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:43:44,248][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:43:44,287][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:43:44,311][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:43:44,312][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:43:44,312][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:43:44,963][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:43:45,418][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:43:45,919][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:43:46,413][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:43:46,908][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:43:47,403][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:43:47,897][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:43:48,395][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:43:48,891][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:43:49,388][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:43:49,881][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:43:50,375][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:43:50,871][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:43:51,365][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:43:51,859][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:43:52,358][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:43:52,855][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:43:53,355][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:43:53,863][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:43:54,360][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:43:54,858][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:43:55,366][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:43:55,864][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:43:56,361][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:43:56,863][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:43:57,356][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:43:57,852][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:43:58,346][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:43:58,842][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:43:59,336][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:43:59,830][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:44:00,327][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:44:00,822][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:44:01,316][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:44:01,822][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:44:02,316][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:44:02,810][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:44:03,300][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:44:03,794][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:44:04,287][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:44:04,780][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:44:05,276][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:44:05,772][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:44:06,268][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:44:06,768][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:44:07,268][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:44:07,764][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:44:08,264][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:44:08,762][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:44:09,259][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:44:09,754][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:44:10,251][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:44:10,760][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:44:11,259][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:44:11,756][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:44:12,253][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:44:12,750][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:44:13,247][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:44:13,749][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:44:14,248][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:44:14,763][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:44:15,263][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:44:15,760][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:44:16,256][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:44:16,755][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10179 tokens. -[2025-11-09 18:44:17,503][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 55.56%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 -[2025-11-09 18:44:18,211][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:44:18,213][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:44:18,214][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:44:20,250][__main__][INFO] - Iteration 51 took 52s (29.42% Gen, 66.67% Train). Generation: 15s, Training: 34s. Estimated remaining time: 42h 43m 40s. Estimated total time: 43h 29m 3s. Time estimates for 10 more iterations: 8m 41s, 100 more iterations: 1h 26m 58s, 500 more iterations: 7h 14m 50s. -[2025-11-09 18:44:20,253][__main__][INFO] - Starting iteration 51. -[2025-11-09 18:44:20,741][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 5 and human policies 1. -[2025-11-09 18:44:20,742][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:44:36,263][__main__][INFO] - Number of regex retries in iteration 51: 0 -[2025-11-09 18:44:36,263][__main__][INFO] - agents played in iteration 51 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:44:37,048][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:44:37,073][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:44:37,097][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:44:37,119][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:44:37,119][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:44:37,120][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:44:37,713][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:44:38,166][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:44:38,661][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:44:39,155][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:44:39,656][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:44:40,149][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:44:40,641][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:44:41,138][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:44:41,630][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:44:42,126][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:44:42,649][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:44:43,140][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:44:43,632][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:44:44,127][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:44:44,619][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:44:45,122][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:44:45,616][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:44:46,111][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:44:46,608][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:44:47,103][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:44:47,597][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:44:48,095][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:44:48,592][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:44:49,088][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:44:49,586][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:44:50,081][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:44:50,576][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:44:51,072][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:44:51,568][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:44:52,065][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:44:52,560][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:44:53,054][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:44:53,550][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:44:54,046][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:44:54,541][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:44:55,033][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:44:55,536][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:44:56,031][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:44:56,531][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:44:57,024][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:44:57,514][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:44:58,008][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:44:58,502][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:44:58,996][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:44:59,489][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:44:59,983][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:45:00,483][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:45:00,979][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:45:01,471][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:45:01,967][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:45:02,459][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:45:02,951][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:45:03,444][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:45:03,937][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:45:04,438][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:45:04,931][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:45:05,427][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:45:05,935][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:45:06,431][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:45:06,927][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:45:07,422][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:45:07,917][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:45:08,416][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:45:08,914][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:45:09,413][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10382 tokens. -[2025-11-09 18:45:10,126][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 55.62%, Block Peak % of device VRAM: 62.18%, ΔTime: 00:00:32 -[2025-11-09 18:45:10,872][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:45:10,874][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:45:10,876][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:45:11,814][__main__][INFO] - Iteration 52 took 51s (30.39% Gen, 67.77% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 47m 24s. Estimated total time: 42h 33m 39s. Time estimates for 10 more iterations: 8m 30s, 100 more iterations: 1h 25m 7s, 500 more iterations: 7h 5m 36s. -[2025-11-09 18:45:11,816][__main__][INFO] - Starting iteration 52. -[2025-11-09 18:45:12,283][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 5 and human policies 1. -[2025-11-09 18:45:12,283][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:45:20,272][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls +By proposing to take all 10 hats, I ensure I get the full value of my high-valued items. This leaves Bob with the books and balls, which he values more highly, ensuring a fairer distribution. did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 21:30:53,023][__main__][INFO] - Number of regex retries in iteration 27: 2 +[2026-03-25 21:30:53,024][__main__][INFO] - agents played in iteration 27 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:30:53,799][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:30:53,825][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:30:53,849][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:30:53,873][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:30:53,874][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:30:53,875][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:30:54,506][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:30:54,958][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:30:55,460][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:30:55,956][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:30:56,455][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:30:56,951][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:30:57,447][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:30:57,948][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:30:58,445][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:30:58,941][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:30:59,440][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:30:59,936][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:31:00,433][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:31:00,930][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:31:01,428][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:31:01,929][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:31:02,429][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:31:02,927][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:31:03,423][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:31:03,918][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:31:04,413][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:31:04,906][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:31:05,401][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:31:05,896][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:31:06,390][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:31:06,887][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:31:07,384][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:31:07,882][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:31:08,378][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:31:08,874][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:31:09,376][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:31:09,871][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:31:10,370][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:31:10,866][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:31:11,361][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:31:11,855][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:31:12,351][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:31:12,846][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:31:13,341][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:31:13,837][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:31:14,332][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:31:14,828][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:31:15,322][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:31:15,818][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:31:16,314][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:31:16,809][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:31:17,304][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:31:17,799][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:31:18,293][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:31:18,788][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:31:19,280][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:31:19,774][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:31:20,270][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:31:20,766][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:31:21,261][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:31:21,756][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:31:22,251][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:31:22,744][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:31:23,238][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:31:23,731][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:31:24,224][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:31:24,716][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:31:25,210][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:31:25,703][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:31:26,196][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9846 tokens. +[2026-03-25 21:31:26,862][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 13.94%, Current % of VRAM taken: 60.37%, Block Peak % of device VRAM: 61.84%, ΔTime: 00:00:32 +[2026-03-25 21:31:27,581][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:31:27,583][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:31:27,585][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:31:28,237][__main__][INFO] - Iteration 28 took 49s (29.40% Gen, 69.29% Train). Generation: 14s, Training: 34s. Estimated remaining time: 41h 8m 7s. Estimated total time: 41h 33m 55s. Time estimates for 10 more iterations: 8m 18s, 100 more iterations: 1h 23m 7s, 500 more iterations: 6h 55m 39s. +[2026-03-25 21:31:28,239][__main__][INFO] - Starting iteration 28. +[2026-03-25 21:31:28,640][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 2 and human policies 1. +[2026-03-25 21:31:28,641][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:31:29,653][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 21:31:42,248][__main__][INFO] - Number of regex retries in iteration 28: 1 +[2026-03-25 21:31:42,249][__main__][INFO] - agents played in iteration 28 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:31:43,024][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.43%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:31:43,048][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.43%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:31:43,071][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.43%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:31:43,094][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.43%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:31:43,095][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:31:43,096][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:31:43,717][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:31:44,170][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:31:44,669][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:31:45,164][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:31:45,659][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:31:46,155][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:31:46,651][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:31:47,146][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:31:47,645][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:31:48,145][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:31:48,645][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:31:49,144][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:31:49,639][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:31:50,133][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:31:50,631][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:31:51,127][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:31:51,624][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:31:52,120][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:31:52,616][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:31:53,113][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:31:53,607][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:31:54,101][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:31:54,596][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:31:55,091][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:31:55,586][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:31:56,080][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:31:56,575][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:31:57,070][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:31:57,567][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:31:58,061][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:31:58,556][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:31:59,051][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:31:59,549][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:32:00,045][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:32:00,545][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:32:01,042][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:32:01,538][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:32:02,036][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:32:02,533][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:32:03,029][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:32:03,525][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:32:04,021][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:32:04,516][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:32:05,010][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:32:05,506][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:32:06,002][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:32:06,497][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:32:06,992][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:32:07,487][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:32:07,981][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:32:08,475][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:32:08,969][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:32:09,464][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:32:09,958][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:32:10,453][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:32:10,948][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:32:11,445][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:32:11,941][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:32:12,435][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:32:12,931][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:32:13,426][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:32:13,922][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:32:14,416][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:32:14,911][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:32:15,405][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9878 tokens. +[2026-03-25 21:32:16,081][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 13.95%, Current % of VRAM taken: 60.37%, Block Peak % of device VRAM: 61.79%, ΔTime: 00:00:32 +[2026-03-25 21:32:16,797][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:32:16,799][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:32:16,800][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:32:17,445][__main__][INFO] - Iteration 29 took 48s (27.88% Gen, 70.79% Train). Generation: 13s, Training: 34s. Estimated remaining time: 40h 13m 38s. Estimated total time: 40h 40m 15s. Time estimates for 10 more iterations: 8m 8s, 100 more iterations: 1h 21m 20s, 500 more iterations: 6h 46m 42s. +[2026-03-25 21:32:17,447][__main__][INFO] - Starting iteration 29. +[2026-03-25 21:32:17,852][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 2 and human policies 1. +[2026-03-25 21:32:17,852][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:32:26,678][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 21:32:31,331][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 21:32:32,087][__main__][INFO] - Number of regex retries in iteration 29: 2 +[2026-03-25 21:32:32,088][__main__][INFO] - agents played in iteration 29 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:32:32,883][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.42%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:32:32,907][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.42%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:32:32,931][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.42%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:32:32,954][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.42%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:32:32,955][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:32:32,955][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:32:33,580][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:32:34,035][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:32:34,540][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:32:35,036][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:32:35,537][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:32:36,038][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:32:36,535][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:32:37,032][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:32:37,531][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:32:38,031][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:32:38,530][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:32:39,031][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:32:39,528][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:32:40,026][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:32:40,524][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:32:41,020][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:32:41,516][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:32:42,012][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:32:42,510][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:32:43,009][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:32:43,509][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:32:44,008][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:32:44,506][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:32:45,003][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:32:45,499][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:32:45,994][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:32:46,491][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:32:46,984][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:32:47,478][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:32:47,974][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:32:48,468][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:32:48,961][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:32:49,455][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:32:49,952][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:32:50,447][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:32:50,941][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:32:51,436][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:32:51,931][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:32:52,426][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:32:52,920][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:32:53,414][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:32:53,910][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:32:54,406][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:32:54,900][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:32:55,394][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:32:55,891][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:32:56,389][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:32:56,885][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:32:57,382][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:32:57,878][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:32:58,372][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:32:58,868][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:32:59,362][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:32:59,857][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:33:00,351][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:33:00,845][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:33:01,339][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:33:01,836][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:33:02,333][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:33:02,837][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:33:03,333][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:33:03,831][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:33:04,328][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:33:04,830][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:33:05,328][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9979 tokens. +[2026-03-25 21:33:06,010][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 60.48%, Block Peak % of device VRAM: 62.05%, ΔTime: 00:00:32 +[2026-03-25 21:33:06,734][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:33:06,736][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:33:06,738][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:33:07,384][__main__][INFO] - Iteration 30 took 49s (28.74% Gen, 69.95% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 49m 11s. Estimated total time: 41h 16m 38s. Time estimates for 10 more iterations: 8m 15s, 100 more iterations: 1h 22m 33s, 500 more iterations: 6h 52m 46s. +[2026-03-25 21:33:07,387][__main__][INFO] - Starting iteration 30. +[2026-03-25 21:33:07,789][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 2 and human policies 1. +[2026-03-25 21:33:07,789][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:33:18,583][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 21:33:21,580][__main__][INFO] - Number of regex retries in iteration 30: 1 +[2026-03-25 21:33:21,581][__main__][INFO] - agents played in iteration 30 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:33:22,357][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:33:22,380][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:33:22,404][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:33:22,427][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:33:22,428][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:33:22,428][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:33:23,080][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:33:23,537][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:33:24,040][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:33:24,536][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:33:25,037][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:33:25,533][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:33:26,029][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:33:26,523][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:33:27,018][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:33:27,515][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:33:28,013][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:33:28,529][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:33:29,031][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:33:29,528][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:33:30,025][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:33:30,522][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:33:31,018][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:33:31,515][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:33:32,013][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:33:32,512][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:33:33,013][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:33:33,510][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:33:34,009][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:33:34,507][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:33:35,006][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:33:35,502][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:33:35,998][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:33:36,496][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:33:36,993][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:33:37,491][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:33:37,988][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:33:38,485][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:33:38,986][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:33:39,483][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:33:39,984][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:33:40,481][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:33:40,976][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:33:41,473][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:33:41,970][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:33:42,467][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:33:42,963][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:33:43,462][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:33:43,957][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:33:44,451][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:33:44,947][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:33:45,443][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:33:45,937][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:33:46,434][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:33:46,931][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:33:47,428][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:33:47,925][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:33:48,422][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:33:48,917][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:33:49,413][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:33:49,909][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:33:50,408][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:33:50,906][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:33:51,409][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:33:51,905][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:33:52,401][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:33:52,902][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:33:53,401][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:33:53,899][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:33:54,399][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:33:54,893][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9970 tokens. +[2026-03-25 21:33:55,582][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 60.49%, Block Peak % of device VRAM: 62.00%, ΔTime: 00:00:32 +[2026-03-25 21:33:56,298][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:33:56,301][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:33:56,303][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:33:57,534][__main__][INFO] - Iteration 31 took 49s (27.72% Gen, 69.80% Train). Generation: 13s, Training: 34s. Estimated remaining time: 40h 58m 59s. Estimated total time: 41h 27m 16s. Time estimates for 10 more iterations: 8m 17s, 100 more iterations: 1h 22m 54s, 500 more iterations: 6h 54m 32s. +[2026-03-25 21:33:57,537][__main__][INFO] - Starting iteration 31. +[2026-03-25 21:33:57,938][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 3 and human policies 1. +[2026-03-25 21:33:57,938][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:34:08,136][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls -Given Bob's values for hats (1), books (10), and balls (10), and my own values for hats (10), books (10), and balls (1), it makes sense to propose taking all the hats since they have the highest value for me compared to Bob's value for them. Books and balls are not as valuable for me as they are for Bob, so proposing to take none of them maximizes my potential points for the round. did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:45:25,769][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:45:27,924][__main__][INFO] - Number of regex retries in iteration 52: 2 -[2025-11-09 18:45:27,925][__main__][INFO] - agents played in iteration 52 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:45:28,697][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:45:28,725][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:45:28,751][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:45:28,774][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:45:28,775][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:45:28,776][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:45:29,364][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:45:30,108][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:45:30,606][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:45:31,103][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:45:31,596][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:45:32,095][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:45:32,594][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:45:33,090][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:45:33,586][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:45:34,080][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:45:34,583][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:45:35,099][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:45:35,593][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:45:36,099][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:45:36,593][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:45:37,088][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:45:37,585][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:45:38,077][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:45:38,579][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:45:39,074][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:45:39,573][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:45:40,067][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:45:40,559][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:45:41,052][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:45:41,546][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:45:42,040][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:45:42,535][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:45:43,027][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:45:43,520][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:45:44,013][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:45:44,506][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:45:45,000][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:45:45,492][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:45:45,989][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:45:46,492][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:45:46,988][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:45:47,486][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:45:47,980][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:45:48,476][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:45:48,971][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:45:49,466][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:45:49,960][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:45:50,451][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:45:50,948][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:45:51,440][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:45:51,944][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:45:52,435][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:45:52,928][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:45:53,422][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:45:53,914][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:45:54,413][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:45:54,910][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:45:55,406][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:45:55,902][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:45:56,401][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:45:56,901][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:45:57,398][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:45:57,899][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:45:58,398][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:45:58,893][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:45:59,392][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:45:59,889][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:46:00,386][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:46:00,882][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:46:01,383][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10292 tokens. -[2025-11-09 18:46:02,080][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 55.64%, Block Peak % of device VRAM: 62.16%, ΔTime: 00:00:32 -[2025-11-09 18:46:02,782][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:46:02,784][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:46:02,787][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:46:03,660][__main__][INFO] - Iteration 53 took 51s (30.45% Gen, 67.85% Train). Generation: 15s, Training: 34s. Estimated remaining time: 42h 1m 47s. Estimated total time: 42h 48m 54s. Time estimates for 10 more iterations: 8m 33s, 100 more iterations: 1h 25m 37s, 500 more iterations: 7h 8m 9s. -[2025-11-09 18:46:03,663][__main__][INFO] - Starting iteration 53. -[2025-11-09 18:46:04,122][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 5 and human policies 1. -[2025-11-09 18:46:04,123][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:46:18,757][__main__][INFO] - Number of regex retries in iteration 53: 0 -[2025-11-09 18:46:18,758][__main__][INFO] - agents played in iteration 53 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:46:19,535][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:46:19,560][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:46:19,584][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:46:19,606][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:46:19,607][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:46:19,608][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:46:20,201][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:46:20,653][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:46:21,150][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:46:21,646][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:46:22,144][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:46:22,641][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:46:23,138][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:46:23,636][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:46:24,129][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:46:24,624][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:46:25,117][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:46:25,611][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:46:26,109][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:46:26,601][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:46:27,095][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:46:27,589][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:46:28,080][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:46:28,575][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:46:29,073][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:46:29,572][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:46:30,085][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:46:30,579][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:46:31,075][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:46:31,604][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:46:32,097][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:46:32,596][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:46:33,088][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:46:33,584][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:46:34,095][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:46:34,590][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:46:35,086][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:46:35,580][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:46:36,076][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:46:36,593][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:46:37,089][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:46:37,587][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:46:38,084][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:46:38,580][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:46:39,073][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:46:39,566][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:46:40,058][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:46:40,567][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:46:41,060][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:46:41,552][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:46:42,061][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:46:42,553][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:46:43,048][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:46:43,543][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:46:44,035][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:46:44,528][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:46:45,022][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:46:45,514][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:46:46,007][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:46:46,503][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:46:47,001][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:46:47,497][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:46:47,990][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:46:48,499][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:46:48,990][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:46:49,487][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:46:49,986][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:46:50,483][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:46:50,996][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:46:51,495][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:46:51,993][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10031 tokens. -[2025-11-09 18:46:52,710][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 55.58%, Block Peak % of device VRAM: 62.03%, ΔTime: 00:00:32 -[2025-11-09 18:46:53,429][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:46:53,430][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:46:53,432][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:46:54,310][__main__][INFO] - Iteration 54 took 50s (29.16% Gen, 69.09% Train). Generation: 14s, Training: 34s. Estimated remaining time: 41h 1m 27s. Estimated total time: 41h 49m 24s. Time estimates for 10 more iterations: 8m 21s, 100 more iterations: 1h 23m 38s, 500 more iterations: 6h 58m 14s. -[2025-11-09 18:46:54,312][__main__][INFO] - Starting iteration 54. -[2025-11-09 18:46:54,787][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 5 and human policies 1. -[2025-11-09 18:46:54,787][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:47:04,528][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls +This proposal maximizes my share of high-value items (hats) while Bob values books more highly. Since my value for hats (10) is much higher than Bob’s (1), and my value for books (10) is similar to his, but I propose to take all hats, I am certain to get the maximum value from this proposal. My value for balls is the same as Bob's, but since I don't value them as highly as hats and especially not as highly as books, and given the proportional allocation risk if I were to take any, it’s best to not take any at all. did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 21:34:13,055][__main__][INFO] - Number of regex retries in iteration 31: 1 +[2026-03-25 21:34:13,056][__main__][INFO] - agents played in iteration 31 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:34:13,879][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:34:13,902][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:34:13,926][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:34:13,949][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:34:13,949][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:34:13,950][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:34:14,635][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:34:15,089][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:34:15,591][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:34:16,093][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:34:16,592][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:34:17,088][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:34:17,588][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:34:18,086][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:34:18,584][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:34:19,082][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:34:19,577][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:34:20,073][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:34:20,571][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:34:21,071][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:34:21,567][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:34:22,064][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:34:22,561][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:34:23,058][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:34:23,556][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:34:24,054][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:34:24,552][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:34:25,048][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:34:25,543][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:34:26,039][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:34:26,534][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:34:27,032][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:34:27,533][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:34:28,031][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:34:28,529][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:34:29,026][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:34:29,523][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:34:30,019][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:34:30,516][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:34:31,011][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:34:31,507][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:34:32,005][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:34:32,501][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:34:32,994][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:34:33,488][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:34:33,983][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:34:34,477][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:34:34,971][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:34:35,466][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:34:35,962][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:34:36,461][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:34:36,959][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:34:37,453][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:34:37,946][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:34:38,442][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:34:38,937][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:34:39,432][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:34:39,932][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:34:40,430][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:34:40,926][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:34:41,422][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:34:41,918][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:34:42,413][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:34:42,909][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:34:43,402][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:34:43,894][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:34:44,387][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:34:44,881][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:34:45,375][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:34:45,868][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:34:46,361][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9912 tokens. +[2026-03-25 21:34:47,062][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 13.98%, Current % of VRAM taken: 60.41%, Block Peak % of device VRAM: 61.83%, ΔTime: 00:00:32 +[2026-03-25 21:34:47,779][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:34:47,782][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:34:47,783][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:34:48,432][__main__][INFO] - Iteration 32 took 50s (29.94% Gen, 68.77% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 35m 37s. Estimated total time: 42h 4m 45s. Time estimates for 10 more iterations: 8m 24s, 100 more iterations: 1h 24m 9s, 500 more iterations: 7h 0m 47s. +[2026-03-25 21:34:48,435][__main__][INFO] - Starting iteration 32. +[2026-03-25 21:34:48,836][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 3 and human policies 1. +[2026-03-25 21:34:48,837][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:34:51,371][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 21:34:55,642][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls -Given the values: -- You value hats at 10, books at 1, and balls at 1. -- Bob values hats at 1, books at 10, and balls at 10. +This proposal takes advantage of the significant value you place on hats while assuming that Alice values them much less. Given her lower valuation of hats, proposing to take all 10 hats ensures the maximum value for you in this round without much risk of the total sum exceeding the available items. Since books and balls have lower mutual values, splitting them would not significantly boost your points given the value Alice places on them. did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 21:35:03,629][__main__][INFO] - Number of regex retries in iteration 32: 2 +[2026-03-25 21:35:03,630][__main__][INFO] - agents played in iteration 32 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:35:04,411][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:35:04,435][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:35:04,458][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:35:04,482][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:35:04,483][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:35:04,483][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:35:05,102][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:35:05,553][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:35:06,056][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:35:06,552][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:35:07,048][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:35:07,544][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:35:08,040][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:35:08,537][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:35:09,034][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:35:09,532][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:35:10,028][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:35:10,526][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:35:11,025][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:35:11,520][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:35:12,016][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:35:12,514][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:35:13,010][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:35:13,506][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:35:14,004][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:35:14,500][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:35:14,996][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:35:15,492][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:35:15,989][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:35:16,484][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:35:16,981][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:35:17,478][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:35:17,975][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:35:18,476][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:35:18,972][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:35:19,471][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:35:19,968][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:35:20,468][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:35:20,965][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:35:21,459][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:35:21,954][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:35:22,448][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:35:22,943][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:35:23,437][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:35:23,931][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:35:24,426][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:35:24,920][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:35:25,414][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:35:25,909][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:35:26,405][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:35:26,900][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:35:27,394][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:35:27,888][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:35:28,382][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:35:28,877][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:35:29,371][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:35:29,865][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:35:30,360][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:35:30,855][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:35:31,350][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:35:31,848][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:35:32,346][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:35:32,840][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:35:33,334][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:35:33,828][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:35:34,323][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:35:34,818][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:35:35,313][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:35:35,808][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:35:36,302][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:35:36,796][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9930 tokens. +[2026-03-25 21:35:37,476][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.03%, Current % of VRAM taken: 60.46%, Block Peak % of device VRAM: 61.91%, ΔTime: 00:00:32 +[2026-03-25 21:35:38,200][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:35:38,202][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:35:38,205][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:35:38,862][__main__][INFO] - Iteration 33 took 50s (29.57% Gen, 69.11% Train). Generation: 14s, Training: 34s. Estimated remaining time: 41h 11m 21s. Estimated total time: 41h 41m 20s. Time estimates for 10 more iterations: 8m 20s, 100 more iterations: 1h 23m 22s, 500 more iterations: 6h 56m 53s. +[2026-03-25 21:35:38,865][__main__][INFO] - Starting iteration 33. +[2026-03-25 21:35:39,268][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 3 and human policies 1. +[2026-03-25 21:35:39,269][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:35:41,613][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 21:35:52,288][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls -Since there are 10 hats, and your value for them is significantly higher than Bob's, it's optimal to claim all the hats. Bob values books and balls more than you do, so they are better off keeping those items. By proposing to take all the hats, you ensure you get the maximum value from this round. did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:47:11,378][__main__][INFO] - Number of regex retries in iteration 54: 1 -[2025-11-09 18:47:11,379][__main__][INFO] - agents played in iteration 54 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:47:12,150][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:47:12,172][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:47:12,194][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:47:12,216][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:47:12,216][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:47:12,218][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:47:12,811][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:47:13,282][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:47:13,784][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:47:14,278][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:47:14,778][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:47:15,276][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:47:15,783][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:47:16,279][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:47:16,771][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:47:17,270][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:47:17,766][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:47:18,267][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:47:18,764][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:47:19,258][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:47:19,756][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:47:20,252][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:47:20,748][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:47:21,244][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:47:21,737][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:47:22,232][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:47:22,728][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:47:23,223][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:47:23,717][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:47:24,212][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:47:24,707][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:47:25,206][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:47:25,703][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:47:26,200][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:47:26,693][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:47:27,188][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:47:27,689][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:47:28,185][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:47:28,681][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:47:29,176][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:47:29,670][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:47:30,162][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:47:30,655][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:47:31,149][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:47:31,646][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:47:32,139][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:47:32,631][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:47:33,126][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:47:33,620][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:47:34,113][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:47:34,606][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:47:35,099][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:47:35,593][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:47:36,087][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:47:36,581][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:47:37,078][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:47:37,573][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:47:38,067][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:47:38,565][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:47:39,063][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:47:39,580][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:47:40,077][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:47:40,574][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:47:41,073][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:47:41,568][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:47:42,074][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:47:42,575][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:47:43,073][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:47:43,583][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:47:44,081][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:47:44,578][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10305 tokens. -[2025-11-09 18:47:45,289][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.04%, Current % of VRAM taken: 55.54%, Block Peak % of device VRAM: 62.12%, ΔTime: 00:00:32 -[2025-11-09 18:47:45,995][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:47:45,997][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:47:45,999][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:47:46,924][__main__][INFO] - Iteration 55 took 52s (31.82% Gen, 66.40% Train). Generation: 16s, Training: 34s. Estimated remaining time: 42h 38m 4s. Estimated total time: 43h 26m 54s. Time estimates for 10 more iterations: 8m 41s, 100 more iterations: 1h 26m 53s, 500 more iterations: 7h 14m 29s. -[2025-11-09 18:47:46,926][__main__][INFO] - Starting iteration 55. -[2025-11-09 18:47:47,387][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 5 and human policies 1. -[2025-11-09 18:47:47,387][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:47:51,759][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:48:02,726][__main__][INFO] - Number of regex retries in iteration 55: 1 -[2025-11-09 18:48:02,727][__main__][INFO] - agents played in iteration 55 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:48:03,498][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:48:03,525][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:48:03,552][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:48:03,574][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:48:03,575][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:48:03,576][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:48:04,173][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:48:04,625][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:48:05,125][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:48:05,622][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:48:06,116][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:48:06,616][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:48:07,112][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:48:07,606][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:48:08,104][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:48:08,595][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:48:09,089][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:48:09,589][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:48:10,083][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:48:10,577][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:48:11,074][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:48:11,571][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:48:12,067][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:48:12,561][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:48:13,060][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:48:13,559][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:48:14,052][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:48:14,549][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:48:15,050][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:48:15,545][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:48:16,043][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:48:16,535][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:48:17,030][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:48:17,525][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:48:18,021][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:48:18,534][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:48:19,034][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:48:19,531][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:48:20,027][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:48:20,521][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:48:21,018][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:48:21,516][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:48:22,008][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:48:22,503][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:48:22,995][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:48:23,489][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:48:23,992][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:48:24,488][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:48:24,983][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:48:25,479][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:48:25,974][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:48:26,470][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:48:26,964][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:48:27,458][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:48:27,951][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:48:28,448][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:48:28,945][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:48:29,446][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:48:29,943][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:48:30,437][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:48:30,935][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:48:31,434][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:48:31,934][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:48:32,434][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:48:32,932][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:48:33,431][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:48:33,931][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:48:34,432][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:48:34,930][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:48:35,433][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:48:35,930][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10224 tokens. -[2025-11-09 18:48:36,633][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.09%, Current % of VRAM taken: 55.58%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:32 -[2025-11-09 18:48:37,362][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:48:37,364][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:48:37,366][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:48:38,412][__main__][INFO] - Iteration 56 took 51s (30.06% Gen, 67.89% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 41m 34s. Estimated total time: 42h 31m 15s. Time estimates for 10 more iterations: 8m 30s, 100 more iterations: 1h 25m 2s, 500 more iterations: 7h 5m 12s. -[2025-11-09 18:48:38,414][__main__][INFO] - Starting iteration 56. -[2025-11-09 18:48:38,895][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 5 and human policies 1. -[2025-11-09 18:48:38,896][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:48:40,770][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:48:53,997][__main__][INFO] - Number of regex retries in iteration 56: 1 -[2025-11-09 18:48:53,998][__main__][INFO] - agents played in iteration 56 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:48:54,769][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:48:54,794][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:48:54,818][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:48:54,839][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:48:54,840][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:48:54,841][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:48:55,428][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:48:55,882][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:48:56,383][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:48:56,879][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:48:57,373][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:48:57,869][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:48:58,377][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:48:58,873][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:48:59,370][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:48:59,871][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:49:00,371][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:49:00,879][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:49:01,376][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:49:01,870][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:49:02,386][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:49:02,879][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:49:03,376][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:49:03,869][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:49:04,364][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:49:04,866][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:49:05,361][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:49:05,855][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:49:06,351][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:49:06,847][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:49:07,341][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:49:07,838][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:49:08,333][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:49:08,831][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:49:09,325][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:49:09,821][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:49:10,318][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:49:10,811][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:49:11,304][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:49:11,797][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:49:12,294][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:49:12,799][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:49:13,297][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:49:13,791][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:49:14,287][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:49:14,782][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:49:15,279][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:49:15,779][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:49:16,273][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:49:16,769][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:49:17,267][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:49:17,763][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:49:18,260][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:49:18,757][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:49:19,252][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:49:19,755][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:49:20,255][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:49:20,751][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:49:21,251][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:49:21,748][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:49:22,250][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:49:22,750][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:49:23,247][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:49:23,744][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:49:24,245][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:49:24,743][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:49:25,245][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:49:25,747][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:49:26,255][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:49:26,758][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:49:27,265][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10362 tokens. -[2025-11-09 18:49:28,004][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 55.69%, Block Peak % of device VRAM: 62.10%, ΔTime: 00:00:32 -[2025-11-09 18:49:28,715][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:49:28,718][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:49:28,719][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:49:29,932][__main__][INFO] - Iteration 57 took 51s (29.59% Gen, 68.03% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 41m 19s. Estimated total time: 42h 31m 52s. Time estimates for 10 more iterations: 8m 30s, 100 more iterations: 1h 25m 3s, 500 more iterations: 7h 5m 18s. -[2025-11-09 18:49:29,934][__main__][INFO] - Starting iteration 57. -[2025-11-09 18:49:30,414][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 5 and human policies 1. -[2025-11-09 18:49:30,415][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:49:42,762][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:49:47,992][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:49:48,986][__main__][INFO] - Number of regex retries in iteration 57: 2 -[2025-11-09 18:49:48,987][__main__][INFO] - agents played in iteration 57 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:49:49,774][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:49:49,796][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:49:49,818][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:49:49,840][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:49:49,840][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:49:49,841][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:49:50,430][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:49:50,901][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:49:51,398][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:49:51,897][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:49:52,393][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:49:52,890][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:49:53,392][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:49:53,889][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:49:54,387][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:49:54,883][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:49:55,380][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:49:55,873][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:49:56,369][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:49:56,867][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:49:57,361][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:49:57,856][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:49:58,353][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:49:58,850][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:49:59,345][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:49:59,841][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:50:00,335][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:50:00,828][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:50:01,329][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:50:01,829][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:50:02,323][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:50:02,822][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:50:03,317][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:50:03,812][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:50:04,309][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:50:04,808][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:50:05,314][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:50:05,810][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:50:06,305][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:50:06,802][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:50:07,295][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:50:07,800][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:50:08,294][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:50:08,791][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:50:09,297][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:50:09,793][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:50:10,289][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:50:10,782][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:50:11,278][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:50:11,773][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:50:12,272][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:50:12,769][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:50:13,265][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:50:13,760][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:50:14,254][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:50:14,747][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:50:15,246][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:50:15,747][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:50:16,242][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:50:16,738][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:50:17,237][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:50:17,735][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:50:18,232][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:50:18,733][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:50:19,230][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:50:19,736][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:50:20,236][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:50:20,735][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:50:21,238][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:50:21,738][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:50:22,242][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10474 tokens. -[2025-11-09 18:50:22,915][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 55.63%, Block Peak % of device VRAM: 62.23%, ΔTime: 00:00:32 -[2025-11-09 18:50:23,657][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:50:23,658][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:50:23,665][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:50:24,634][__main__][INFO] - Iteration 58 took 54s (34.25% Gen, 63.96% Train). Generation: 18s, Training: 34s. Estimated remaining time: 44h 19m 32s. Estimated total time: 45h 11m 0s. Time estimates for 10 more iterations: 9m 2s, 100 more iterations: 1h 30m 22s, 500 more iterations: 7h 31m 50s. -[2025-11-09 18:50:24,637][__main__][INFO] - Starting iteration 58. -[2025-11-09 18:50:25,108][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 5 and human policies 1. -[2025-11-09 18:50:25,108][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:50:26,964][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:50:40,179][__main__][INFO] - Number of regex retries in iteration 58: 1 -[2025-11-09 18:50:40,180][__main__][INFO] - agents played in iteration 58 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:50:40,954][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:50:40,981][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:50:41,006][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:50:41,029][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:50:41,030][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:50:41,031][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:50:41,627][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:50:42,077][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:50:42,575][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:50:43,074][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:50:43,568][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:50:44,061][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:50:44,555][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:50:45,051][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:50:45,545][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:50:46,048][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:50:46,545][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:50:47,040][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:50:47,535][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:50:48,029][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:50:48,529][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:50:49,024][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:50:49,517][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:50:50,011][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:50:50,505][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:50:51,014][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:50:51,508][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:50:52,999][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:50:52,492][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:50:52,986][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:50:53,481][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:50:53,976][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:50:54,468][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:50:54,962][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:50:55,457][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:50:55,950][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:50:56,446][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:50:56,938][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:50:57,433][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:50:57,934][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:50:58,428][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:50:58,929][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:50:59,423][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:50:59,919][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:51:00,413][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:51:00,905][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:51:01,397][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:51:01,896][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:51:02,389][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:51:02,899][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:51:03,393][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:51:03,886][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:51:04,394][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:51:04,887][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:51:05,379][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:51:05,872][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:51:06,364][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:51:06,861][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:51:07,359][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:51:07,855][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:51:08,353][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:51:08,853][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:51:09,352][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:51:09,850][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:51:10,352][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:51:10,850][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:51:11,348][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:51:11,846][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:51:12,349][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:51:12,847][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:51:13,350][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10308 tokens. -[2025-11-09 18:51:14,031][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 55.65%, Block Peak % of device VRAM: 62.05%, ΔTime: 00:00:32 -[2025-11-09 18:51:14,767][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:51:14,769][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:51:14,771][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:51:15,927][__main__][INFO] - Iteration 59 took 50s (29.66% Gen, 68.07% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 28m 39s. Estimated total time: 42h 20m 58s. Time estimates for 10 more iterations: 8m 28s, 100 more iterations: 1h 24m 41s, 500 more iterations: 7h 3m 29s. -[2025-11-09 18:51:15,929][__main__][INFO] - Starting iteration 59. -[2025-11-09 18:51:16,394][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 5 and human policies 1. -[2025-11-09 18:51:16,394][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:51:31,598][__main__][INFO] - Number of regex retries in iteration 59: 0 -[2025-11-09 18:51:31,599][__main__][INFO] - agents played in iteration 59 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:51:32,370][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:51:32,397][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:51:32,423][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:51:32,446][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:51:32,446][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:51:32,447][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:51:33,050][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:51:33,503][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:51:34,004][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:51:34,503][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:51:35,002][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:51:35,500][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:51:36,001][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:51:36,514][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:51:37,026][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:51:37,523][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:51:38,021][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:51:38,524][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:51:39,017][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:51:39,512][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:51:40,007][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:51:40,502][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:51:41,000][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:51:41,496][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:51:41,990][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:51:42,498][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:51:42,993][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:51:43,491][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:51:43,987][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:51:44,481][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:51:44,981][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:51:45,482][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:51:45,978][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:51:46,476][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:51:46,972][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:51:47,468][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:51:47,964][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:51:48,462][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:51:48,961][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:51:49,455][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:51:49,952][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:51:50,449][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:51:50,946][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:51:51,445][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:51:51,942][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:51:52,439][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:51:52,936][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:51:53,430][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:51:53,926][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:51:54,420][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:51:54,914][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:51:55,423][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:51:55,915][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:51:56,411][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:51:56,911][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:51:57,408][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:51:57,915][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:51:58,416][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:51:58,917][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:51:59,418][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:51:59,915][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:52:00,417][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:52:00,919][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:52:01,419][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:52:01,928][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:52:02,426][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:52:02,925][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:52:03,428][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:52:03,928][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:52:04,431][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:52:04,931][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10444 tokens. -[2025-11-09 18:52:05,606][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 55.70%, Block Peak % of device VRAM: 62.22%, ΔTime: 00:00:32 -[2025-11-09 18:52:06,347][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:52:06,349][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:52:06,351][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:52:07,243][__main__][INFO] - Iteration 60 took 50s (29.90% Gen, 68.34% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 29m 18s. Estimated total time: 42h 22m 28s. Time estimates for 10 more iterations: 8m 28s, 100 more iterations: 1h 24m 44s, 500 more iterations: 7h 3m 44s. -[2025-11-09 18:52:07,245][__main__][INFO] - Starting iteration 60. -[2025-11-09 18:52:07,711][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 5 and human policies 1. -[2025-11-09 18:52:07,711][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:52:22,294][__main__][INFO] - Number of regex retries in iteration 60: 0 -[2025-11-09 18:52:22,295][__main__][INFO] - agents played in iteration 60 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:52:23,078][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:52:23,100][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:52:23,123][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:52:23,144][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:52:23,145][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:52:23,146][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:52:23,738][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:52:24,189][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:52:24,692][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:52:25,186][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:52:25,685][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:52:26,178][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:52:26,673][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:52:27,172][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:52:27,668][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:52:28,160][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:52:28,652][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:52:29,144][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:52:29,643][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:52:30,137][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:52:30,628][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:52:31,125][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:52:31,618][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:52:32,111][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:52:32,612][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:52:33,105][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:52:33,608][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:52:34,102][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:52:34,595][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:52:35,092][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:52:35,587][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:52:36,084][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:52:36,580][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:52:37,072][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:52:37,575][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:52:38,070][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:52:38,564][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:52:39,056][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:52:39,550][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:52:40,050][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:52:40,544][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:52:41,038][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:52:41,564][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:52:42,057][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:52:42,551][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:52:43,049][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:52:43,542][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:52:44,035][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:52:44,528][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:52:45,022][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:52:45,516][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:52:46,007][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:52:46,500][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:52:46,993][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:52:47,489][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:52:47,983][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:52:48,477][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:52:48,970][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:52:49,472][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:52:49,971][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:52:50,468][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:52:50,965][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:52:51,462][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:52:51,967][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:52:52,465][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:52:52,963][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:52:53,460][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:52:53,957][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:52:54,456][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:52:54,956][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:52:55,455][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10092 tokens. -[2025-11-09 18:52:56,195][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.09%, Current % of VRAM taken: 55.58%, Block Peak % of device VRAM: 62.04%, ΔTime: 00:00:32 -[2025-11-09 18:52:56,926][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:52:56,927][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:52:56,929][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:52:58,882][__main__][INFO] - Iteration 61 took 51s (28.50% Gen, 67.68% Train). Generation: 14s, Training: 34s. Estimated remaining time: 41h 44m 34s. Estimated total time: 42h 38m 36s. Time estimates for 10 more iterations: 8m 31s, 100 more iterations: 1h 25m 17s, 500 more iterations: 7h 6m 26s. -[2025-11-09 18:52:58,884][__main__][INFO] - Starting iteration 61. -[2025-11-09 18:52:59,348][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 6 and human policies 1. -[2025-11-09 18:52:59,348][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:53:05,046][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls +Given the values, I see an opportunity to maximize my points by focusing on the items where my value is higher than Alice's. For hats, her value is 10, which is higher than my 1. For books and balls, my values are 10, equal to Alice's, so there's no clear advantage to taking more of these. Given the rules, if I take all 10 hats, and Alice takes nothing, the distribution will be exactly as my proposal, as the total sum of our proposals will not exceed the item quantity. This will maximize my points for this round since hats are worth the least to me, but I have a significant value for books and balls which I'll focus on in future rounds if needed. did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 21:35:54,782][__main__][INFO] - Number of regex retries in iteration 33: 2 +[2026-03-25 21:35:54,783][__main__][INFO] - agents played in iteration 33 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:35:55,602][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:35:55,626][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:35:55,651][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:35:55,674][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:35:55,675][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:35:55,675][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:35:56,296][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:35:56,751][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:35:57,256][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:35:57,758][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:35:58,256][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:35:58,754][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:35:59,252][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:35:59,760][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:36:00,254][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:36:00,753][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:36:01,249][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:36:01,751][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:36:02,250][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:36:02,748][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:36:03,246][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:36:03,742][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:36:04,238][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:36:04,737][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:36:05,234][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:36:05,731][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:36:06,231][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:36:06,730][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:36:07,227][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:36:07,727][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:36:08,222][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:36:08,718][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:36:09,215][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:36:09,712][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:36:10,209][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:36:10,708][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:36:11,205][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:36:11,703][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:36:12,200][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:36:12,695][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:36:13,191][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:36:13,686][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:36:14,182][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:36:14,677][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:36:15,171][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:36:15,668][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:36:16,166][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:36:16,664][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:36:17,160][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:36:17,658][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:36:18,155][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:36:18,650][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:36:19,145][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:36:19,642][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:36:20,138][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:36:20,636][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:36:21,133][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:36:21,630][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:36:22,127][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:36:22,622][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:36:23,118][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:36:23,614][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:36:24,109][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:36:24,603][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:36:25,098][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:36:25,593][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:36:26,088][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:36:26,581][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:36:27,075][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:36:27,571][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:36:28,065][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9960 tokens. +[2026-03-25 21:36:28,746][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 13.96%, Current % of VRAM taken: 60.39%, Block Peak % of device VRAM: 62.02%, ΔTime: 00:00:32 +[2026-03-25 21:36:29,468][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:36:29,470][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:36:29,472][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:36:30,121][__main__][INFO] - Iteration 34 took 50s (30.51% Gen, 68.21% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 51m 49s. Estimated total time: 42h 22m 39s. Time estimates for 10 more iterations: 8m 28s, 100 more iterations: 1h 24m 45s, 500 more iterations: 7h 3m 46s. +[2026-03-25 21:36:30,123][__main__][INFO] - Starting iteration 34. +[2026-03-25 21:36:30,523][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 3 and human policies 1. +[2026-03-25 21:36:30,524][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:36:44,893][__main__][INFO] - Number of regex retries in iteration 34: 0 +[2026-03-25 21:36:44,894][__main__][INFO] - agents played in iteration 34 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:36:45,698][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.44%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:36:45,721][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.44%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:36:45,744][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.44%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:36:45,768][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.44%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:36:45,768][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:36:45,769][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:36:46,401][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:36:46,854][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:36:47,354][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:36:47,851][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:36:48,349][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:36:48,847][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:36:49,343][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:36:49,844][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:36:50,341][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:36:50,837][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:36:51,332][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:36:51,828][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:36:52,323][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:36:52,818][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:36:53,313][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:36:53,809][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:36:54,306][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:36:54,802][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:36:55,296][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:36:55,791][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:36:56,287][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:36:56,784][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:36:57,277][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:36:57,772][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:36:58,267][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:36:58,763][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:36:59,256][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:36:59,753][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:37:00,249][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:37:00,746][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:37:01,240][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:37:01,736][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:37:02,232][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:37:02,729][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:37:03,224][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:37:03,721][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:37:04,215][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:37:04,711][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:37:05,205][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:37:05,700][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:37:06,193][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:37:06,690][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:37:07,210][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:37:07,708][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:37:08,211][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:37:08,710][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:37:09,210][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:37:09,707][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:37:10,206][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:37:10,702][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:37:11,195][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:37:11,688][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:37:12,181][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:37:12,674][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:37:13,166][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:37:13,661][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:37:14,154][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:37:14,650][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:37:15,147][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:37:15,645][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:37:16,139][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:37:16,635][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:37:17,133][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:37:17,628][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:37:18,125][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9999 tokens. +[2026-03-25 21:37:18,807][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.05%, Current % of VRAM taken: 60.48%, Block Peak % of device VRAM: 62.04%, ΔTime: 00:00:32 +[2026-03-25 21:37:19,552][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:37:19,554][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:37:19,555][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:37:20,204][__main__][INFO] - Iteration 35 took 49s (28.92% Gen, 69.77% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 52m 23s. Estimated total time: 41h 24m 2s. Time estimates for 10 more iterations: 8m 16s, 100 more iterations: 1h 22m 48s, 500 more iterations: 6h 54m 0s. +[2026-03-25 21:37:20,206][__main__][INFO] - Starting iteration 35. +[2026-03-25 21:37:20,606][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 3 and human policies 1. +[2026-03-25 21:37:20,607][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:37:30,368][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 21:37:35,000][__main__][INFO] - Number of regex retries in iteration 35: 1 +[2026-03-25 21:37:35,001][__main__][INFO] - agents played in iteration 35 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:37:35,791][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:37:35,815][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:37:35,838][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:37:35,861][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:37:35,862][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:37:35,862][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:37:36,516][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:37:36,970][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:37:37,470][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:37:37,972][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:37:38,471][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:37:38,970][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:37:39,469][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:37:39,967][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:37:40,463][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:37:40,963][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:37:41,466][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:37:41,960][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:37:42,457][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:37:42,955][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:37:43,453][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:37:43,950][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:37:44,446][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:37:44,943][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:37:45,440][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:37:45,935][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:37:46,432][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:37:46,929][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:37:47,426][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:37:47,923][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:37:48,423][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:37:48,919][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:37:49,440][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:37:49,940][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:37:50,438][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:37:50,935][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:37:51,433][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:37:51,929][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:37:52,426][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:37:52,923][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:37:53,418][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:37:53,914][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:37:54,411][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:37:54,908][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:37:55,406][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:37:55,900][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:37:56,395][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:37:56,890][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:37:57,387][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:37:57,881][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:37:58,377][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:37:58,875][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:37:59,374][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:37:59,871][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:38:00,365][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:38:00,862][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:38:01,358][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:38:01,857][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:38:02,355][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:38:02,853][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:38:03,349][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:38:03,844][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:38:04,339][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:38:04,836][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:38:05,333][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:38:05,832][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:38:06,330][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:38:06,831][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:38:07,333][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:38:07,830][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:38:08,330][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10062 tokens. +[2026-03-25 21:38:09,020][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.58%, Block Peak % of device VRAM: 62.02%, ΔTime: 00:00:32 +[2026-03-25 21:38:09,749][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:38:09,752][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:38:09,754][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:38:10,409][__main__][INFO] - Iteration 36 took 49s (28.90% Gen, 69.78% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 57m 41s. Estimated total time: 41h 30m 11s. Time estimates for 10 more iterations: 8m 18s, 100 more iterations: 1h 23m 0s, 500 more iterations: 6h 55m 1s. +[2026-03-25 21:38:10,412][__main__][INFO] - Starting iteration 36. +[2026-03-25 21:38:10,812][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 3 and human policies 1. +[2026-03-25 21:38:10,813][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:38:24,504][__main__][INFO] - Number of regex retries in iteration 36: 0 +[2026-03-25 21:38:24,504][__main__][INFO] - agents played in iteration 36 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:38:25,286][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:38:25,310][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:38:25,333][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:38:25,356][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:38:25,357][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:38:25,357][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:38:26,004][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:38:26,460][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:38:26,961][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:38:27,458][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:38:27,954][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:38:28,455][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:38:28,952][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:38:29,452][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:38:29,948][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:38:30,443][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:38:30,938][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:38:31,434][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:38:31,932][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:38:32,429][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:38:32,924][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:38:33,420][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:38:33,916][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:38:34,414][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:38:34,910][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:38:35,411][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:38:35,907][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:38:36,405][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:38:36,900][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:38:37,396][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:38:37,896][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:38:38,392][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:38:38,889][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:38:39,386][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:38:39,882][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:38:40,377][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:38:40,873][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:38:41,370][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:38:41,867][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:38:42,362][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:38:42,858][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:38:43,354][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:38:43,852][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:38:44,349][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:38:44,848][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:38:45,344][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:38:45,840][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:38:46,335][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:38:46,830][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:38:47,326][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:38:47,821][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:38:48,317][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:38:48,813][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:38:49,310][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:38:49,806][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:38:50,304][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:38:50,801][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:38:51,296][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:38:51,792][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:38:52,288][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:38:52,786][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:38:53,281][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:38:53,777][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:38:54,272][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:38:54,768][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:38:55,264][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:38:55,759][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:38:56,254][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:38:56,749][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:38:57,247][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:38:57,743][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10044 tokens. +[2026-03-25 21:38:58,436][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.04%, Current % of VRAM taken: 60.47%, Block Peak % of device VRAM: 61.94%, ΔTime: 00:00:32 +[2026-03-25 21:38:59,156][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:38:59,158][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:38:59,160][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:38:59,810][__main__][INFO] - Iteration 37 took 48s (27.94% Gen, 70.73% Train). Generation: 13s, Training: 34s. Estimated remaining time: 40h 16m 34s. Estimated total time: 40h 49m 53s. Time estimates for 10 more iterations: 8m 9s, 100 more iterations: 1h 21m 39s, 500 more iterations: 6h 48m 18s. +[2026-03-25 21:38:59,812][__main__][INFO] - Starting iteration 37. +[2026-03-25 21:39:00,214][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 3 and human policies 1. +[2026-03-25 21:39:00,214][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:39:14,182][__main__][INFO] - Number of regex retries in iteration 37: 0 +[2026-03-25 21:39:14,184][__main__][INFO] - agents played in iteration 37 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:39:14,991][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:39:15,014][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:39:15,038][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:39:15,060][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:39:15,061][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:39:15,062][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:39:15,705][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:39:16,159][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:39:16,658][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:39:17,152][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:39:17,649][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:39:18,144][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:39:18,637][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:39:19,132][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:39:19,629][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:39:20,126][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:39:20,625][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:39:21,124][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:39:21,620][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:39:22,116][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:39:22,613][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:39:23,113][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:39:23,611][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:39:24,108][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:39:24,609][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:39:25,106][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:39:25,603][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:39:26,101][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:39:26,599][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:39:27,096][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:39:27,594][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:39:28,114][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:39:28,616][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:39:29,115][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:39:29,612][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:39:30,111][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:39:30,607][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:39:31,107][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:39:31,606][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:39:32,109][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:39:32,607][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:39:33,107][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:39:33,604][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:39:34,104][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:39:34,601][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:39:35,100][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:39:35,597][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:39:36,094][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:39:36,589][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:39:37,086][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:39:37,582][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:39:38,078][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:39:38,571][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:39:39,067][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:39:39,562][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:39:40,056][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:39:40,551][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:39:41,045][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:39:41,542][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:39:42,039][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:39:42,535][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:39:43,030][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:39:43,527][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:39:44,023][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:39:44,520][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:39:45,016][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:39:45,512][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:39:46,011][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:39:46,507][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:39:47,004][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:39:47,500][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10001 tokens. +[2026-03-25 21:39:48,196][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 60.49%, Block Peak % of device VRAM: 61.99%, ΔTime: 00:00:32 +[2026-03-25 21:39:48,919][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:39:48,921][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:39:48,923][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:39:49,589][__main__][INFO] - Iteration 38 took 49s (28.29% Gen, 70.36% Train). Generation: 13s, Training: 34s. Estimated remaining time: 40h 34m 37s. Estimated total time: 41h 8m 46s. Time estimates for 10 more iterations: 8m 13s, 100 more iterations: 1h 22m 17s, 500 more iterations: 6h 51m 27s. +[2026-03-25 21:39:49,591][__main__][INFO] - Starting iteration 38. +[2026-03-25 21:39:49,993][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 3 and human policies 1. +[2026-03-25 21:39:49,994][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:40:04,261][__main__][INFO] - Number of regex retries in iteration 38: 0 +[2026-03-25 21:40:04,261][__main__][INFO] - agents played in iteration 38 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:40:05,041][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:40:05,064][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:40:05,088][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:40:05,111][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:40:05,111][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:40:05,112][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:40:05,707][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:40:06,159][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:40:06,659][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:40:07,156][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:40:07,652][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:40:08,148][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:40:08,644][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:40:09,139][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:40:09,635][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:40:10,132][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:40:10,630][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:40:11,128][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:40:11,625][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:40:12,125][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:40:12,625][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:40:13,123][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:40:13,619][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:40:14,115][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:40:14,612][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:40:15,106][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:40:15,601][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:40:16,095][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:40:16,591][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:40:17,086][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:40:17,580][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:40:18,075][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:40:18,571][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:40:19,069][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:40:19,564][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:40:20,062][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:40:20,558][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:40:21,053][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:40:21,548][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:40:22,045][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:40:22,541][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:40:23,036][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:40:23,531][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:40:24,028][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:40:24,523][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:40:25,018][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:40:25,514][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:40:26,009][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:40:26,506][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:40:27,001][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:40:27,496][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:40:27,990][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:40:28,484][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:40:28,980][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:40:29,476][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:40:29,971][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:40:30,465][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:40:30,961][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:40:31,457][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:40:31,955][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:40:32,450][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:40:32,947][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:40:33,442][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:40:33,938][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:40:34,434][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:40:34,931][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:40:35,427][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:40:35,924][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:40:36,422][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:40:36,918][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:40:37,415][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9981 tokens. +[2026-03-25 21:40:38,100][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 60.49%, Block Peak % of device VRAM: 61.85%, ΔTime: 00:00:32 +[2026-03-25 21:40:38,857][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:40:38,860][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:40:38,861][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:40:39,513][__main__][INFO] - Iteration 39 took 49s (28.81% Gen, 69.87% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 41m 2s. Estimated total time: 41h 16m 2s. Time estimates for 10 more iterations: 8m 15s, 100 more iterations: 1h 22m 32s, 500 more iterations: 6h 52m 40s. +[2026-03-25 21:40:39,516][__main__][INFO] - Starting iteration 39. +[2026-03-25 21:40:39,919][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 3 and human policies 1. +[2026-03-25 21:40:39,920][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:40:44,094][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 21:40:53,792][__main__][INFO] - Number of regex retries in iteration 39: 1 +[2026-03-25 21:40:53,793][__main__][INFO] - agents played in iteration 39 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:40:54,578][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:40:54,601][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:40:54,624][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:40:54,647][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:40:54,648][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:40:54,648][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:40:55,265][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:40:55,719][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:40:56,221][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:40:56,718][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:40:57,218][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:40:57,712][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:40:58,209][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:40:58,707][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:40:59,203][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:40:59,703][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:41:00,200][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:41:00,696][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:41:01,193][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:41:01,693][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:41:02,195][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:41:02,692][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:41:03,188][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:41:03,687][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:41:04,183][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:41:04,683][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:41:05,181][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:41:05,676][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:41:06,173][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:41:06,669][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:41:07,168][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:41:07,665][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:41:08,164][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:41:08,684][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:41:09,185][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:41:09,685][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:41:10,182][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:41:10,686][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:41:11,186][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:41:11,681][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:41:12,173][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:41:12,669][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:41:13,165][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:41:13,660][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:41:14,155][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:41:14,649][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:41:15,143][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:41:15,637][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:41:16,133][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:41:16,630][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:41:17,126][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:41:17,621][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:41:18,118][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:41:18,615][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:41:19,113][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:41:19,607][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:41:20,101][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:41:20,595][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:41:21,089][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:41:21,584][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:41:22,079][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:41:22,573][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:41:23,067][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:41:23,561][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:41:24,059][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:41:24,559][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:41:25,055][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:41:25,553][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:41:26,049][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:41:26,547][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:41:27,045][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9953 tokens. +[2026-03-25 21:41:27,728][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.09%, Current % of VRAM taken: 60.52%, Block Peak % of device VRAM: 62.04%, ΔTime: 00:00:32 +[2026-03-25 21:41:28,454][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:41:28,458][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:41:28,459][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:41:29,112][__main__][INFO] - Iteration 40 took 49s (28.20% Gen, 70.47% Train). Generation: 13s, Training: 34s. Estimated remaining time: 40h 23m 51s. Estimated total time: 40h 59m 40s. Time estimates for 10 more iterations: 8m 11s, 100 more iterations: 1h 21m 59s, 500 more iterations: 6h 49m 56s. +[2026-03-25 21:41:29,115][__main__][INFO] - Starting iteration 40. +[2026-03-25 21:41:29,516][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 3 and human policies 1. +[2026-03-25 21:41:29,517][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:41:43,385][__main__][INFO] - Number of regex retries in iteration 40: 0 +[2026-03-25 21:41:43,386][__main__][INFO] - agents played in iteration 40 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:41:44,168][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:41:44,192][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:41:44,216][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:41:44,240][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:41:44,241][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:41:44,241][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:41:44,900][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:41:45,354][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:41:45,860][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:41:46,360][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:41:46,859][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:41:47,359][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:41:47,858][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:41:48,360][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:41:48,859][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:41:49,358][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:41:49,862][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:41:50,362][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:41:50,860][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:41:51,357][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:41:51,859][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:41:52,356][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:41:52,856][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:41:53,353][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:41:53,850][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:41:54,349][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:41:54,847][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:41:55,348][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:41:55,845][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:41:56,341][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:41:56,837][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:41:57,337][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:41:57,842][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:41:58,342][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:41:58,842][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:41:59,343][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:41:59,843][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:42:00,342][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:42:00,839][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:42:01,337][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:42:01,838][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:42:02,335][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:42:02,830][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:42:03,327][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:42:03,824][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:42:04,319][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:42:04,813][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:42:05,311][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:42:05,809][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:42:06,314][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:42:06,813][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:42:07,311][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:42:07,807][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:42:08,303][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:42:08,800][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:42:09,296][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:42:09,790][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:42:10,285][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:42:10,781][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:42:11,276][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:42:11,772][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:42:12,267][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:42:12,763][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:42:13,259][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:42:13,754][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:42:14,249][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:42:14,744][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:42:15,238][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:42:15,734][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:42:16,231][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:42:16,727][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9967 tokens. +[2026-03-25 21:42:17,420][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.01%, Current % of VRAM taken: 60.44%, Block Peak % of device VRAM: 62.00%, ΔTime: 00:00:32 +[2026-03-25 21:42:18,145][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:42:18,147][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:42:18,148][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:42:19,385][__main__][INFO] - Iteration 41 took 49s (27.81% Gen, 69.71% Train). Generation: 13s, Training: 34s. Estimated remaining time: 40h 56m 49s. Estimated total time: 41h 33m 28s. Time estimates for 10 more iterations: 8m 18s, 100 more iterations: 1h 23m 6s, 500 more iterations: 6h 55m 34s. +[2026-03-25 21:42:19,388][__main__][INFO] - Starting iteration 41. +[2026-03-25 21:42:19,789][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 4 and human policies 1. +[2026-03-25 21:42:19,790][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:42:34,040][__main__][INFO] - Number of regex retries in iteration 41: 0 +[2026-03-25 21:42:34,041][__main__][INFO] - agents played in iteration 41 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:42:34,818][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.45%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:42:34,841][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.45%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:42:34,865][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.45%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:42:34,889][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.45%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:42:34,889][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:42:34,890][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:42:35,548][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:42:36,002][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:42:36,504][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:42:37,000][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:42:37,498][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:42:38,002][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:42:38,497][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:42:38,993][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:42:39,491][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:42:39,987][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:42:40,482][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:42:40,976][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:42:41,470][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:42:41,965][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:42:42,461][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:42:42,957][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:42:43,451][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:42:43,947][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:42:44,443][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:42:44,938][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:42:45,434][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:42:45,932][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:42:46,429][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:42:46,923][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:42:47,420][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:42:47,916][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:42:48,413][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:42:48,909][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:42:49,404][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:42:49,898][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:42:50,393][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:42:50,888][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:42:51,384][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:42:51,880][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:42:52,380][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:42:52,877][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:42:53,373][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:42:53,868][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:42:54,364][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:42:54,860][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:42:55,358][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:42:55,854][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:42:56,349][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:42:56,846][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:42:57,342][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:42:57,837][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:42:58,332][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:42:58,828][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:42:59,324][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:42:59,821][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:43:00,319][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:43:00,815][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:43:01,313][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:43:01,812][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:43:02,311][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:43:02,818][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:43:03,314][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:43:03,809][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:43:04,303][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:43:04,799][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:43:05,297][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:43:05,792][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:43:06,289][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:43:06,785][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:43:07,281][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10073 tokens. +[2026-03-25 21:43:07,946][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.02%, Current % of VRAM taken: 60.45%, Block Peak % of device VRAM: 62.14%, ΔTime: 00:00:32 +[2026-03-25 21:43:08,670][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:43:08,673][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:43:08,674][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:43:09,330][__main__][INFO] - Iteration 42 took 49s (28.77% Gen, 69.91% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 39m 35s. Estimated total time: 41h 17m 4s. Time estimates for 10 more iterations: 8m 15s, 100 more iterations: 1h 22m 34s, 500 more iterations: 6h 52m 50s. +[2026-03-25 21:43:09,333][__main__][INFO] - Starting iteration 42. +[2026-03-25 21:43:09,734][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 4 and human policies 1. +[2026-03-25 21:43:09,734][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:43:11,203][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 21:43:23,681][__main__][INFO] - Number of regex retries in iteration 42: 1 +[2026-03-25 21:43:23,682][__main__][INFO] - agents played in iteration 42 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:43:24,482][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:43:24,505][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:43:24,529][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:43:24,553][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:43:24,554][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:43:24,554][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:43:25,194][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:43:25,651][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:43:26,162][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:43:26,658][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:43:27,159][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:43:27,657][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:43:28,154][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:43:28,650][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:43:29,146][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:43:29,642][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:43:30,137][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:43:30,633][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:43:31,131][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:43:31,629][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:43:32,128][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:43:32,626][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:43:33,121][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:43:33,618][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:43:34,113][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:43:34,607][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:43:35,102][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:43:35,598][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:43:36,092][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:43:36,589][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:43:37,084][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:43:37,578][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:43:38,072][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:43:38,567][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:43:39,064][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:43:39,558][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:43:40,054][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:43:40,549][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:43:41,045][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:43:41,545][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:43:42,043][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:43:42,543][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:43:43,041][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:43:43,538][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:43:44,043][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:43:44,547][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:43:45,043][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:43:45,544][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:43:46,041][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:43:46,540][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:43:47,038][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:43:47,535][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:43:48,031][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:43:48,530][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:43:49,026][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:43:49,522][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:43:50,016][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:43:50,511][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:43:51,005][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:43:51,498][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:43:51,991][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:43:52,484][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:43:52,977][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:43:53,475][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:43:53,972][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:43:54,475][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:43:54,971][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:43:55,470][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:43:55,967][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:43:56,463][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:43:56,965][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9999 tokens. +[2026-03-25 21:43:57,656][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 60.49%, Block Peak % of device VRAM: 62.12%, ΔTime: 00:00:32 +[2026-03-25 21:43:58,384][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:43:58,386][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:43:58,388][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:43:59,056][__main__][INFO] - Iteration 43 took 49s (28.28% Gen, 70.36% Train). Generation: 13s, Training: 34s. Estimated remaining time: 40h 27m 50s. Estimated total time: 41h 6m 8s. Time estimates for 10 more iterations: 8m 13s, 100 more iterations: 1h 22m 12s, 500 more iterations: 6h 51m 1s. +[2026-03-25 21:43:59,059][__main__][INFO] - Starting iteration 43. +[2026-03-25 21:43:59,462][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 4 and human policies 1. +[2026-03-25 21:43:59,463][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:44:00,894][mllm.models.large_language_model_local][WARNING] - Response Proposal: 5 hats, 0 books, 5 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 21:44:13,546][__main__][INFO] - Number of regex retries in iteration 43: 1 +[2026-03-25 21:44:13,547][__main__][INFO] - agents played in iteration 43 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:44:14,332][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:44:14,356][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:44:14,379][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:44:14,403][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:44:14,403][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:44:14,404][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:44:15,056][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:44:15,507][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:44:16,011][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:44:16,511][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:44:17,011][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:44:17,510][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:44:18,014][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:44:18,512][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:44:19,012][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:44:19,510][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:44:20,010][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:44:20,508][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:44:21,007][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:44:21,509][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:44:22,005][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:44:22,502][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:44:23,001][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:44:23,497][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:44:23,993][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:44:24,489][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:44:24,986][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:44:25,482][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:44:25,979][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:44:26,475][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:44:26,972][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:44:27,467][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:44:27,965][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:44:28,462][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:44:28,961][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:44:29,457][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:44:29,954][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:44:30,451][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:44:30,953][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:44:31,452][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:44:31,952][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:44:32,449][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:44:32,948][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:44:33,448][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:44:33,949][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:44:34,447][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:44:34,950][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:44:35,445][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:44:35,942][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:44:36,440][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:44:36,936][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:44:37,436][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:44:37,934][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:44:38,430][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:44:38,927][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:44:39,424][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:44:39,921][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:44:40,416][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:44:40,915][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:44:41,416][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:44:41,913][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:44:42,410][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:44:42,907][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:44:43,404][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:44:43,902][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:44:44,399][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:44:44,900][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:44:45,397][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:44:45,894][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:44:46,396][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:44:46,895][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10067 tokens. +[2026-03-25 21:44:47,585][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 61.97%, ΔTime: 00:00:32 +[2026-03-25 21:44:48,307][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:44:48,309][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:44:48,311][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:44:48,975][__main__][INFO] - Iteration 44 took 49s (28.45% Gen, 70.21% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 36m 31s. Estimated total time: 41h 15m 39s. Time estimates for 10 more iterations: 8m 15s, 100 more iterations: 1h 22m 31s, 500 more iterations: 6h 52m 36s. +[2026-03-25 21:44:48,977][__main__][INFO] - Starting iteration 44. +[2026-03-25 21:44:49,379][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 4 and human policies 1. +[2026-03-25 21:44:49,380][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:44:51,079][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 21:45:04,632][__main__][INFO] - Number of regex retries in iteration 44: 1 +[2026-03-25 21:45:04,634][__main__][INFO] - agents played in iteration 44 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:45:05,436][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:45:05,459][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:45:05,484][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:45:05,507][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:45:05,508][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:45:05,508][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:45:06,173][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:45:06,630][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:45:07,137][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:45:07,639][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:45:08,136][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:45:08,632][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:45:09,132][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:45:09,629][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:45:10,128][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:45:10,626][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:45:11,124][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:45:11,621][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:45:12,117][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:45:12,613][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:45:13,111][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:45:13,612][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:45:14,110][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:45:14,608][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:45:15,107][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:45:15,606][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:45:16,104][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:45:16,602][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:45:17,099][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:45:17,599][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:45:18,097][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:45:18,596][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:45:19,092][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:45:19,597][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:45:20,096][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:45:20,593][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:45:21,090][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:45:21,586][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:45:22,084][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:45:22,580][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:45:23,077][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:45:23,573][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:45:24,069][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:45:24,567][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:45:25,067][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:45:25,566][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:45:26,067][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:45:26,566][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:45:27,064][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:45:27,565][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:45:28,062][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:45:28,564][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:45:29,060][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:45:29,557][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:45:30,052][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:45:30,550][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:45:31,053][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:45:31,551][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:45:32,052][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:45:32,550][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:45:33,048][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:45:33,547][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:45:34,043][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:45:34,543][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:45:35,043][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:45:35,547][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:45:36,047][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:45:36,546][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:45:37,044][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:45:37,542][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:45:38,042][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10138 tokens. +[2026-03-25 21:45:38,740][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.17%, ΔTime: 00:00:32 +[2026-03-25 21:45:39,460][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:45:39,463][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:45:39,465][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:45:40,123][__main__][INFO] - Iteration 45 took 50s (30.06% Gen, 68.64% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 37m 13s. Estimated total time: 42h 17m 13s. Time estimates for 10 more iterations: 8m 27s, 100 more iterations: 1h 24m 34s, 500 more iterations: 7h 2m 52s. +[2026-03-25 21:45:40,125][__main__][INFO] - Starting iteration 45. +[2026-03-25 21:45:40,526][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 4 and human policies 1. +[2026-03-25 21:45:40,527][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:45:54,986][__main__][INFO] - Number of regex retries in iteration 45: 0 +[2026-03-25 21:45:54,987][__main__][INFO] - agents played in iteration 45 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:45:55,802][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:45:55,826][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:45:55,849][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:45:55,873][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:45:55,873][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:45:55,874][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:45:56,521][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:45:56,975][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:45:57,476][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:45:57,975][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:45:58,477][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:45:58,980][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:45:59,477][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:45:59,974][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:46:00,471][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:46:00,968][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:46:01,465][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:46:01,967][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:46:02,463][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:46:02,960][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:46:03,455][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:46:03,950][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:46:04,447][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:46:04,947][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:46:05,463][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:46:05,964][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:46:06,464][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:46:06,962][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:46:07,466][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:46:07,965][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:46:08,465][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:46:08,960][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:46:09,456][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:46:09,953][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:46:10,450][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:46:10,947][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:46:11,444][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:46:11,938][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:46:12,434][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:46:12,930][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:46:13,426][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:46:13,920][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:46:14,414][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:46:14,909][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:46:15,404][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:46:15,899][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:46:16,393][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:46:16,888][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:46:17,383][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:46:17,878][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:46:18,373][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:46:18,872][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:46:19,369][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:46:19,865][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:46:20,361][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:46:20,859][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:46:21,357][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:46:21,855][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:46:22,353][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:46:22,850][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:46:23,349][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:46:23,845][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:46:24,341][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:46:24,839][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:46:25,335][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:46:25,833][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:46:26,331][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:46:26,828][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:46:27,323][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:46:27,817][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:46:28,316][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10138 tokens. +[2026-03-25 21:46:28,965][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.58%, Block Peak % of device VRAM: 62.07%, ΔTime: 00:00:32 +[2026-03-25 21:46:29,697][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:46:29,700][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:46:29,701][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:46:30,349][__main__][INFO] - Iteration 46 took 49s (29.02% Gen, 69.67% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 50m 21s. Estimated total time: 41h 31m 11s. Time estimates for 10 more iterations: 8m 18s, 100 more iterations: 1h 23m 2s, 500 more iterations: 6h 55m 11s. +[2026-03-25 21:46:30,352][__main__][INFO] - Starting iteration 46. +[2026-03-25 21:46:30,750][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 4 and human policies 1. +[2026-03-25 21:46:30,751][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:46:45,085][__main__][INFO] - Number of regex retries in iteration 46: 0 +[2026-03-25 21:46:45,086][__main__][INFO] - agents played in iteration 46 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:46:45,906][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:46:45,930][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:46:45,954][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:46:45,977][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:46:45,978][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:46:45,978][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:46:46,600][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:46:47,058][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:46:47,562][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:46:48,067][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:46:48,566][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:46:49,064][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:46:49,563][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:46:50,062][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:46:50,560][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:46:51,056][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:46:51,555][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:46:52,055][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:46:52,553][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:46:53,051][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:46:53,548][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:46:54,044][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:46:54,540][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:46:55,037][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:46:55,534][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:46:56,030][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:46:56,528][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:46:57,025][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:46:57,521][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:46:58,020][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:46:58,517][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:46:59,011][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:46:59,510][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:47:00,005][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:47:00,499][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:47:00,995][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:47:01,490][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:47:01,989][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:47:02,486][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:47:02,987][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:47:03,487][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:47:03,983][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:47:04,482][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:47:04,979][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:47:05,477][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:47:05,974][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:47:06,473][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:47:06,970][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:47:07,469][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:47:07,967][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:47:08,466][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:47:08,964][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:47:09,459][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:47:09,959][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:47:10,454][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:47:10,950][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:47:11,448][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:47:11,947][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:47:12,447][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:47:12,943][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:47:13,442][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:47:13,938][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:47:14,436][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:47:14,932][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:47:15,430][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:47:15,928][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:47:16,425][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:47:16,921][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:47:17,417][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:47:17,911][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:47:18,405][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10036 tokens. +[2026-03-25 21:47:19,073][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 60.49%, Block Peak % of device VRAM: 62.08%, ΔTime: 00:00:32 +[2026-03-25 21:47:19,795][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:47:19,798][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:47:19,799][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:47:20,451][__main__][INFO] - Iteration 47 took 49s (28.84% Gen, 69.84% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 43m 22s. Estimated total time: 41h 25m 2s. Time estimates for 10 more iterations: 8m 17s, 100 more iterations: 1h 22m 50s, 500 more iterations: 6h 54m 10s. +[2026-03-25 21:47:20,453][__main__][INFO] - Starting iteration 47. +[2026-03-25 21:47:20,854][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 4 and human policies 1. +[2026-03-25 21:47:20,855][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:47:26,382][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 21:47:35,367][__main__][INFO] - Number of regex retries in iteration 47: 1 +[2026-03-25 21:47:35,368][__main__][INFO] - agents played in iteration 47 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:47:36,150][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:47:36,174][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:47:36,197][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:47:36,220][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:47:36,221][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:47:36,221][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:47:36,835][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:47:37,292][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:47:37,794][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:47:38,292][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:47:38,793][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:47:39,294][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:47:39,792][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:47:40,292][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:47:40,788][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:47:41,287][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:47:41,783][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:47:42,284][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:47:42,778][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:47:43,278][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:47:43,776][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:47:44,272][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:47:44,769][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:47:45,268][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:47:45,768][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:47:46,266][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:47:46,764][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:47:47,262][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:47:47,760][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:47:48,258][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:47:48,756][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:47:49,254][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:47:49,753][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:47:50,251][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:47:50,749][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:47:51,245][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:47:51,741][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:47:52,238][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:47:52,737][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:47:53,235][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:47:53,732][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:47:54,229][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:47:54,725][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:47:55,225][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:47:55,721][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:47:56,216][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:47:56,712][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:47:57,207][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:47:57,704][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:47:58,199][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:47:58,693][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:47:59,189][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:47:59,685][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:48:00,181][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:48:00,677][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:48:01,176][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:48:01,671][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:48:02,169][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:48:02,663][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:48:03,158][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:48:03,654][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:48:04,150][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:48:04,647][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:48:05,143][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:48:05,638][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:48:06,134][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:48:06,632][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:48:07,130][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:48:07,625][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:48:08,120][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:48:08,616][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10174 tokens. +[2026-03-25 21:48:09,289][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 60.50%, Block Peak % of device VRAM: 62.05%, ΔTime: 00:00:32 +[2026-03-25 21:48:10,016][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:48:10,018][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:48:10,020][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:48:10,673][__main__][INFO] - Iteration 48 took 49s (29.13% Gen, 69.55% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 48m 28s. Estimated total time: 41h 30m 58s. Time estimates for 10 more iterations: 8m 18s, 100 more iterations: 1h 23m 1s, 500 more iterations: 6h 55m 9s. +[2026-03-25 21:48:10,675][__main__][INFO] - Starting iteration 48. +[2026-03-25 21:48:11,076][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 4 and human policies 1. +[2026-03-25 21:48:11,077][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:48:25,087][__main__][INFO] - Number of regex retries in iteration 48: 0 +[2026-03-25 21:48:25,088][__main__][INFO] - agents played in iteration 48 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:48:25,865][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:48:25,889][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:48:25,912][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:48:25,935][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:48:25,936][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:48:25,936][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:48:26,536][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:48:26,987][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:48:27,496][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:48:27,997][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:48:28,495][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:48:28,998][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:48:29,496][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:48:29,995][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:48:30,492][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:48:30,988][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:48:31,482][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:48:31,977][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:48:32,471][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:48:32,965][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:48:33,460][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:48:33,954][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:48:34,450][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:48:34,947][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:48:35,443][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:48:35,939][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:48:36,435][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:48:36,931][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:48:37,427][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:48:37,922][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:48:38,417][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:48:38,912][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:48:39,408][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:48:39,902][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:48:40,397][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:48:40,892][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:48:41,388][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:48:41,883][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:48:42,377][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:48:42,873][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:48:43,369][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:48:43,865][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:48:44,362][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:48:44,858][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:48:45,355][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:48:45,852][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:48:46,348][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:48:46,844][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:48:47,339][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:48:47,834][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:48:48,330][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:48:48,825][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:48:49,320][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:48:49,815][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:48:50,310][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:48:50,805][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:48:51,301][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:48:51,802][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:48:52,300][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:48:52,795][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:48:53,294][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:48:53,794][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:48:54,295][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:48:54,793][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:48:55,290][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:48:55,787][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:48:56,289][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:48:56,788][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:48:57,290][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:48:57,786][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:48:58,283][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10017 tokens. +[2026-03-25 21:48:58,965][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 60.53%, Block Peak % of device VRAM: 61.93%, ΔTime: 00:00:32 +[2026-03-25 21:48:59,693][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:48:59,696][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:48:59,697][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:49:00,363][__main__][INFO] - Iteration 49 took 49s (28.43% Gen, 70.22% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 21m 1s. Estimated total time: 41h 4m 21s. Time estimates for 10 more iterations: 8m 12s, 100 more iterations: 1h 22m 8s, 500 more iterations: 6h 50m 43s. +[2026-03-25 21:49:00,365][__main__][INFO] - Starting iteration 49. +[2026-03-25 21:49:00,765][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 4 and human policies 1. +[2026-03-25 21:49:00,766][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:49:15,392][__main__][INFO] - Number of regex retries in iteration 49: 0 +[2026-03-25 21:49:15,393][__main__][INFO] - agents played in iteration 49 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:49:16,183][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:49:16,207][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:49:16,230][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:49:16,254][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:49:16,254][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:49:16,255][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:49:16,891][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:49:17,346][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:49:17,850][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:49:18,350][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:49:18,851][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:49:19,350][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:49:19,849][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:49:20,354][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:49:20,851][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:49:21,349][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:49:21,850][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:49:22,351][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:49:22,848][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:49:23,348][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:49:23,844][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:49:24,343][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:49:24,839][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:49:25,334][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:49:25,830][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:49:26,325][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:49:26,818][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:49:27,312][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:49:27,808][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:49:28,303][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:49:28,799][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:49:29,295][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:49:29,795][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:49:30,291][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:49:30,788][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:49:31,282][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:49:31,783][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:49:32,279][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:49:32,776][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:49:33,274][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:49:33,770][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:49:34,268][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:49:34,761][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:49:35,261][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:49:35,757][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:49:36,256][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:49:36,755][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:49:37,253][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:49:37,748][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:49:38,242][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:49:38,738][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:49:39,235][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:49:39,729][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:49:40,223][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:49:40,718][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:49:41,215][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:49:41,711][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:49:42,207][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:49:42,704][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:49:43,209][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:49:43,709][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:49:44,208][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:49:44,707][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:49:45,201][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:49:45,695][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:49:46,191][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:49:46,686][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:49:47,181][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:49:47,676][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:49:48,172][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:49:48,669][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10060 tokens. +[2026-03-25 21:49:49,340][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 60.49%, Block Peak % of device VRAM: 62.08%, ΔTime: 00:00:32 +[2026-03-25 21:49:50,036][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:49:50,039][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:49:50,040][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:49:50,702][__main__][INFO] - Iteration 50 took 49s (29.29% Gen, 69.38% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 52m 42s. Estimated total time: 41h 36m 52s. Time estimates for 10 more iterations: 8m 19s, 100 more iterations: 1h 23m 13s, 500 more iterations: 6h 56m 8s. +[2026-03-25 21:49:50,705][__main__][INFO] - Starting iteration 50. +[2026-03-25 21:49:51,105][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 4 and human policies 1. +[2026-03-25 21:49:51,106][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:49:55,975][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 21:49:57,408][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 21:50:05,263][__main__][INFO] - Number of regex retries in iteration 50: 2 +[2026-03-25 21:50:05,264][__main__][INFO] - agents played in iteration 50 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:50:06,042][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:50:06,066][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:50:06,090][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:50:06,113][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:50:06,114][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:50:06,114][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:50:06,750][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:50:07,206][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:50:07,708][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:50:08,203][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:50:08,699][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:50:09,195][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:50:09,690][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:50:10,188][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:50:10,685][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:50:11,180][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:50:11,675][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:50:12,171][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:50:12,669][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:50:13,165][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:50:13,661][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:50:14,156][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:50:14,652][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:50:15,158][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:50:15,657][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:50:16,157][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:50:16,657][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:50:17,156][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:50:17,654][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:50:18,155][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:50:18,658][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:50:19,154][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:50:19,652][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:50:20,150][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:50:20,649][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:50:21,146][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:50:21,642][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:50:22,139][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:50:22,635][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:50:23,131][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:50:23,626][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:50:24,120][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:50:24,616][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:50:25,112][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:50:25,609][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:50:26,105][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:50:26,599][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:50:27,098][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:50:27,597][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:50:28,094][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:50:28,591][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:50:29,087][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:50:29,584][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:50:30,083][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:50:30,581][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:50:31,080][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:50:31,578][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:50:32,080][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:50:32,578][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:50:33,078][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:50:33,576][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:50:34,079][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:50:34,577][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:50:35,073][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:50:35,571][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:50:36,067][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:50:36,563][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:50:37,059][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:50:37,555][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:50:38,052][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:50:38,548][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10111 tokens. +[2026-03-25 21:50:39,219][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 60.51%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:32 +[2026-03-25 21:50:39,940][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:50:39,942][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:50:39,943][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:50:41,162][__main__][INFO] - Iteration 51 took 50s (28.28% Gen, 69.28% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 57m 52s. Estimated total time: 41h 42m 53s. Time estimates for 10 more iterations: 8m 20s, 100 more iterations: 1h 23m 25s, 500 more iterations: 6h 57m 8s. +[2026-03-25 21:50:41,165][__main__][INFO] - Starting iteration 51. +[2026-03-25 21:50:41,564][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 5 and human policies 1. +[2026-03-25 21:50:41,564][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:50:56,502][__main__][INFO] - Number of regex retries in iteration 51: 0 +[2026-03-25 21:50:56,503][__main__][INFO] - agents played in iteration 51 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:50:57,282][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:50:57,305][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:50:57,329][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:50:57,352][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:50:57,352][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:50:57,353][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:50:57,971][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:50:58,423][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:50:58,920][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:50:59,419][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:50:59,913][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:51:00,408][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:51:00,912][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:51:01,408][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:51:01,905][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:51:02,405][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:51:02,925][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:51:03,421][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:51:03,916][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:51:04,416][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:51:04,911][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:51:05,408][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:51:05,905][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:51:06,402][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:51:06,903][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:51:07,398][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:51:07,896][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:51:08,395][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:51:08,897][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:51:09,399][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:51:09,901][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:51:10,399][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:51:10,898][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:51:11,394][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:51:11,894][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:51:12,392][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:51:12,891][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:51:13,387][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:51:13,889][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:51:14,388][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:51:14,884][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:51:15,385][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:51:15,891][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:51:16,390][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:51:16,889][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:51:17,388][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:51:17,888][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:51:18,383][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:51:18,879][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:51:19,378][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:51:19,873][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:51:20,370][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:51:20,869][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:51:21,368][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:51:21,865][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:51:22,361][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:51:22,856][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:51:23,352][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:51:23,848][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:51:24,344][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:51:24,840][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:51:25,338][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:51:25,834][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:51:26,329][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:51:26,825][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:51:27,322][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:51:27,820][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:51:28,316][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:51:28,814][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:51:29,313][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:51:29,810][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10214 tokens. +[2026-03-25 21:51:30,482][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 60.52%, Block Peak % of device VRAM: 62.14%, ΔTime: 00:00:32 +[2026-03-25 21:51:31,202][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:51:31,204][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:51:31,206][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:51:31,864][__main__][INFO] - Iteration 52 took 50s (29.70% Gen, 68.99% Train). Generation: 14s, Training: 34s. Estimated remaining time: 41h 9m 10s. Estimated total time: 41h 55m 1s. Time estimates for 10 more iterations: 8m 23s, 100 more iterations: 1h 23m 50s, 500 more iterations: 6h 59m 10s. +[2026-03-25 21:51:31,866][__main__][INFO] - Starting iteration 52. +[2026-03-25 21:51:32,267][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 5 and human policies 1. +[2026-03-25 21:51:32,267][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:51:46,335][__main__][INFO] - Number of regex retries in iteration 52: 0 +[2026-03-25 21:51:46,336][__main__][INFO] - agents played in iteration 52 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:51:47,125][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:51:47,148][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:51:47,171][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:51:47,194][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:51:47,195][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:51:47,195][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:51:47,799][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:51:48,250][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:51:48,750][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:51:49,247][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:51:49,743][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:51:50,240][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:51:50,742][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:51:51,238][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:51:51,735][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:51:52,254][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:51:52,750][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:51:53,248][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:51:53,745][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:51:54,243][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:51:54,740][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:51:55,241][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:51:55,739][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:51:56,235][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:51:56,731][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:51:57,229][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:51:57,725][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:51:58,221][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:51:58,717][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:51:59,212][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:51:59,709][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:52:00,206][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:52:00,702][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:52:01,199][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:52:01,695][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:52:02,193][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:52:02,689][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:52:03,186][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:52:03,683][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:52:04,188][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:52:04,689][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:52:05,188][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:52:05,687][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:52:06,184][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:52:06,684][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:52:07,184][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:52:07,706][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:52:08,206][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:52:08,701][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:52:09,198][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:52:09,694][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:52:10,191][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:52:10,689][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:52:11,186][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:52:11,683][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:52:12,182][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:52:12,680][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:52:13,180][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:52:13,677][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:52:14,177][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:52:14,680][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:52:15,178][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:52:15,674][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:52:16,173][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:52:16,671][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:52:17,168][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:52:17,664][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:52:18,161][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:52:18,658][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:52:19,155][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:52:19,654][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10176 tokens. +[2026-03-25 21:52:20,319][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.18%, ΔTime: 00:00:32 +[2026-03-25 21:52:21,035][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:52:21,037][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:52:21,039][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:52:21,691][__main__][INFO] - Iteration 53 took 49s (28.46% Gen, 70.21% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 24m 34s. Estimated total time: 41h 11m 15s. Time estimates for 10 more iterations: 8m 14s, 100 more iterations: 1h 22m 22s, 500 more iterations: 6h 51m 52s. +[2026-03-25 21:52:21,694][__main__][INFO] - Starting iteration 53. +[2026-03-25 21:52:22,093][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 5 and human policies 1. +[2026-03-25 21:52:22,094][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:52:36,913][__main__][INFO] - Number of regex retries in iteration 53: 0 +[2026-03-25 21:52:36,914][__main__][INFO] - agents played in iteration 53 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:52:37,701][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:52:37,724][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:52:37,748][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:52:37,770][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:52:37,771][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:52:37,771][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:52:38,376][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:52:38,827][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:52:39,332][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:52:39,830][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:52:40,330][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:52:40,828][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:52:41,325][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:52:41,821][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:52:42,316][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:52:42,811][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:52:43,310][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:52:43,806][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:52:44,302][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:52:44,798][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:52:45,294][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:52:45,791][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:52:46,285][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:52:46,788][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:52:47,287][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:52:47,785][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:52:48,286][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:52:48,781][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:52:49,286][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:52:49,784][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:52:50,280][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:52:50,800][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:52:51,298][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:52:51,803][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:52:52,298][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:52:52,795][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:52:53,293][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:52:53,791][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:52:54,286][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:52:54,789][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:52:55,289][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:52:55,787][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:52:56,285][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:52:56,782][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:52:57,278][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:52:57,781][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:52:58,279][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:52:58,775][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:52:59,271][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:52:59,768][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:53:00,266][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:53:00,762][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:53:01,259][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:53:01,757][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:53:02,254][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:53:02,750][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:53:03,245][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:53:03,741][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:53:04,237][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:53:04,734][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:53:05,229][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:53:05,723][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:53:06,218][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:53:06,714][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:53:07,210][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:53:07,714][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:53:08,213][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:53:08,712][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:53:09,210][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:53:09,708][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:53:10,205][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10081 tokens. +[2026-03-25 21:53:10,900][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 60.54%, Block Peak % of device VRAM: 62.07%, ΔTime: 00:00:32 +[2026-03-25 21:53:11,621][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:53:11,624][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:53:11,625][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:53:12,282][__main__][INFO] - Iteration 54 took 50s (29.53% Gen, 69.16% Train). Generation: 14s, Training: 34s. Estimated remaining time: 41h 1m 55s. Estimated total time: 41h 49m 27s. Time estimates for 10 more iterations: 8m 21s, 100 more iterations: 1h 23m 38s, 500 more iterations: 6h 58m 14s. +[2026-03-25 21:53:12,284][__main__][INFO] - Starting iteration 54. +[2026-03-25 21:53:12,685][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 5 and human policies 1. +[2026-03-25 21:53:12,686][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:53:14,335][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 21:53:14,408][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 21:53:27,614][__main__][INFO] - Number of regex retries in iteration 54: 2 +[2026-03-25 21:53:27,615][__main__][INFO] - agents played in iteration 54 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:53:28,396][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:53:28,419][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:53:28,442][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:53:28,465][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:53:28,466][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:53:28,466][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:53:29,101][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:53:29,554][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:53:30,055][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:53:30,558][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:53:31,059][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:53:31,558][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:53:32,057][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:53:32,559][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:53:33,055][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:53:33,552][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:53:34,050][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:53:34,547][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:53:35,045][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:53:35,543][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:53:36,042][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:53:36,539][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:53:37,036][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:53:37,534][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:53:38,032][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:53:38,531][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:53:39,028][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:53:39,524][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:53:40,041][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:53:40,541][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:53:41,037][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:53:41,536][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:53:42,034][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:53:42,534][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:53:43,031][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:53:43,531][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:53:44,029][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:53:44,528][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:53:45,027][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:53:45,523][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:53:46,019][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:53:46,515][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:53:47,012][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:53:47,510][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:53:48,007][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:53:48,506][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:53:49,001][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:53:49,497][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:53:49,994][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:53:50,490][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:53:50,986][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:53:51,481][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:53:51,978][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:53:52,473][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:53:52,971][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:53:53,468][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:53:53,969][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:53:54,465][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:53:54,965][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:53:55,464][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:53:55,961][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:53:56,462][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:53:56,958][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:53:57,455][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:53:57,950][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:53:58,448][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:53:58,945][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:53:59,442][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:53:59,937][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:54:00,436][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:54:00,931][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10233 tokens. +[2026-03-25 21:54:01,587][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.02%, Current % of VRAM taken: 60.45%, Block Peak % of device VRAM: 62.11%, ΔTime: 00:00:32 +[2026-03-25 21:54:02,308][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:54:02,311][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:54:02,312][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:54:02,995][__main__][INFO] - Iteration 55 took 50s (29.68% Gen, 68.97% Train). Generation: 14s, Training: 34s. Estimated remaining time: 41h 7m 8s. Estimated total time: 41h 55m 31s. Time estimates for 10 more iterations: 8m 23s, 100 more iterations: 1h 23m 51s, 500 more iterations: 6h 59m 15s. +[2026-03-25 21:54:02,997][__main__][INFO] - Starting iteration 55. +[2026-03-25 21:54:03,398][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 5 and human policies 1. +[2026-03-25 21:54:03,399][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:54:18,365][__main__][INFO] - Number of regex retries in iteration 55: 0 +[2026-03-25 21:54:18,367][__main__][INFO] - agents played in iteration 55 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:54:19,152][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:54:19,175][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:54:19,199][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:54:19,223][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:54:19,223][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:54:19,224][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:54:19,859][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:54:20,312][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:54:20,813][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:54:21,310][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:54:21,809][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:54:22,310][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:54:22,807][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:54:23,307][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:54:23,803][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:54:24,299][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:54:24,794][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:54:25,289][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:54:25,785][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:54:26,281][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:54:26,777][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:54:27,274][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:54:27,771][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:54:28,269][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:54:28,766][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:54:29,263][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:54:29,758][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:54:30,255][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:54:30,753][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:54:31,252][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:54:31,756][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:54:32,253][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:54:32,749][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:54:33,245][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:54:33,741][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:54:34,238][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:54:34,736][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:54:35,233][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:54:35,730][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:54:36,228][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:54:36,727][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:54:37,223][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:54:37,719][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:54:38,215][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:54:38,711][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:54:39,206][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:54:39,702][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:54:40,199][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:54:40,695][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:54:41,214][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:54:41,712][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:54:42,209][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:54:42,706][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:54:43,202][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:54:43,700][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:54:44,196][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:54:44,693][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:54:45,192][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:54:45,691][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:54:46,189][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:54:46,689][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:54:47,186][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:54:47,684][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:54:48,182][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:54:48,677][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:54:49,174][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:54:49,672][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:54:50,168][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:54:50,663][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:54:51,160][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:54:51,660][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10162 tokens. +[2026-03-25 21:54:52,317][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 60.60%, Block Peak % of device VRAM: 62.03%, ΔTime: 00:00:32 +[2026-03-25 21:54:53,039][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:54:53,042][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:54:53,044][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:54:53,697][__main__][INFO] - Iteration 56 took 50s (29.76% Gen, 68.94% Train). Generation: 14s, Training: 34s. Estimated remaining time: 41h 5m 43s. Estimated total time: 41h 54m 57s. Time estimates for 10 more iterations: 8m 22s, 100 more iterations: 1h 23m 49s, 500 more iterations: 6h 59m 9s. +[2026-03-25 21:54:53,699][__main__][INFO] - Starting iteration 56. +[2026-03-25 21:54:54,097][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 5 and human policies 1. +[2026-03-25 21:54:54,098][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:55:08,399][__main__][INFO] - Number of regex retries in iteration 56: 0 +[2026-03-25 21:55:08,400][__main__][INFO] - agents played in iteration 56 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:55:09,201][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:55:09,226][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:55:09,250][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:55:09,274][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:55:09,274][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:55:09,275][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:55:09,898][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:55:10,353][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:55:10,857][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:55:11,354][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:55:11,851][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:55:12,350][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:55:12,848][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:55:13,350][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:55:13,845][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:55:14,342][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:55:14,838][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:55:15,338][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:55:15,836][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:55:16,334][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:55:16,830][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:55:17,324][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:55:17,822][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:55:18,316][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:55:18,811][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:55:19,315][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:55:19,812][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:55:20,309][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:55:20,811][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:55:21,309][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:55:21,805][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:55:22,300][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:55:22,796][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:55:23,293][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:55:23,789][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:55:24,285][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:55:24,782][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:55:25,277][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:55:25,773][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:55:26,271][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:55:26,767][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:55:27,262][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:55:27,759][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:55:28,256][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:55:28,752][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:55:29,248][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:55:29,744][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:55:30,239][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:55:30,734][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:55:31,233][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:55:31,732][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:55:32,230][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:55:32,724][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:55:33,219][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:55:33,714][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:55:34,210][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:55:34,705][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:55:35,201][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:55:35,697][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:55:36,195][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:55:36,693][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:55:37,192][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:55:37,687][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:55:38,182][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:55:38,683][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:55:39,178][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:55:39,673][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:55:40,168][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:55:40,665][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:55:41,162][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:55:41,661][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10195 tokens. +[2026-03-25 21:55:42,336][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 60.61%, Block Peak % of device VRAM: 61.98%, ΔTime: 00:00:32 +[2026-03-25 21:55:43,057][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:55:43,059][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:55:43,061][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:55:43,718][__main__][INFO] - Iteration 57 took 49s (28.82% Gen, 69.85% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 30m 58s. Estimated total time: 41h 21m 1s. Time estimates for 10 more iterations: 8m 16s, 100 more iterations: 1h 22m 42s, 500 more iterations: 6h 53m 30s. +[2026-03-25 21:55:43,720][__main__][INFO] - Starting iteration 57. +[2026-03-25 21:55:44,121][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 5 and human policies 1. +[2026-03-25 21:55:44,121][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:55:57,917][__main__][INFO] - Number of regex retries in iteration 57: 0 +[2026-03-25 21:55:57,918][__main__][INFO] - agents played in iteration 57 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:55:58,691][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:55:58,714][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:55:58,738][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:55:58,761][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:55:58,761][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:55:58,762][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:55:59,398][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:55:59,857][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:56:00,368][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:56:00,872][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:56:01,380][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:56:01,887][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:56:02,390][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:56:02,893][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:56:03,394][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:56:03,897][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:56:04,396][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:56:04,894][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:56:05,397][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:56:05,896][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:56:06,397][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:56:06,898][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:56:07,398][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:56:07,896][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:56:08,395][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:56:08,893][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:56:09,395][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:56:09,893][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:56:10,391][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:56:10,888][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:56:11,386][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:56:11,886][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:56:12,385][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:56:12,883][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:56:13,382][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:56:13,880][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:56:14,380][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:56:14,878][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:56:15,376][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:56:15,873][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:56:16,369][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:56:16,865][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:56:17,363][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:56:17,865][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:56:18,363][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:56:18,864][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:56:19,360][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:56:19,857][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:56:20,355][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:56:20,854][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:56:21,351][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:56:21,851][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:56:22,353][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:56:22,850][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:56:23,349][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:56:23,844][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:56:24,340][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:56:24,838][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:56:25,339][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:56:25,836][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:56:26,336][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:56:26,831][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:56:27,326][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:56:27,823][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:56:28,319][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:56:28,815][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:56:29,312][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:56:29,809][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:56:30,305][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:56:30,800][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:56:31,298][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10304 tokens. +[2026-03-25 21:56:31,980][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.57%, Block Peak % of device VRAM: 62.34%, ΔTime: 00:00:32 +[2026-03-25 21:56:32,700][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:56:32,703][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:56:32,704][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:56:33,362][__main__][INFO] - Iteration 58 took 49s (28.02% Gen, 70.64% Train). Generation: 13s, Training: 34s. Estimated remaining time: 40h 11m 13s. Estimated total time: 41h 2m 6s. Time estimates for 10 more iterations: 8m 12s, 100 more iterations: 1h 22m 4s, 500 more iterations: 6h 50m 21s. +[2026-03-25 21:56:33,365][__main__][INFO] - Starting iteration 58. +[2026-03-25 21:56:33,764][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 5 and human policies 1. +[2026-03-25 21:56:33,765][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:56:48,376][__main__][INFO] - Number of regex retries in iteration 58: 0 +[2026-03-25 21:56:48,377][__main__][INFO] - agents played in iteration 58 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:56:49,143][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:56:49,166][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:56:49,191][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:56:49,214][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:56:49,215][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:56:49,215][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:56:49,877][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:56:50,334][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:56:50,838][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:56:51,338][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:56:51,840][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:56:52,337][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:56:52,834][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:56:53,333][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:56:53,838][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:56:54,337][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:56:54,835][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:56:55,339][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:56:55,837][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:56:56,342][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:56:56,839][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:56:57,358][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:56:57,857][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:56:58,355][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:56:58,854][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:56:59,358][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:56:59,854][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:57:00,350][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:57:00,847][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:57:01,344][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:57:01,842][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:57:02,343][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:57:02,841][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:57:03,337][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:57:03,833][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:57:04,330][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:57:04,828][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:57:05,329][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:57:05,825][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:57:06,324][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:57:06,825][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:57:07,324][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:57:07,819][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:57:08,315][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:57:08,811][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:57:09,308][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:57:09,809][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:57:10,305][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:57:10,801][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:57:11,298][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:57:11,799][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:57:12,296][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:57:12,793][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:57:13,289][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:57:13,784][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:57:14,280][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:57:14,775][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:57:15,274][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:57:15,771][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:57:16,270][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:57:16,765][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:57:17,265][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:57:17,762][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:57:18,257][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:57:18,756][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:57:19,255][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:57:19,751][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:57:20,248][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:57:20,745][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:57:21,242][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:57:21,738][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10245 tokens. +[2026-03-25 21:57:22,409][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.09%, Current % of VRAM taken: 60.52%, Block Peak % of device VRAM: 62.00%, ΔTime: 00:00:32 +[2026-03-25 21:57:23,130][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:57:23,132][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:57:23,133][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:57:23,791][__main__][INFO] - Iteration 59 took 50s (29.21% Gen, 69.48% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 49m 39s. Estimated total time: 41h 41m 22s. Time estimates for 10 more iterations: 8m 20s, 100 more iterations: 1h 23m 22s, 500 more iterations: 6h 56m 53s. +[2026-03-25 21:57:23,793][__main__][INFO] - Starting iteration 59. +[2026-03-25 21:57:24,192][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 5 and human policies 1. +[2026-03-25 21:57:24,193][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:57:34,079][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 21:57:38,586][__main__][INFO] - Number of regex retries in iteration 59: 1 +[2026-03-25 21:57:38,587][__main__][INFO] - agents played in iteration 59 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:57:39,348][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:57:39,372][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:57:39,397][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:57:39,421][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:57:39,422][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:57:39,422][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:57:40,064][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:57:40,517][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:57:41,020][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:57:41,519][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:57:42,018][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:57:42,515][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:57:43,013][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:57:43,510][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:57:44,007][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:57:44,504][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:57:45,001][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:57:45,498][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:57:45,995][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:57:46,492][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:57:46,991][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:57:47,488][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:57:47,991][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:57:48,489][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:57:48,988][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:57:49,484][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:57:49,982][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:57:50,478][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:57:50,975][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:57:51,473][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:57:51,970][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:57:52,470][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:57:52,968][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:57:53,465][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:57:53,968][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:57:54,464][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:57:54,963][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:57:55,463][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:57:55,960][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:57:56,458][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:57:56,958][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:57:57,456][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:57:57,955][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:57:58,453][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:57:58,952][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:57:59,450][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:57:59,949][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:58:00,444][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:58:00,940][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:58:01,435][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:58:01,933][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:58:02,432][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:58:02,929][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:58:03,426][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:58:03,922][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:58:04,421][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:58:04,922][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:58:05,425][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:58:05,924][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:58:06,426][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:58:06,925][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:58:07,425][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:58:07,922][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:58:08,420][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:58:08,921][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:58:09,417][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:58:09,917][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:58:10,417][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:58:10,933][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:58:11,440][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:58:11,940][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10246 tokens. +[2026-03-25 21:58:12,613][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 60.66%, Block Peak % of device VRAM: 62.21%, ΔTime: 00:00:32 +[2026-03-25 21:58:13,339][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:58:13,342][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:58:13,343][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:58:14,037][__main__][INFO] - Iteration 60 took 49s (28.88% Gen, 69.73% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 39m 42s. Estimated total time: 41h 32m 16s. Time estimates for 10 more iterations: 8m 18s, 100 more iterations: 1h 23m 4s, 500 more iterations: 6h 55m 22s. +[2026-03-25 21:58:14,040][__main__][INFO] - Starting iteration 60. +[2026-03-25 21:58:14,437][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 5 and human policies 1. +[2026-03-25 21:58:14,438][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:58:28,088][__main__][INFO] - Number of regex retries in iteration 60: 0 +[2026-03-25 21:58:28,089][__main__][INFO] - agents played in iteration 60 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:58:28,846][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:58:28,870][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:58:28,893][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:58:28,916][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:58:28,916][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:58:28,917][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:58:29,551][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:58:30,007][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:58:30,506][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:58:31,006][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:58:31,505][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:58:32,004][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:58:32,503][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:58:33,000][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:58:33,497][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:58:33,991][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:58:34,484][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:58:34,979][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:58:35,474][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:58:35,971][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:58:36,469][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:58:36,965][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:58:37,461][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:58:37,957][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:58:38,453][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:58:38,950][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:58:39,445][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:58:39,940][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:58:40,435][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:58:40,931][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:58:41,428][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:58:41,929][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:58:42,429][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:58:42,925][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:58:43,430][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:58:43,931][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:58:44,433][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:58:44,932][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:58:45,432][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:58:45,931][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:58:46,428][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:58:46,924][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:58:47,421][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:58:47,917][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:58:48,414][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:58:48,911][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:58:49,407][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:58:49,903][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:58:50,398][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:58:50,893][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:58:51,389][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:58:51,884][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:58:52,379][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:58:52,874][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:58:53,370][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:58:53,866][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:58:54,362][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:58:54,858][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:58:55,355][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:58:55,853][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:58:56,349][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:58:56,845][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:58:57,344][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:58:57,839][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:58:58,335][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:58:58,832][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:58:59,329][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:58:59,825][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:59:00,320][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:59:00,816][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:59:01,312][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10055 tokens. +[2026-03-25 21:59:01,989][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 60.49%, Block Peak % of device VRAM: 62.17%, ΔTime: 00:00:32 +[2026-03-25 21:59:02,718][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:59:02,720][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:59:02,722][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:59:04,032][__main__][INFO] - Iteration 61 took 49s (27.52% Gen, 69.83% Train). Generation: 13s, Training: 34s. Estimated remaining time: 40h 26m 22s. Estimated total time: 41h 19m 46s. Time estimates for 10 more iterations: 8m 15s, 100 more iterations: 1h 22m 39s, 500 more iterations: 6h 53m 17s. +[2026-03-25 21:59:04,035][__main__][INFO] - Starting iteration 61. +[2026-03-25 21:59:04,434][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 6 and human policies 1. +[2026-03-25 21:59:04,435][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:59:19,518][__main__][INFO] - Number of regex retries in iteration 61: 0 +[2026-03-25 21:59:19,519][__main__][INFO] - agents played in iteration 61 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 21:59:20,303][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:59:20,327][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:59:20,350][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:59:20,373][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 21:59:20,374][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 21:59:20,374][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 21:59:21,005][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 21:59:21,464][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 21:59:21,966][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 21:59:22,471][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 21:59:22,971][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 21:59:23,469][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 21:59:23,972][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 21:59:24,471][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 21:59:24,971][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 21:59:25,466][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 21:59:25,961][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 21:59:26,457][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 21:59:26,952][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 21:59:27,447][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 21:59:27,942][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 21:59:28,435][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 21:59:28,933][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 21:59:29,431][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 21:59:29,926][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 21:59:30,423][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 21:59:30,921][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 21:59:31,417][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 21:59:31,918][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 21:59:32,414][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 21:59:32,911][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 21:59:33,408][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 21:59:33,903][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 21:59:34,398][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 21:59:34,892][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 21:59:35,384][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 21:59:35,879][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 21:59:36,374][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 21:59:36,870][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 21:59:37,369][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 21:59:37,868][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 21:59:38,364][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 21:59:38,861][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 21:59:39,360][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 21:59:39,858][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 21:59:40,353][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 21:59:40,849][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 21:59:41,344][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 21:59:41,839][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 21:59:42,337][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 21:59:42,833][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 21:59:43,327][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 21:59:43,821][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 21:59:44,316][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 21:59:44,811][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 21:59:45,307][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 21:59:45,806][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 21:59:46,302][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 21:59:46,800][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 21:59:47,296][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 21:59:47,792][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 21:59:48,288][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 21:59:48,781][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 21:59:49,280][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 21:59:49,775][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 21:59:50,272][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 21:59:50,766][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 21:59:51,261][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 21:59:51,757][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 21:59:52,255][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 21:59:52,754][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10147 tokens. +[2026-03-25 21:59:53,424][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 60.61%, Block Peak % of device VRAM: 62.35%, ΔTime: 00:00:32 +[2026-03-25 21:59:54,150][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 21:59:54,152][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 21:59:54,154][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 21:59:54,820][__main__][INFO] - Iteration 62 took 50s (29.94% Gen, 68.74% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 5m 3s. Estimated total time: 41h 59m 18s. Time estimates for 10 more iterations: 8m 23s, 100 more iterations: 1h 23m 58s, 500 more iterations: 6h 59m 53s. +[2026-03-25 21:59:54,822][__main__][INFO] - Starting iteration 62. +[2026-03-25 21:59:55,224][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 6 and human policies 1. +[2026-03-25 21:59:55,224][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 21:59:57,312][mllm.models.large_language_model_local][WARNING] - Response Proposition: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:00:00,834][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:00:06,976][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:00:10,591][__main__][INFO] - Number of regex retries in iteration 62: 3 +[2026-03-25 22:00:10,591][__main__][INFO] - agents played in iteration 62 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:00:11,370][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:00:11,394][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:00:11,417][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:00:11,440][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:00:11,440][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:00:11,441][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:00:12,052][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:00:12,510][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:00:13,013][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:00:13,512][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:00:14,010][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:00:14,508][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:00:15,007][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:00:15,516][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:00:16,013][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:00:16,512][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:00:17,008][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:00:17,502][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:00:18,000][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:00:18,496][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:00:18,993][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:00:19,489][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:00:19,985][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:00:20,482][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:00:20,980][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:00:21,478][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:00:21,975][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:00:22,473][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:00:22,969][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:00:23,464][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:00:23,960][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:00:24,456][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:00:24,953][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:00:25,451][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:00:25,948][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:00:26,445][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:00:26,942][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:00:27,438][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:00:27,936][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:00:28,433][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:00:28,936][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:00:29,436][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:00:29,936][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:00:30,436][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:00:30,932][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:00:31,431][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:00:31,930][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:00:32,431][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:00:32,932][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:00:33,431][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:00:33,931][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:00:34,431][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:00:34,932][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:00:35,432][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:00:35,932][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:00:36,430][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:00:36,930][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:00:37,427][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:00:37,923][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:00:38,420][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:00:38,917][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:00:39,413][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:00:39,910][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:00:40,412][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:00:40,910][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:00:41,407][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:00:41,904][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:00:42,402][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:00:42,903][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:00:43,401][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:00:43,897][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10381 tokens. +[2026-03-25 22:00:44,567][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 60.54%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:32 +[2026-03-25 22:00:45,286][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:00:45,288][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:00:45,290][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:00:45,961][__main__][INFO] - Iteration 63 took 50s (30.29% Gen, 68.39% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 21m 48s. Estimated total time: 42h 16m 53s. Time estimates for 10 more iterations: 8m 27s, 100 more iterations: 1h 24m 33s, 500 more iterations: 7h 2m 48s. +[2026-03-25 22:00:45,964][__main__][INFO] - Starting iteration 63. +[2026-03-25 22:00:46,364][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 6 and human policies 1. +[2026-03-25 22:00:46,364][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:00:50,056][mllm.models.large_language_model_local][WARNING] - Response Proposal: 5 hats, 5 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:01:03,038][__main__][INFO] - Number of regex retries in iteration 63: 1 +[2026-03-25 22:01:03,039][__main__][INFO] - agents played in iteration 63 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:01:03,820][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:01:03,843][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:01:03,867][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:01:03,889][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:01:03,890][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:01:03,891][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:01:04,510][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:01:04,963][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:01:05,471][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:01:05,970][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:01:06,470][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:01:06,968][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:01:07,470][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:01:07,967][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:01:08,464][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:01:08,968][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:01:09,463][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:01:09,961][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:01:10,457][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:01:10,984][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:01:11,483][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:01:11,980][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:01:12,480][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:01:12,977][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:01:13,476][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:01:13,974][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:01:14,473][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:01:14,971][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:01:15,470][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:01:15,969][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:01:16,464][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:01:16,963][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:01:17,459][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:01:17,956][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:01:18,455][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:01:18,956][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:01:19,455][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:01:19,954][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:01:20,454][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:01:20,952][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:01:21,448][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:01:21,945][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:01:22,440][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:01:22,937][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:01:23,434][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:01:23,931][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:01:24,427][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:01:24,922][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:01:25,416][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:01:25,914][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:01:26,411][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:01:26,907][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:01:27,402][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:01:27,897][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:01:28,393][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:01:28,893][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:01:29,392][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:01:29,891][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:01:30,390][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:01:30,889][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:01:31,392][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:01:31,905][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:01:32,406][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:01:32,905][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:01:33,403][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:01:33,901][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:01:34,400][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:01:34,899][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:01:35,399][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:01:35,899][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:01:36,397][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10205 tokens. +[2026-03-25 22:01:37,077][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.57%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 +[2026-03-25 22:01:37,804][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:01:37,806][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:01:37,808][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:01:38,513][__main__][INFO] - Iteration 64 took 52s (31.97% Gen, 66.67% Train). Generation: 16s, Training: 34s. Estimated remaining time: 42h 31m 31s. Estimated total time: 43h 27m 29s. Time estimates for 10 more iterations: 8m 41s, 100 more iterations: 1h 26m 54s, 500 more iterations: 7h 14m 34s. +[2026-03-25 22:01:38,518][__main__][INFO] - Starting iteration 64. +[2026-03-25 22:01:38,918][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 6 and human policies 1. +[2026-03-25 22:01:38,919][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:01:44,315][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:01:46,314][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:01:54,190][__main__][INFO] - Number of regex retries in iteration 64: 2 +[2026-03-25 22:01:54,192][__main__][INFO] - agents played in iteration 64 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:01:54,949][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:01:54,973][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:01:54,996][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:01:55,019][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:01:55,020][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:01:55,020][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:01:55,688][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:01:56,141][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:01:56,645][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:01:57,143][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:01:57,641][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:01:58,139][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:01:58,642][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:01:59,142][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:01:59,638][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:02:00,140][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:02:00,644][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:02:01,141][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:02:01,639][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:02:02,141][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:02:02,640][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:02:03,137][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:02:03,636][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:02:04,139][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:02:04,638][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:02:05,136][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:02:05,637][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:02:06,134][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:02:06,632][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:02:07,130][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:02:07,628][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:02:08,126][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:02:08,624][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:02:09,121][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:02:09,618][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:02:10,114][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:02:10,611][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:02:11,107][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:02:11,603][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:02:12,100][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:02:12,597][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:02:13,094][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:02:13,590][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:02:14,087][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:02:14,582][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:02:15,080][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:02:15,576][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:02:16,073][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:02:16,568][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:02:17,066][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:02:17,562][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:02:18,058][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:02:18,555][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:02:19,049][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:02:19,545][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:02:20,046][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:02:20,542][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:02:21,040][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:02:21,536][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:02:22,037][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:02:22,536][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:02:23,035][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:02:23,534][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:02:24,032][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:02:24,531][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:02:25,029][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:02:25,526][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:02:26,023][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:02:26,520][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:02:27,018][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:02:27,516][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10158 tokens. +[2026-03-25 22:02:28,226][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 60.53%, Block Peak % of device VRAM: 62.04%, ΔTime: 00:00:32 +[2026-03-25 22:02:28,957][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:02:28,960][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:02:28,961][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:02:29,659][__main__][INFO] - Iteration 65 took 50s (30.10% Gen, 68.52% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 20m 15s. Estimated total time: 42h 17m 4s. Time estimates for 10 more iterations: 8m 27s, 100 more iterations: 1h 24m 34s, 500 more iterations: 7h 2m 50s. +[2026-03-25 22:02:29,662][__main__][INFO] - Starting iteration 65. +[2026-03-25 22:02:30,062][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 6 and human policies 1. +[2026-03-25 22:02:30,062][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:02:35,257][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 1, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:02:44,412][__main__][INFO] - Number of regex retries in iteration 65: 1 +[2026-03-25 22:02:44,413][__main__][INFO] - agents played in iteration 65 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:02:45,167][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:02:45,190][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:02:45,214][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:02:45,236][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:02:45,237][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:02:45,238][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:02:45,895][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:02:46,351][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:02:46,852][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:02:47,348][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:02:47,846][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:02:48,347][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:02:48,843][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:02:49,340][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:02:49,838][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:02:50,336][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:02:50,834][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:02:51,334][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:02:51,833][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:02:52,331][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:02:52,828][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:02:53,324][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:02:53,819][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:02:54,315][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:02:54,813][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:02:55,312][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:02:55,811][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:02:56,318][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:02:56,817][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:02:57,314][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:02:57,813][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:02:58,333][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:02:58,835][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:02:59,335][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:02:59,833][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:03:00,334][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:03:00,836][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:03:01,337][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:03:01,837][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:03:02,336][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:03:02,834][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:03:03,332][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:03:03,830][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:03:04,331][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:03:04,830][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:03:05,327][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:03:05,824][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:03:06,323][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:03:06,823][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:03:07,321][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:03:07,819][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:03:08,316][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:03:08,813][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:03:09,311][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:03:09,813][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:03:10,310][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:03:10,810][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:03:11,305][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:03:11,804][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:03:12,299][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:03:12,795][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:03:13,293][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:03:13,790][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:03:14,288][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:03:14,788][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:03:15,288][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:03:15,787][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:03:16,285][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:03:16,780][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:03:17,278][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:03:17,777][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10219 tokens. +[2026-03-25 22:03:18,460][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 60.62%, Block Peak % of device VRAM: 62.23%, ΔTime: 00:00:32 +[2026-03-25 22:03:19,188][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:03:19,190][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:03:19,192][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:03:19,852][__main__][INFO] - Iteration 66 took 49s (28.82% Gen, 69.85% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 31m 54s. Estimated total time: 41h 29m 33s. Time estimates for 10 more iterations: 8m 17s, 100 more iterations: 1h 22m 59s, 500 more iterations: 6h 54m 55s. +[2026-03-25 22:03:19,855][__main__][INFO] - Starting iteration 66. +[2026-03-25 22:03:20,254][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 6 and human policies 1. +[2026-03-25 22:03:20,255][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:03:22,180][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:03:35,571][__main__][INFO] - Number of regex retries in iteration 66: 1 +[2026-03-25 22:03:35,572][__main__][INFO] - agents played in iteration 66 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:03:36,332][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:03:36,356][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:03:36,379][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:03:36,402][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:03:36,403][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:03:36,403][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:03:37,018][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:03:37,472][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:03:37,977][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:03:38,477][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:03:38,975][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:03:39,475][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:03:39,971][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:03:40,472][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:03:40,973][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:03:41,471][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:03:41,971][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:03:42,466][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:03:42,966][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:03:43,463][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:03:43,959][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:03:44,456][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:03:44,952][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:03:45,449][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:03:45,949][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:03:46,449][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:03:46,947][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:03:47,444][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:03:47,940][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:03:48,439][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:03:48,939][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:03:49,435][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:03:49,934][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:03:50,431][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:03:50,927][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:03:51,424][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:03:51,923][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:03:52,418][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:03:52,915][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:03:53,416][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:03:53,915][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:03:54,411][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:03:54,911][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:03:55,411][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:03:55,908][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:03:56,406][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:03:56,908][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:03:57,402][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:03:57,901][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:03:58,396][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:03:58,894][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:03:59,391][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:03:59,887][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:04:00,385][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:04:00,879][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:04:01,376][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:04:01,874][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:04:02,373][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:04:02,873][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:04:03,369][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:04:03,866][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:04:04,363][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:04:04,859][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:04:05,356][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:04:05,850][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:04:06,345][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:04:06,840][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:04:07,336][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:04:07,833][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:04:08,328][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:04:08,824][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10262 tokens. +[2026-03-25 22:04:09,510][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.02%, Current % of VRAM taken: 60.45%, Block Peak % of device VRAM: 62.02%, ΔTime: 00:00:32 +[2026-03-25 22:04:10,239][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:04:10,241][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:04:10,243][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:04:10,934][__main__][INFO] - Iteration 67 took 50s (30.22% Gen, 68.41% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 15m 31s. Estimated total time: 42h 14m 1s. Time estimates for 10 more iterations: 8m 26s, 100 more iterations: 1h 24m 28s, 500 more iterations: 7h 2m 20s. +[2026-03-25 22:04:10,936][__main__][INFO] - Starting iteration 67. +[2026-03-25 22:04:11,337][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 6 and human policies 1. +[2026-03-25 22:04:11,338][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:04:26,738][__main__][INFO] - Number of regex retries in iteration 67: 0 +[2026-03-25 22:04:26,740][__main__][INFO] - agents played in iteration 67 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:04:27,528][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:04:27,552][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:04:27,576][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:04:27,599][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:04:27,600][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:04:27,601][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:04:28,257][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:04:28,739][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:04:29,248][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:04:29,749][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:04:30,256][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:04:30,756][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:04:31,257][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:04:31,758][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:04:32,261][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:04:32,756][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:04:33,262][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:04:33,761][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:04:34,259][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:04:34,757][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:04:35,258][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:04:35,758][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:04:36,260][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:04:36,765][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:04:37,265][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:04:37,764][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:04:38,263][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:04:38,762][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:04:39,262][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:04:39,762][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:04:40,263][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:04:40,761][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:04:41,260][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:04:41,760][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:04:42,259][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:04:42,757][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:04:43,255][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:04:43,752][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:04:44,253][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:04:44,749][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:04:45,246][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:04:45,742][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:04:46,241][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:04:46,739][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:04:47,236][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:04:47,732][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:04:48,227][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:04:48,736][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:04:49,237][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:04:49,737][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:04:50,239][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:04:50,738][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:04:51,239][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:04:51,739][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:04:52,241][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:04:52,740][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:04:53,238][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:04:53,736][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:04:54,235][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:04:54,733][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:04:55,231][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:04:55,730][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:04:56,227][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:04:56,724][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:04:57,224][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:04:57,724][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:04:58,220][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:04:58,720][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:04:59,219][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:04:59,718][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:05:00,217][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10387 tokens. +[2026-03-25 22:05:00,905][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.25%, Current % of VRAM taken: 60.68%, Block Peak % of device VRAM: 62.21%, ΔTime: 00:00:32 +[2026-03-25 22:05:01,629][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:05:01,632][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:05:01,633][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:05:02,247][__main__][INFO] - Iteration 68 took 50s (30.25% Gen, 68.54% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 26m 9s. Estimated total time: 42h 25m 31s. Time estimates for 10 more iterations: 8m 29s, 100 more iterations: 1h 24m 51s, 500 more iterations: 7h 4m 15s. +[2026-03-25 22:05:02,250][__main__][INFO] - Starting iteration 68. +[2026-03-25 22:05:02,651][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 6 and human policies 1. +[2026-03-25 22:05:02,651][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:05:20,218][__main__][INFO] - Number of regex retries in iteration 68: 0 +[2026-03-25 22:05:20,220][__main__][INFO] - agents played in iteration 68 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:05:20,998][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:05:21,021][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:05:21,044][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:05:21,067][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:05:21,068][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:05:21,068][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:05:21,714][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:05:22,171][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:05:22,677][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:05:23,176][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:05:23,677][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:05:24,177][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:05:24,676][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:05:25,174][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:05:25,669][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:05:26,167][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:05:26,668][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:05:27,168][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:05:27,667][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:05:28,165][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:05:28,665][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:05:29,163][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:05:29,663][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:05:30,157][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:05:30,655][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:05:31,153][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:05:31,648][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:05:32,146][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:05:32,641][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:05:33,139][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:05:33,635][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:05:34,130][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:05:34,630][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:05:35,125][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:05:35,620][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:05:36,117][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:05:36,613][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:05:37,108][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:05:37,602][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:05:38,101][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:05:38,597][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:05:39,093][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:05:39,592][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:05:40,090][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:05:40,590][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:05:41,087][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:05:41,587][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:05:42,083][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:05:42,582][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:05:43,080][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:05:43,576][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:05:44,072][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:05:44,572][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:05:45,071][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:05:45,568][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:05:46,065][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:05:46,562][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:05:47,058][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:05:47,555][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:05:48,051][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:05:48,548][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:05:49,046][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:05:49,542][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:05:50,045][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:05:50,541][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:05:51,038][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:05:51,536][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:05:52,031][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:05:52,526][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:05:53,024][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:05:53,520][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10281 tokens. +[2026-03-25 22:05:54,212][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 60.54%, Block Peak % of device VRAM: 62.07%, ΔTime: 00:00:32 +[2026-03-25 22:05:54,965][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:05:54,968][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:05:54,969][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:05:55,684][__main__][INFO] - Iteration 69 took 53s (33.13% Gen, 65.52% Train). Generation: 17s, Training: 34s. Estimated remaining time: 43h 11m 27s. Estimated total time: 44h 11m 42s. Time estimates for 10 more iterations: 8m 50s, 100 more iterations: 1h 28m 23s, 500 more iterations: 7h 21m 57s. +[2026-03-25 22:05:55,690][__main__][INFO] - Starting iteration 69. +[2026-03-25 22:05:56,091][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 6 and human policies 1. +[2026-03-25 22:05:56,092][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:05:59,527][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:06:12,362][__main__][INFO] - Number of regex retries in iteration 69: 1 +[2026-03-25 22:06:12,363][__main__][INFO] - agents played in iteration 69 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:06:13,126][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:06:13,149][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:06:13,173][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:06:13,196][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:06:13,196][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:06:13,197][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:06:13,856][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:06:14,313][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:06:14,815][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:06:15,321][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:06:15,820][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:06:16,320][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:06:16,821][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:06:17,319][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:06:17,821][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:06:18,321][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:06:18,819][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:06:19,322][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:06:19,823][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:06:20,329][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:06:20,827][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:06:21,324][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:06:21,821][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:06:22,321][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:06:22,824][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:06:23,323][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:06:23,832][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:06:24,330][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:06:24,832][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:06:25,333][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:06:25,834][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:06:26,330][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:06:26,826][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:06:27,319][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:06:27,816][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:06:28,311][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:06:28,807][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:06:29,301][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:06:29,796][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:06:30,296][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:06:30,796][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:06:31,295][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:06:31,797][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:06:32,298][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:06:32,795][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:06:33,302][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:06:33,803][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:06:34,299][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:06:34,796][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:06:35,292][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:06:35,787][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:06:36,283][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:06:36,780][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:06:37,276][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:06:37,771][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:06:38,268][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:06:38,766][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:06:39,264][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:06:39,767][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:06:40,264][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:06:40,761][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:06:41,257][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:06:41,758][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:06:42,257][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:06:42,755][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:06:43,254][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:06:43,755][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:06:44,254][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:06:44,753][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:06:45,250][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:06:45,754][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10280 tokens. +[2026-03-25 22:06:46,447][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.13%, ΔTime: 00:00:32 +[2026-03-25 22:06:47,170][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:06:47,172][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:06:47,174][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:06:47,838][__main__][INFO] - Iteration 70 took 51s (31.44% Gen, 67.27% Train). Generation: 16s, Training: 34s. Estimated remaining time: 42h 6m 15s. Estimated total time: 43h 7m 23s. Time estimates for 10 more iterations: 8m 37s, 100 more iterations: 1h 26m 14s, 500 more iterations: 7h 11m 13s. +[2026-03-25 22:06:47,841][__main__][INFO] - Starting iteration 70. +[2026-03-25 22:06:48,241][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 6 and human policies 1. +[2026-03-25 22:06:48,242][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:07:03,279][__main__][INFO] - Number of regex retries in iteration 70: 0 +[2026-03-25 22:07:03,280][__main__][INFO] - agents played in iteration 70 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:07:04,061][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:07:04,085][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:07:04,108][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:07:04,131][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:07:04,132][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:07:04,133][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:07:04,787][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:07:05,245][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:07:05,750][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:07:06,250][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:07:06,747][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:07:07,243][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:07:07,740][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:07:08,239][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:07:08,736][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:07:09,233][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:07:09,729][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:07:10,234][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:07:10,731][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:07:11,233][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:07:11,732][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:07:12,231][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:07:12,733][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:07:13,233][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:07:13,731][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:07:14,233][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:07:14,733][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:07:15,232][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:07:15,730][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:07:16,227][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:07:16,724][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:07:17,223][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:07:17,719][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:07:18,219][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:07:18,718][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:07:19,218][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:07:19,734][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:07:20,231][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:07:20,736][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:07:21,235][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:07:21,735][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:07:22,231][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:07:22,728][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:07:23,222][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:07:23,717][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:07:24,213][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:07:24,712][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:07:25,208][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:07:25,704][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:07:26,199][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:07:26,695][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:07:27,190][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:07:27,686][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:07:28,182][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:07:28,679][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:07:29,174][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:07:29,671][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:07:30,167][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:07:30,663][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:07:31,165][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:07:31,662][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:07:32,159][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:07:32,657][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:07:33,155][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:07:33,653][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:07:34,151][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:07:34,649][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:07:35,146][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:07:35,645][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:07:36,141][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:07:36,642][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10222 tokens. +[2026-03-25 22:07:37,321][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.09%, Current % of VRAM taken: 60.51%, Block Peak % of device VRAM: 62.09%, ΔTime: 00:00:32 +[2026-03-25 22:07:38,042][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:07:38,044][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:07:38,046][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:07:39,435][__main__][INFO] - Iteration 71 took 51s (29.37% Gen, 67.91% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 37m 44s. Estimated total time: 42h 39m 43s. Time estimates for 10 more iterations: 8m 31s, 100 more iterations: 1h 25m 19s, 500 more iterations: 7h 6m 37s. +[2026-03-25 22:07:39,438][__main__][INFO] - Starting iteration 71. +[2026-03-25 22:07:39,837][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 7 and human policies 1. +[2026-03-25 22:07:39,838][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:07:55,024][__main__][INFO] - Number of regex retries in iteration 71: 0 +[2026-03-25 22:07:55,025][__main__][INFO] - agents played in iteration 71 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:07:55,803][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:07:55,828][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:07:55,852][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:07:55,876][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:07:55,877][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:07:55,877][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:07:56,528][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:07:56,985][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:07:57,487][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:07:57,987][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:07:58,486][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:07:58,983][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:07:59,483][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:07:59,982][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:08:00,483][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:08:00,981][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:08:01,485][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:08:01,986][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:08:02,485][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:08:02,984][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:08:03,484][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:08:03,983][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:08:04,483][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:08:04,983][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:08:05,483][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:08:05,981][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:08:06,479][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:08:06,981][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:08:07,480][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:08:07,979][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:08:08,478][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:08:08,973][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:08:09,468][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:08:09,964][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:08:10,463][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:08:10,962][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:08:11,461][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:08:11,957][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:08:12,457][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:08:12,962][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:08:13,463][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:08:13,964][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:08:14,467][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:08:14,967][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:08:15,468][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:08:15,967][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:08:16,466][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:08:16,962][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:08:17,458][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:08:17,955][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:08:18,451][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:08:18,946][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:08:19,444][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:08:19,942][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:08:20,437][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:08:20,934][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:08:21,433][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:08:21,929][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:08:22,425][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:08:22,921][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:08:23,420][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:08:23,916][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:08:24,416][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:08:24,913][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:08:25,411][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:08:25,913][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:08:26,412][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:08:26,908][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:08:27,408][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:08:27,908][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:08:28,407][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10430 tokens. +[2026-03-25 22:08:29,089][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 60.64%, Block Peak % of device VRAM: 62.34%, ΔTime: 00:00:32 +[2026-03-25 22:08:29,818][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:08:29,820][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:08:29,822][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:08:30,487][__main__][INFO] - Iteration 72 took 50s (29.98% Gen, 68.70% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 9m 41s. Estimated total time: 42h 12m 32s. Time estimates for 10 more iterations: 8m 26s, 100 more iterations: 1h 24m 25s, 500 more iterations: 7h 2m 5s. +[2026-03-25 22:08:30,490][__main__][INFO] - Starting iteration 72. +[2026-03-25 22:08:30,890][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 7 and human policies 1. +[2026-03-25 22:08:30,891][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:08:35,213][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:08:45,864][__main__][INFO] - Number of regex retries in iteration 72: 1 +[2026-03-25 22:08:45,865][__main__][INFO] - agents played in iteration 72 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:08:46,635][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:08:46,658][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:08:46,682][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:08:46,704][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:08:46,705][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:08:46,706][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:08:47,341][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:08:47,792][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:08:48,295][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:08:48,792][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:08:49,287][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:08:49,783][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:08:50,281][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:08:50,780][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:08:51,278][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:08:51,775][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:08:52,272][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:08:52,773][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:08:53,270][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:08:53,766][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:08:54,266][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:08:54,763][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:08:55,261][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:08:55,759][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:08:56,256][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:08:56,753][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:08:57,251][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:08:57,748][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:08:58,246][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:08:58,743][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:08:59,239][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:08:59,740][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:09:00,235][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:09:00,734][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:09:01,233][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:09:01,733][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:09:02,233][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:09:02,729][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:09:03,224][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:09:03,722][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:09:04,217][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:09:04,715][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:09:05,213][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:09:05,712][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:09:06,208][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:09:06,705][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:09:07,201][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:09:07,697][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:09:08,192][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:09:08,686][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:09:09,182][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:09:09,677][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:09:10,174][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:09:10,669][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:09:11,165][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:09:11,662][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:09:12,162][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:09:12,659][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:09:13,158][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:09:13,653][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:09:14,152][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:09:14,648][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:09:15,146][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:09:15,643][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:09:16,140][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:09:16,637][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:09:17,133][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:09:17,627][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:09:18,123][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:09:18,619][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:09:19,116][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10240 tokens. +[2026-03-25 22:09:19,791][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.04%, Current % of VRAM taken: 60.47%, Block Peak % of device VRAM: 61.97%, ΔTime: 00:00:32 +[2026-03-25 22:09:20,516][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:09:20,518][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:09:20,520][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:09:21,244][__main__][INFO] - Iteration 73 took 50s (29.74% Gen, 68.82% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 54m 2s. Estimated total time: 41h 57m 43s. Time estimates for 10 more iterations: 8m 23s, 100 more iterations: 1h 23m 55s, 500 more iterations: 6h 59m 37s. +[2026-03-25 22:09:21,247][__main__][INFO] - Starting iteration 73. +[2026-03-25 22:09:21,646][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 7 and human policies 1. +[2026-03-25 22:09:21,647][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:09:23,658][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:09:28,842][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:09:36,882][__main__][INFO] - Number of regex retries in iteration 73: 2 +[2026-03-25 22:09:36,883][__main__][INFO] - agents played in iteration 73 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:09:37,663][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:09:37,688][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:09:37,712][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:09:37,736][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:09:37,736][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:09:37,737][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:09:38,386][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:09:38,845][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:09:39,348][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:09:39,851][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:09:40,350][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:09:40,846][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:09:41,343][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:09:41,843][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:09:42,340][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:09:42,836][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:09:43,334][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:09:43,829][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:09:44,324][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:09:44,820][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:09:45,315][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:09:45,815][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:09:46,311][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:09:46,810][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:09:47,305][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:09:47,800][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:09:48,296][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:09:48,792][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:09:49,294][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:09:49,791][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:09:50,289][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:09:50,784][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:09:51,280][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:09:51,774][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:09:52,270][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:09:52,766][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:09:53,262][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:09:53,757][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:09:54,251][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:09:54,747][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:09:55,242][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:09:55,739][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:09:56,236][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:09:56,753][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:09:57,248][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:09:57,744][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:09:58,239][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:09:58,737][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:09:59,235][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:09:59,733][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:10:00,230][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:10:00,730][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:10:01,225][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:10:01,723][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:10:02,224][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:10:02,719][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:10:03,221][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:10:03,720][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:10:04,215][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:10:04,714][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:10:05,212][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:10:05,711][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:10:06,210][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:10:06,708][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:10:07,205][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:10:07,700][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:10:08,195][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:10:08,692][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:10:09,191][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:10:09,690][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:10:10,190][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10276 tokens. +[2026-03-25 22:10:10,872][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 60.64%, Block Peak % of device VRAM: 62.13%, ΔTime: 00:00:32 +[2026-03-25 22:10:11,598][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:10:11,600][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:10:11,602][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:10:12,268][__main__][INFO] - Iteration 74 took 50s (30.10% Gen, 68.58% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 6m 35s. Estimated total time: 42h 11m 7s. Time estimates for 10 more iterations: 8m 26s, 100 more iterations: 1h 24m 22s, 500 more iterations: 7h 1m 51s. +[2026-03-25 22:10:12,270][__main__][INFO] - Starting iteration 74. +[2026-03-25 22:10:12,671][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 7 and human policies 1. +[2026-03-25 22:10:12,672][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:10:25,836][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:10:28,204][__main__][INFO] - Number of regex retries in iteration 74: 1 +[2026-03-25 22:10:28,205][__main__][INFO] - agents played in iteration 74 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:10:28,981][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:10:29,004][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:10:29,027][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:10:29,052][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:10:29,052][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:10:29,053][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:10:29,740][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:10:30,198][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:10:30,701][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:10:31,202][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:10:31,707][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:10:32,207][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:10:32,705][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:10:33,205][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:10:33,702][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:10:34,204][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:10:34,702][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:10:35,203][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:10:35,702][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:10:36,201][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:10:36,708][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:10:37,207][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:10:37,704][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:10:38,208][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:10:38,705][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:10:39,201][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:10:39,704][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:10:40,204][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:10:40,702][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:10:41,201][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:10:41,700][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:10:42,199][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:10:42,714][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:10:43,216][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:10:43,718][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:10:44,217][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:10:44,716][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:10:45,215][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:10:45,715][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:10:46,215][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:10:46,712][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:10:47,211][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:10:47,711][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:10:48,210][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:10:48,712][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:10:49,209][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:10:49,705][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:10:50,202][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:10:50,699][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:10:51,198][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:10:51,694][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:10:52,191][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:10:52,687][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:10:53,181][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:10:53,679][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:10:54,182][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:10:54,680][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:10:55,179][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:10:55,676][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:10:56,175][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:10:56,671][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:10:57,173][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:10:57,669][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:10:58,170][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:10:58,672][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:10:59,176][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:10:59,676][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:11:00,176][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:11:00,672][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:11:01,173][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:11:01,671][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10362 tokens. +[2026-03-25 22:11:02,357][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 60.60%, Block Peak % of device VRAM: 62.21%, ΔTime: 00:00:32 +[2026-03-25 22:11:03,082][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:11:03,085][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:11:03,087][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:11:03,755][__main__][INFO] - Iteration 75 took 51s (30.41% Gen, 68.28% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 28m 51s. Estimated total time: 42h 34m 15s. Time estimates for 10 more iterations: 8m 30s, 100 more iterations: 1h 25m 8s, 500 more iterations: 7h 5m 42s. +[2026-03-25 22:11:03,758][__main__][INFO] - Starting iteration 75. +[2026-03-25 22:11:04,157][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 7 and human policies 1. +[2026-03-25 22:11:04,158][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:11:17,210][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:11:28,403][__main__][INFO] - Number of regex retries in iteration 75: 1 +[2026-03-25 22:11:28,404][__main__][INFO] - agents played in iteration 75 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:11:29,181][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:11:29,204][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:11:29,228][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:11:29,251][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:11:29,251][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:11:29,252][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:11:29,904][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:11:30,363][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:11:30,873][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:11:31,377][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:11:31,880][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:11:32,384][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:11:32,883][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:11:33,382][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:11:33,887][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:11:34,384][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:11:34,889][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:11:35,387][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:11:35,886][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:11:36,384][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:11:36,884][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:11:37,380][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:11:37,880][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:11:38,378][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:11:38,876][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:11:39,376][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:11:39,877][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:11:40,376][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:11:40,873][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:11:41,369][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:11:41,869][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:11:42,364][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:11:42,861][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:11:43,360][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:11:43,858][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:11:44,355][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:11:44,852][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:11:45,350][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:11:45,845][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:11:46,344][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:11:46,841][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:11:47,338][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:11:47,838][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:11:48,332][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:11:48,831][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:11:49,330][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:11:49,829][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:11:50,328][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:11:50,828][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:11:51,323][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:11:51,823][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:11:52,322][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:11:52,819][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:11:53,319][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:11:53,818][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:11:54,314][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:11:54,814][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:11:55,309][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:11:55,807][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:11:56,306][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:11:56,801][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:11:57,299][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:11:57,796][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:11:58,293][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:11:58,792][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:11:59,291][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:11:59,787][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:12:00,283][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:12:00,782][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:12:01,280][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:12:01,780][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10511 tokens. +[2026-03-25 22:12:02,458][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 60.64%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:32 +[2026-03-25 22:12:03,199][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:12:03,201][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:12:03,203][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:12:03,876][__main__][INFO] - Iteration 76 took 59s (40.60% Gen, 58.27% Train). Generation: 24s, Training: 34s. Estimated remaining time: 48h 39m 34s. Estimated total time: 49h 45m 57s. Time estimates for 10 more iterations: 9m 57s, 100 more iterations: 1h 39m 31s, 500 more iterations: 8h 17m 39s. +[2026-03-25 22:12:03,878][__main__][INFO] - Starting iteration 76. +[2026-03-25 22:12:04,276][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 7 and human policies 1. +[2026-03-25 22:12:04,276][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:12:19,371][__main__][INFO] - Number of regex retries in iteration 76: 0 +[2026-03-25 22:12:19,372][__main__][INFO] - agents played in iteration 76 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:12:20,228][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:12:20,255][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:12:20,280][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:12:20,304][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:12:20,305][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:12:20,305][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:12:20,961][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:12:21,416][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:12:21,921][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:12:22,420][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:12:22,925][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:12:23,425][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:12:23,923][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:12:24,420][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:12:24,920][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:12:25,418][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:12:25,927][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:12:26,426][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:12:26,926][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:12:27,423][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:12:27,941][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:12:28,440][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:12:28,942][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:12:29,446][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:12:29,948][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:12:30,448][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:12:30,950][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:12:31,447][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:12:31,949][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:12:32,449][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:12:32,950][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:12:33,458][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:12:33,959][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:12:34,460][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:12:34,957][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:12:35,456][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:12:35,957][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:12:36,458][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:12:36,956][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:12:37,453][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:12:37,952][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:12:38,453][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:12:38,949][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:12:39,444][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:12:39,942][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:12:40,437][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:12:40,934][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:12:41,430][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:12:41,926][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:12:42,423][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:12:42,920][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:12:43,416][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:12:43,914][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:12:44,416][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:12:44,915][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:12:45,416][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:12:45,917][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:12:46,414][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:12:46,914][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:12:47,415][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:12:47,916][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:12:48,416][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:12:48,915][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:12:49,415][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:12:49,911][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:12:50,407][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:12:50,902][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:12:51,397][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:12:51,892][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:12:52,387][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:12:52,882][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10307 tokens. +[2026-03-25 22:12:53,566][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 60.50%, Block Peak % of device VRAM: 62.17%, ΔTime: 00:00:32 +[2026-03-25 22:12:54,285][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:12:54,288][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:12:54,289][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:12:55,024][__main__][INFO] - Iteration 77 took 50s (29.75% Gen, 68.80% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 10m 11s. Estimated total time: 42h 17m 26s. Time estimates for 10 more iterations: 8m 27s, 100 more iterations: 1h 24m 34s, 500 more iterations: 7h 2m 54s. +[2026-03-25 22:12:55,026][__main__][INFO] - Starting iteration 77. +[2026-03-25 22:12:55,429][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 7 and human policies 1. +[2026-03-25 22:12:55,430][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:12:57,530][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:12:57,587][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:13:05,327][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 balls, 10 books, 10 hats did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:13:10,866][__main__][INFO] - Number of regex retries in iteration 77: 3 +[2026-03-25 22:13:10,867][__main__][INFO] - agents played in iteration 77 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:13:11,662][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:13:11,686][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:13:11,709][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:13:11,732][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:13:11,733][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:13:11,733][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:13:12,398][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:13:12,857][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:13:13,362][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:13:13,867][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:13:14,374][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:13:14,872][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:13:15,371][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:13:15,870][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:13:16,367][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:13:16,867][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:13:17,371][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:13:17,870][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:13:18,370][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:13:18,867][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:13:19,366][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:13:19,868][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:13:20,368][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:13:20,864][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:13:21,360][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:13:21,858][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:13:22,354][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:13:22,851][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:13:23,347][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:13:23,843][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:13:24,340][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:13:24,835][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:13:25,332][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:13:25,833][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:13:26,330][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:13:26,828][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:13:27,332][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:13:27,828][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:13:28,324][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:13:28,824][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:13:29,323][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:13:29,827][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:13:30,328][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:13:30,827][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:13:31,326][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:13:31,823][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:13:32,320][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:13:32,823][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:13:33,323][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:13:33,821][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:13:34,321][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:13:34,819][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:13:35,322][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:13:35,820][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:13:36,324][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:13:36,831][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:13:37,332][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:13:37,834][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:13:38,334][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:13:38,835][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:13:39,342][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:13:39,843][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:13:40,342][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:13:40,839][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:13:41,345][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:13:41,843][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:13:42,341][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:13:42,838][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:13:43,335][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:13:43,832][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:13:44,329][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10327 tokens. +[2026-03-25 22:13:45,053][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.09%, Current % of VRAM taken: 60.52%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 +[2026-03-25 22:13:45,781][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:13:45,784][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:13:45,785][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:13:46,474][__main__][INFO] - Iteration 78 took 51s (30.24% Gen, 68.41% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 24m 9s. Estimated total time: 42h 32m 15s. Time estimates for 10 more iterations: 8m 30s, 100 more iterations: 1h 25m 4s, 500 more iterations: 7h 5m 22s. +[2026-03-25 22:13:46,477][__main__][INFO] - Starting iteration 78. +[2026-03-25 22:13:46,882][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 7 and human policies 1. +[2026-03-25 22:13:46,883][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:13:50,232][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:13:54,827][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:14:02,790][__main__][INFO] - Number of regex retries in iteration 78: 2 +[2026-03-25 22:14:02,791][__main__][INFO] - agents played in iteration 78 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:14:03,569][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:14:03,593][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:14:03,616][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:14:03,639][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:14:03,640][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:14:03,640][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:14:04,292][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:14:04,754][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:14:05,259][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:14:05,759][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:14:06,259][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:14:06,761][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:14:07,260][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:14:07,759][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:14:08,257][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:14:08,754][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:14:09,249][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:14:09,745][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:14:10,240][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:14:10,738][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:14:11,235][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:14:11,734][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:14:12,229][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:14:12,723][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:14:13,221][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:14:13,718][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:14:14,217][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:14:14,713][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:14:15,213][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:14:15,709][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:14:16,205][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:14:16,705][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:14:17,203][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:14:17,701][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:14:18,203][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:14:18,701][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:14:19,200][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:14:19,701][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:14:20,200][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:14:20,697][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:14:21,195][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:14:21,692][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:14:22,188][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:14:22,684][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:14:23,180][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:14:23,678][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:14:24,174][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:14:24,670][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:14:25,169][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:14:25,665][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:14:26,162][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:14:26,662][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:14:27,159][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:14:27,657][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:14:28,157][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:14:28,656][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:14:29,154][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:14:29,654][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:14:30,150][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:14:30,650][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:14:31,146][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:14:31,645][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:14:32,144][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:14:32,641][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:14:33,139][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:14:33,637][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:14:34,136][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:14:34,634][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:14:35,134][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:14:35,632][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:14:36,131][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10437 tokens. +[2026-03-25 22:14:36,811][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.21%, ΔTime: 00:00:32 +[2026-03-25 22:14:37,543][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:14:37,546][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:14:37,547][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:14:38,254][__main__][INFO] - Iteration 79 took 51s (30.97% Gen, 67.66% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 39m 38s. Estimated total time: 42h 48m 36s. Time estimates for 10 more iterations: 8m 33s, 100 more iterations: 1h 25m 37s, 500 more iterations: 7h 8m 6s. +[2026-03-25 22:14:38,256][__main__][INFO] - Starting iteration 79. +[2026-03-25 22:14:38,659][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 7 and human policies 1. +[2026-03-25 22:14:38,659][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:14:53,637][__main__][INFO] - Number of regex retries in iteration 79: 0 +[2026-03-25 22:14:53,638][__main__][INFO] - agents played in iteration 79 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:14:54,407][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:14:54,432][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:14:54,456][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:14:54,479][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:14:54,480][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:14:54,481][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:14:55,129][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:14:55,582][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:14:56,088][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:14:56,587][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:14:57,087][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:14:57,586][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:14:58,089][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:14:58,589][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:14:59,087][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:14:59,591][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:15:00,089][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:15:00,585][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:15:01,083][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:15:01,583][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:15:02,083][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:15:02,582][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:15:03,081][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:15:03,580][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:15:04,080][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:15:04,579][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:15:05,076][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:15:05,576][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:15:06,075][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:15:06,577][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:15:07,077][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:15:07,577][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:15:08,080][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:15:08,604][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:15:09,104][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:15:09,600][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:15:10,100][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:15:10,600][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:15:11,095][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:15:11,595][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:15:12,092][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:15:12,592][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:15:13,095][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:15:13,593][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:15:14,092][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:15:14,593][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:15:15,093][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:15:15,591][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:15:16,087][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:15:16,586][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:15:17,081][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:15:17,576][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:15:18,073][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:15:18,570][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:15:19,067][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:15:19,566][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:15:20,066][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:15:20,568][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:15:21,066][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:15:21,568][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:15:22,067][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:15:22,566][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:15:23,061][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:15:23,559][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:15:24,060][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:15:24,560][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:15:25,055][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:15:25,551][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:15:26,048][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:15:26,545][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:15:27,041][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10309 tokens. +[2026-03-25 22:15:27,722][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 60.52%, Block Peak % of device VRAM: 62.27%, ΔTime: 00:00:32 +[2026-03-25 22:15:28,463][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:15:28,466][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:15:28,467][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:15:29,149][__main__][INFO] - Iteration 80 took 50s (29.67% Gen, 68.98% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 54m 44s. Estimated total time: 42h 4m 33s. Time estimates for 10 more iterations: 8m 24s, 100 more iterations: 1h 24m 9s, 500 more iterations: 7h 0m 45s. +[2026-03-25 22:15:29,152][__main__][INFO] - Starting iteration 80. +[2026-03-25 22:15:29,560][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 7 and human policies 1. +[2026-03-25 22:15:29,561][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:15:42,274][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:15:44,657][__main__][INFO] - Number of regex retries in iteration 80: 1 +[2026-03-25 22:15:44,658][__main__][INFO] - agents played in iteration 80 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:15:45,430][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:15:45,454][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:15:45,477][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:15:45,500][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:15:45,501][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:15:45,501][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:15:46,154][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:15:46,611][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:15:47,115][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:15:47,615][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:15:48,114][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:15:48,614][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:15:49,117][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:15:49,615][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:15:50,119][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:15:50,618][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:15:51,116][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:15:51,614][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:15:52,113][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:15:52,609][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:15:53,110][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:15:53,606][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:15:54,106][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:15:54,607][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:15:55,105][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:15:55,601][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:15:56,100][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:15:56,599][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:15:57,098][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:15:57,597][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:15:58,097][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:15:58,593][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:15:59,090][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:15:59,586][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:16:00,082][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:16:00,579][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:16:01,077][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:16:01,575][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:16:02,075][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:16:02,574][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:16:03,070][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:16:03,572][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:16:04,069][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:16:04,566][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:16:05,063][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:16:05,563][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:16:06,064][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:16:06,563][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:16:07,063][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:16:07,562][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:16:08,063][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:16:08,561][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:16:09,060][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:16:09,562][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:16:10,061][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:16:10,588][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:16:11,093][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:16:11,595][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:16:12,095][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:16:12,599][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:16:13,102][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:16:13,602][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:16:14,102][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:16:14,601][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:16:15,098][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:16:15,595][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:16:16,095][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:16:16,593][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:16:17,089][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:16:17,592][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:16:18,093][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10445 tokens. +[2026-03-25 22:16:18,788][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 60.63%, Block Peak % of device VRAM: 62.47%, ΔTime: 00:00:32 +[2026-03-25 22:16:19,508][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:16:19,511][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:16:19,512][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:16:20,752][__main__][INFO] - Iteration 81 took 51s (29.49% Gen, 68.08% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 28m 58s. Estimated total time: 42h 39m 38s. Time estimates for 10 more iterations: 8m 31s, 100 more iterations: 1h 25m 19s, 500 more iterations: 7h 6m 36s. +[2026-03-25 22:16:20,755][__main__][INFO] - Starting iteration 81. +[2026-03-25 22:16:21,155][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 8 and human policies 1. +[2026-03-25 22:16:21,156][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:16:25,936][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:16:29,171][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls -This proposal aims to maximize my expected value by focusing on the items where I have a higher per-item value compared to Bob. Given my values and Bob's values, I perceive hats as the most valuable item for me, followed by books and then balls, where Bob's values are higher. did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:53:15,634][__main__][INFO] - Number of regex retries in iteration 61: 1 -[2025-11-09 18:53:15,634][__main__][INFO] - agents played in iteration 61 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:53:16,414][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:53:16,439][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:53:16,464][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:53:16,486][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:53:16,486][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:53:16,487][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:53:17,090][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:53:17,544][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:53:18,046][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:53:18,545][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:53:19,041][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:53:19,547][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:53:20,047][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:53:20,545][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:53:21,042][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:53:21,535][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:53:22,029][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:53:22,521][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:53:23,014][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:53:23,509][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:53:24,001][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:53:24,494][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:53:24,992][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:53:25,487][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:53:25,983][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:53:26,477][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:53:26,971][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:53:27,468][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:53:27,964][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:53:28,457][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:53:28,960][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:53:29,453][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:53:29,948][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:53:30,446][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:53:30,938][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:53:31,441][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:53:31,933][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:53:32,427][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:53:32,918][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:53:33,415][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:53:33,910][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:53:34,414][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:53:34,911][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:53:35,407][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:53:35,901][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:53:36,395][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:53:36,892][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:53:37,386][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:53:37,880][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:53:38,378][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:53:38,870][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:53:39,366][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:53:39,860][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:53:40,353][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:53:40,861][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:53:41,355][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:53:41,850][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:53:42,349][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:53:42,848][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:53:43,355][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:53:43,856][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:53:44,356][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:53:44,867][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:53:45,370][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:53:45,874][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:53:46,371][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:53:46,870][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:53:47,370][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:53:47,869][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:53:48,368][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:53:48,871][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10328 tokens. -[2025-11-09 18:53:49,570][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 55.65%, Block Peak % of device VRAM: 62.53%, ΔTime: 00:00:32 -[2025-11-09 18:53:50,339][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:53:50,340][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:53:50,342][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:53:51,257][__main__][INFO] - Iteration 62 took 51s (31.37% Gen, 66.86% Train). Generation: 16s, Training: 34s. Estimated remaining time: 42h 20m 36s. Estimated total time: 43h 15m 30s. Time estimates for 10 more iterations: 8m 39s, 100 more iterations: 1h 26m 31s, 500 more iterations: 7h 12m 35s. -[2025-11-09 18:53:51,259][__main__][INFO] - Starting iteration 62. -[2025-11-09 18:53:51,727][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 6 and human policies 1. -[2025-11-09 18:53:51,728][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:54:06,992][__main__][INFO] - Number of regex retries in iteration 62: 0 -[2025-11-09 18:54:06,993][__main__][INFO] - agents played in iteration 62 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:54:07,789][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:54:07,817][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:54:07,843][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:54:07,872][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:54:07,873][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:54:07,875][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:54:08,480][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:54:08,932][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:54:09,448][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:54:09,943][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:54:10,440][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:54:10,937][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:54:11,432][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:54:11,927][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:54:12,422][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:54:12,917][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:54:13,410][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:54:13,902][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:54:14,397][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:54:14,892][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:54:15,385][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:54:15,880][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:54:16,376][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:54:16,872][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:54:17,372][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:54:17,866][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:54:18,362][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:54:18,856][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:54:19,349][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:54:19,853][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:54:20,347][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:54:20,839][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:54:21,350][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:54:21,845][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:54:22,340][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:54:22,831][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:54:23,325][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:54:23,827][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:54:24,320][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:54:24,814][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:54:25,315][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:54:25,812][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:54:26,304][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:54:26,797][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:54:27,291][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:54:27,787][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:54:28,281][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:54:28,776][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:54:29,270][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:54:29,765][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:54:30,267][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:54:30,764][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:54:31,260][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:54:31,757][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:54:32,252][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:54:32,746][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:54:33,241][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:54:33,733][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:54:34,252][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:54:34,751][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:54:35,253][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:54:35,753][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:54:36,252][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:54:36,764][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:54:37,260][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:54:37,757][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:54:38,255][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:54:38,757][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:54:39,256][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:54:39,752][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:54:40,252][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10506 tokens. -[2025-11-09 18:54:40,947][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 55.68%, Block Peak % of device VRAM: 62.23%, ΔTime: 00:00:32 -[2025-11-09 18:54:41,714][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:54:41,716][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:54:41,718][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:54:42,628][__main__][INFO] - Iteration 63 took 50s (29.99% Gen, 68.22% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 29m 19s. Estimated total time: 42h 25m 4s. Time estimates for 10 more iterations: 8m 29s, 100 more iterations: 1h 24m 50s, 500 more iterations: 7h 4m 10s. -[2025-11-09 18:54:42,631][__main__][INFO] - Starting iteration 63. -[2025-11-09 18:54:43,114][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 6 and human policies 1. -[2025-11-09 18:54:43,116][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:54:57,974][__main__][INFO] - Number of regex retries in iteration 63: 0 -[2025-11-09 18:54:57,975][__main__][INFO] - agents played in iteration 63 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:54:58,821][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:54:58,847][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:54:58,873][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:54:58,896][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:54:58,896][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:54:58,897][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:54:59,542][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:54:59,991][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:55:00,496][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:55:00,990][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:55:01,484][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:55:01,976][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:55:02,473][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:55:02,976][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:55:03,471][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:55:03,972][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:55:04,472][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:55:04,968][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:55:05,460][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:55:05,956][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:55:06,477][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:55:06,981][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:55:07,479][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:55:07,974][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:55:08,469][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:55:08,966][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:55:09,471][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:55:09,968][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:55:10,463][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:55:10,961][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:55:11,455][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:55:11,950][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:55:12,443][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:55:12,939][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:55:13,448][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:55:13,944][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:55:14,440][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:55:14,950][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:55:15,447][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:55:15,941][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:55:16,433][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:55:16,927][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:55:17,424][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:55:17,920][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:55:18,411][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:55:18,909][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:55:19,401][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:55:19,893][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:55:20,388][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:55:20,879][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:55:21,387][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:55:21,877][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:55:22,371][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:55:22,867][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:55:23,358][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:55:23,867][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:55:24,371][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:55:24,875][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:55:25,387][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:55:25,888][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:55:26,393][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:55:26,891][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:55:27,393][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:55:27,897][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:55:28,395][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:55:28,896][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:55:29,394][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:55:29,893][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:55:30,393][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:55:30,889][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:55:31,385][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10365 tokens. -[2025-11-09 18:55:32,091][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 55.67%, Block Peak % of device VRAM: 62.48%, ΔTime: 00:00:32 -[2025-11-09 18:55:32,870][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:55:32,871][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:55:32,873][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:55:33,823][__main__][INFO] - Iteration 64 took 50s (29.30% Gen, 68.82% Train). Generation: 14s, Training: 34s. Estimated remaining time: 41h 18m 50s. Estimated total time: 42h 15m 27s. Time estimates for 10 more iterations: 8m 27s, 100 more iterations: 1h 24m 30s, 500 more iterations: 7h 2m 34s. -[2025-11-09 18:55:33,825][__main__][INFO] - Starting iteration 64. -[2025-11-09 18:55:34,301][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 6 and human policies 1. -[2025-11-09 18:55:34,302][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:55:50,108][__main__][INFO] - Number of regex retries in iteration 64: 0 -[2025-11-09 18:55:50,108][__main__][INFO] - agents played in iteration 64 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:55:50,901][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:55:50,927][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:55:50,956][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:55:50,980][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:55:50,980][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:55:50,981][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:55:51,625][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:55:52,079][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:55:52,579][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:55:53,076][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:55:53,570][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:55:54,074][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:55:54,572][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:55:55,070][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:55:55,564][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:55:56,061][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:55:56,553][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:55:57,065][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:55:57,560][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:55:58,056][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:55:58,551][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:55:59,044][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:55:59,540][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:56:00,036][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:56:00,531][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:56:01,028][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:56:01,524][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:56:02,019][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:56:02,514][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:56:03,008][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:56:03,508][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:56:04,003][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:56:04,498][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:56:04,999][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:56:05,492][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:56:05,985][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:56:06,481][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:56:06,973][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:56:07,480][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:56:07,974][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:56:08,468][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:56:08,977][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:56:09,471][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:56:09,968][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:56:10,458][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:56:10,952][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:56:11,446][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:56:11,937][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:56:12,430][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:56:12,925][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:56:13,416][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:56:13,908][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:56:14,400][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:56:14,892][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:56:15,400][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:56:15,894][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:56:16,389][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:56:16,896][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:56:17,392][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:56:17,895][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:56:18,393][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:56:18,893][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:56:19,405][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:56:19,904][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:56:20,404][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:56:20,899][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:56:21,397][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:56:21,903][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:56:22,398][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:56:22,895][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:56:23,391][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10358 tokens. -[2025-11-09 18:56:24,115][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 55.61%, Block Peak % of device VRAM: 62.10%, ΔTime: 00:00:32 -[2025-11-09 18:56:24,892][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:56:24,893][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:56:24,895][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:56:25,842][__main__][INFO] - Iteration 65 took 51s (30.67% Gen, 67.49% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 59m 33s. Estimated total time: 42h 57m 2s. Time estimates for 10 more iterations: 8m 35s, 100 more iterations: 1h 25m 54s, 500 more iterations: 7h 9m 30s. -[2025-11-09 18:56:25,844][__main__][INFO] - Starting iteration 65. -[2025-11-09 18:56:26,327][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 6 and human policies 1. -[2025-11-09 18:56:26,327][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:56:41,203][__main__][INFO] - Number of regex retries in iteration 65: 0 -[2025-11-09 18:56:41,203][__main__][INFO] - agents played in iteration 65 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:56:42,017][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:56:42,045][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:56:42,072][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:56:42,095][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:56:42,096][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:56:42,097][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:56:42,767][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:56:43,224][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:56:43,728][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:56:44,229][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:56:44,727][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:56:45,221][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:56:45,719][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:56:46,216][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:56:46,709][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:56:47,203][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:56:47,698][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:56:48,193][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:56:48,688][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:56:49,180][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:56:49,685][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:56:50,178][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:56:50,670][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:56:51,163][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:56:51,659][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:56:52,155][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:56:52,650][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:56:53,147][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:56:53,646][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:56:54,139][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:56:54,633][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:56:55,132][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:56:55,648][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:56:56,151][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:56:56,649][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:56:57,147][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:56:57,652][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:56:58,149][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:56:58,647][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:56:59,144][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:56:59,638][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:57:00,134][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:57:00,630][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:57:01,130][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:57:01,624][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:57:02,117][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:57:02,621][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:57:03,116][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:57:03,610][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:57:04,107][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:57:04,602][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:57:05,106][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:57:05,598][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:57:06,092][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:57:06,591][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:57:07,085][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:57:07,586][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:57:08,083][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:57:08,577][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:57:09,091][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:57:09,590][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:57:10,092][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:57:10,591][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:57:11,095][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:57:11,605][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:57:12,105][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:57:12,604][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:57:13,109][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:57:13,605][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:57:14,104][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:57:14,601][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10480 tokens. -[2025-11-09 18:57:15,236][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 55.72%, Block Peak % of device VRAM: 62.39%, ΔTime: 00:00:32 -[2025-11-09 18:57:15,997][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:57:15,998][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:57:16,000][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:57:16,937][__main__][INFO] - Iteration 66 took 50s (29.39% Gen, 68.75% Train). Generation: 14s, Training: 34s. Estimated remaining time: 41h 12m 11s. Estimated total time: 42h 10m 31s. Time estimates for 10 more iterations: 8m 26s, 100 more iterations: 1h 24m 21s, 500 more iterations: 7h 1m 45s. -[2025-11-09 18:57:16,939][__main__][INFO] - Starting iteration 66. -[2025-11-09 18:57:17,400][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 6 and human policies 1. -[2025-11-09 18:57:17,401][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:57:19,413][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 18:57:32,862][__main__][INFO] - Number of regex retries in iteration 66: 1 -[2025-11-09 18:57:32,863][__main__][INFO] - agents played in iteration 66 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:57:33,690][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:57:33,717][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:57:33,744][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:57:33,767][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:57:33,767][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:57:33,769][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:57:34,411][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:57:34,869][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:57:35,372][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:57:35,869][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:57:36,370][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:57:36,866][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:57:37,362][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:57:37,860][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:57:38,359][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:57:38,857][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:57:39,356][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:57:39,851][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:57:40,349][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:57:40,843][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:57:41,338][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:57:41,833][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:57:42,328][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:57:42,826][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:57:43,320][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:57:43,818][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:57:44,321][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:57:44,819][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:57:45,315][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:57:45,811][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:57:46,309][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:57:46,804][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:57:47,306][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:57:47,804][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:57:48,312][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:57:48,809][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:57:49,303][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:57:49,800][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:57:50,294][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:57:50,794][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:57:51,292][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:57:51,788][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:57:52,296][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:57:52,791][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:57:53,289][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:57:53,796][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:57:54,293][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:57:54,789][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:57:55,286][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:57:55,781][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:57:56,276][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:57:56,770][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:57:57,265][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:57:57,759][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:57:58,253][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:57:58,750][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:57:59,248][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:57:59,745][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:58:00,250][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:58:00,750][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:58:01,250][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:58:01,772][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:58:02,272][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:58:02,769][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:58:03,267][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:58:03,764][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:58:04,271][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:58:04,768][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:58:05,266][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:58:05,766][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:58:06,261][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10355 tokens. -[2025-11-09 18:58:06,970][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 55.57%, Block Peak % of device VRAM: 62.08%, ΔTime: 00:00:32 -[2025-11-09 18:58:07,733][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:58:07,735][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:58:07,737][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:58:08,750][__main__][INFO] - Iteration 67 took 51s (30.11% Gen, 67.91% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 48m 20s. Estimated total time: 42h 47m 32s. Time estimates for 10 more iterations: 8m 33s, 100 more iterations: 1h 25m 35s, 500 more iterations: 7h 7m 55s. -[2025-11-09 18:58:08,753][__main__][INFO] - Starting iteration 67. -[2025-11-09 18:58:09,245][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 6 and human policies 1. -[2025-11-09 18:58:09,245][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:58:24,385][__main__][INFO] - Number of regex retries in iteration 67: 0 -[2025-11-09 18:58:24,386][__main__][INFO] - agents played in iteration 67 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:58:25,259][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:58:25,286][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:58:25,312][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:58:25,334][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:58:25,335][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:58:25,335][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:58:25,990][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:58:26,449][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:58:26,951][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:58:27,449][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:58:27,947][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:58:28,442][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:58:28,937][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:58:29,436][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:58:29,931][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:58:30,436][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:58:30,929][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:58:31,421][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:58:31,916][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:58:32,410][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:58:32,905][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:58:33,413][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:58:33,907][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:58:34,400][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:58:34,894][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:58:35,391][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:58:35,888][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:58:36,383][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:58:36,878][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:58:37,375][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:58:37,871][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:58:38,368][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:58:38,861][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:58:39,356][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:58:39,856][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:58:40,351][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:58:40,845][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:58:41,340][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:58:41,833][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:58:42,339][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:58:42,829][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:58:43,322][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:58:43,820][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:58:44,311][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:58:44,802][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:58:45,292][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:58:45,788][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:58:46,287][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:58:46,788][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:58:47,285][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:58:47,783][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:58:48,282][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:58:48,777][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:58:49,275][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:58:49,772][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:58:50,270][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:58:50,762][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:58:51,255][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:58:51,755][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:58:52,253][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:58:52,756][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:58:53,251][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:58:53,749][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:58:54,258][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:58:54,757][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:58:55,258][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:58:55,754][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:58:56,252][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:58:56,751][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:58:57,247][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:58:57,745][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10512 tokens. -[2025-11-09 18:58:58,450][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.33%, ΔTime: 00:00:32 -[2025-11-09 18:58:59,223][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:58:59,225][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:58:59,226][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:59:00,163][__main__][INFO] - Iteration 68 took 50s (29.73% Gen, 68.42% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 25m 53s. Estimated total time: 42h 25m 56s. Time estimates for 10 more iterations: 8m 29s, 100 more iterations: 1h 24m 51s, 500 more iterations: 7h 4m 19s. -[2025-11-09 18:59:00,165][__main__][INFO] - Starting iteration 68. -[2025-11-09 18:59:00,660][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 6 and human policies 1. -[2025-11-09 18:59:00,661][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 18:59:16,121][__main__][INFO] - Number of regex retries in iteration 68: 0 -[2025-11-09 18:59:16,121][__main__][INFO] - agents played in iteration 68 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 18:59:16,951][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:59:16,976][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:59:17,001][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:59:17,024][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 18:59:17,025][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 18:59:17,025][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 18:59:17,631][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 18:59:18,086][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 18:59:18,590][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 18:59:19,087][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 18:59:19,586][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 18:59:20,088][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 18:59:20,584][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 18:59:21,077][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 18:59:21,573][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 18:59:22,072][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 18:59:22,569][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 18:59:23,065][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 18:59:23,561][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 18:59:24,056][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 18:59:24,562][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 18:59:25,055][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 18:59:25,551][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 18:59:26,045][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 18:59:26,539][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 18:59:27,034][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 18:59:27,529][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 18:59:28,022][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 18:59:28,528][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 18:59:29,023][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 18:59:29,516][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 18:59:30,026][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 18:59:30,525][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 18:59:31,018][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 18:59:31,510][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 18:59:32,006][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 18:59:32,498][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 18:59:32,991][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 18:59:33,484][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 18:59:33,980][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 18:59:34,476][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 18:59:34,972][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 18:59:35,466][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 18:59:35,961][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 18:59:36,462][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 18:59:36,956][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 18:59:37,452][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 18:59:37,944][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 18:59:38,437][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 18:59:38,945][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 18:59:39,437][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 18:59:39,933][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 18:59:40,437][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 18:59:40,929][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 18:59:41,423][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 18:59:41,926][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 18:59:42,418][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 18:59:42,912][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 18:59:43,412][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 18:59:43,911][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 18:59:44,410][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 18:59:44,911][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 18:59:45,409][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 18:59:45,908][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 18:59:46,402][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 18:59:46,913][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 18:59:47,413][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 18:59:47,909][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 18:59:48,407][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 18:59:48,903][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 18:59:49,402][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10586 tokens. -[2025-11-09 18:59:50,104][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 55.71%, Block Peak % of device VRAM: 62.20%, ΔTime: 00:00:32 -[2025-11-09 18:59:50,858][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 18:59:50,860][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 18:59:50,861][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 18:59:51,949][__main__][INFO] - Iteration 69 took 51s (30.14% Gen, 67.73% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 43m 34s. Estimated total time: 42h 44m 29s. Time estimates for 10 more iterations: 8m 32s, 100 more iterations: 1h 25m 28s, 500 more iterations: 7h 7m 24s. -[2025-11-09 18:59:51,952][__main__][INFO] - Starting iteration 69. -[2025-11-09 18:59:52,426][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 6 and human policies 1. -[2025-11-09 18:59:52,427][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:00:08,027][__main__][INFO] - Number of regex retries in iteration 69: 0 -[2025-11-09 19:00:08,028][__main__][INFO] - agents played in iteration 69 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:00:08,852][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:00:08,874][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:00:08,896][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:00:08,918][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:00:08,918][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:00:08,919][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:00:09,521][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:00:09,976][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:00:10,475][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:00:10,973][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:00:11,470][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:00:11,968][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:00:12,470][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:00:12,963][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:00:13,460][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:00:13,958][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:00:14,453][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:00:14,950][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:00:15,444][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:00:15,945][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:00:16,440][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:00:16,963][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:00:17,463][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:00:17,963][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:00:18,463][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:00:18,958][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:00:19,455][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:00:19,950][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:00:20,449][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:00:20,948][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:00:21,445][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:00:21,938][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:00:22,432][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:00:22,925][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:00:23,418][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:00:23,912][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:00:24,404][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:00:24,896][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:00:25,389][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:00:25,885][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:00:26,384][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:00:26,880][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:00:27,376][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:00:27,871][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:00:28,367][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:00:28,866][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:00:29,367][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:00:29,859][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:00:30,367][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:00:30,860][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:00:31,356][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:00:31,851][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:00:32,348][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:00:32,854][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:00:33,349][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:00:33,845][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:00:34,358][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:00:34,854][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:00:35,351][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:00:35,855][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:00:36,352][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:00:36,850][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:00:37,350][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:00:37,852][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:00:38,368][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:00:38,869][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:00:39,365][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:00:39,866][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:00:40,367][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:00:40,873][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:00:41,372][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10575 tokens. -[2025-11-09 19:00:42,065][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 55.72%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 -[2025-11-09 19:00:42,821][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:00:42,822][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:00:42,824][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:00:43,760][__main__][INFO] - Iteration 70 took 51s (30.39% Gen, 67.78% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 44m 56s. Estimated total time: 42h 46m 43s. Time estimates for 10 more iterations: 8m 33s, 100 more iterations: 1h 25m 33s, 500 more iterations: 7h 7m 47s. -[2025-11-09 19:00:43,762][__main__][INFO] - Starting iteration 70. -[2025-11-09 19:00:44,227][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 6 and human policies 1. -[2025-11-09 19:00:44,227][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:00:46,250][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:00:48,658][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:00:59,856][__main__][INFO] - Number of regex retries in iteration 70: 2 -[2025-11-09 19:00:59,857][__main__][INFO] - agents played in iteration 70 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:01:00,662][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:01:00,685][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:01:00,707][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:01:00,728][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:01:00,729][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:01:00,730][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:01:01,322][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:01:01,777][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:01:02,275][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:01:02,771][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:01:03,278][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:01:03,776][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:01:04,273][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:01:04,768][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:01:05,266][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:01:05,765][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:01:06,264][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:01:06,761][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:01:07,269][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:01:07,764][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:01:08,258][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:01:08,769][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:01:09,267][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:01:09,764][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:01:10,258][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:01:10,753][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:01:11,251][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:01:11,746][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:01:12,244][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:01:12,739][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:01:13,237][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:01:13,733][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:01:14,228][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:01:14,726][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:01:15,220][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:01:15,713][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:01:16,209][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:01:16,703][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:01:17,199][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:01:17,700][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:01:18,196][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:01:18,690][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:01:19,185][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:01:19,678][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:01:20,171][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:01:20,667][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:01:21,163][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:01:21,655][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:01:22,150][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:01:22,642][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:01:23,141][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:01:23,634][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:01:24,130][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:01:24,620][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:01:25,113][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:01:25,605][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:01:26,098][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:01:26,594][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:01:27,094][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:01:27,592][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:01:28,092][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:01:28,593][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:01:29,092][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:01:29,594][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:01:30,093][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:01:30,593][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:01:31,103][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:01:31,601][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:01:32,100][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:01:32,602][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:01:33,101][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10389 tokens. -[2025-11-09 19:01:33,826][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 55.60%, Block Peak % of device VRAM: 62.08%, ΔTime: 00:00:32 -[2025-11-09 19:01:34,599][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:01:34,600][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:01:34,602][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:01:36,423][__main__][INFO] - Iteration 71 took 52s (29.94% Gen, 66.57% Train). Generation: 15s, Training: 34s. Estimated remaining time: 42h 27m 10s. Estimated total time: 43h 29m 50s. Time estimates for 10 more iterations: 8m 41s, 100 more iterations: 1h 26m 59s, 500 more iterations: 7h 14m 58s. -[2025-11-09 19:01:36,425][__main__][INFO] - Starting iteration 71. -[2025-11-09 19:01:36,892][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 7 and human policies 1. -[2025-11-09 19:01:36,893][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:01:39,328][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:01:46,392][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:01:53,064][__main__][INFO] - Number of regex retries in iteration 71: 2 -[2025-11-09 19:01:53,065][__main__][INFO] - agents played in iteration 71 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:01:53,856][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:01:53,885][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:01:53,910][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:01:53,934][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:01:53,934][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:01:53,936][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:01:54,563][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:01:55,020][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:01:55,524][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:01:56,026][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:01:56,524][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:01:57,021][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:01:57,517][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:01:58,013][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:01:58,514][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:01:59,011][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:01:59,512][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:02:00,012][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:02:00,515][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:02:01,011][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:02:01,509][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:02:02,005][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:02:02,507][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:02:03,005][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:02:03,501][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:02:03,998][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:02:04,495][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:02:05,000][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:02:05,501][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:02:05,997][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:02:06,494][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:02:06,990][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:02:07,494][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:02:07,988][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:02:08,486][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:02:08,982][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:02:09,479][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:02:09,974][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:02:10,471][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:02:10,970][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:02:11,468][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:02:11,965][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:02:12,462][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:02:12,974][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:02:13,471][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:02:13,971][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:02:14,471][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:02:14,965][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:02:15,459][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:02:15,953][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:02:16,448][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:02:16,942][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:02:17,435][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:02:17,930][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:02:18,424][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:02:18,920][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:02:19,415][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:02:19,915][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:02:20,418][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:02:20,916][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:02:21,418][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:02:21,920][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:02:22,419][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:02:22,918][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:02:23,414][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:02:23,914][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:02:24,418][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:02:24,915][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:02:25,411][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:02:25,911][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:02:26,412][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10687 tokens. -[2025-11-09 19:02:27,109][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 55.72%, Block Peak % of device VRAM: 62.36%, ΔTime: 00:00:32 -[2025-11-09 19:02:27,859][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:02:27,861][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:02:27,863][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:02:28,835][__main__][INFO] - Iteration 72 took 51s (31.13% Gen, 66.99% Train). Generation: 16s, Training: 34s. Estimated remaining time: 42h 13m 38s. Estimated total time: 43h 17m 10s. Time estimates for 10 more iterations: 8m 39s, 100 more iterations: 1h 26m 34s, 500 more iterations: 7h 12m 51s. -[2025-11-09 19:02:28,838][__main__][INFO] - Starting iteration 72. -[2025-11-09 19:02:29,306][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 7 and human policies 1. -[2025-11-09 19:02:29,307][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:02:31,510][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:02:44,437][__main__][INFO] - Number of regex retries in iteration 72: 1 -[2025-11-09 19:02:44,438][__main__][INFO] - agents played in iteration 72 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:02:45,283][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:02:45,313][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:02:45,342][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:02:45,367][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:02:45,367][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:02:45,368][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:02:46,029][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:02:46,481][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:02:46,980][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:02:47,488][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:02:47,983][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:02:48,476][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:02:48,971][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:02:49,468][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:02:49,963][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:02:50,461][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:02:50,958][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:02:51,458][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:02:51,955][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:02:52,450][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:02:52,944][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:02:53,439][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:02:53,932][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:02:54,426][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:02:54,923][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:02:55,418][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:02:55,913][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:02:56,412][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:02:56,907][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:02:57,401][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:02:57,899][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:02:58,395][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:02:58,889][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:02:59,385][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:02:59,880][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:03:00,394][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:03:00,907][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:03:01,405][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:03:01,901][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:03:02,399][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:03:02,894][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:03:03,391][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:03:03,885][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:03:04,384][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:03:04,877][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:03:05,374][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:03:05,884][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:03:06,377][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:03:06,870][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:03:07,363][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:03:07,856][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:03:08,352][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:03:08,846][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:03:09,339][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:03:09,832][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:03:10,328][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:03:10,827][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:03:11,326][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:03:11,825][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:03:12,326][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:03:12,826][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:03:13,329][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:03:13,830][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:03:14,325][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:03:14,822][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:03:15,319][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:03:15,817][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:03:16,313][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:03:16,808][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:03:17,319][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:03:17,813][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10423 tokens. -[2025-11-09 19:03:18,499][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 55.64%, Block Peak % of device VRAM: 62.04%, ΔTime: 00:00:32 -[2025-11-09 19:03:19,253][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:03:19,254][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:03:19,256][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:03:20,205][__main__][INFO] - Iteration 73 took 50s (29.73% Gen, 68.41% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 20m 35s. Estimated total time: 42h 24m 58s. Time estimates for 10 more iterations: 8m 28s, 100 more iterations: 1h 24m 49s, 500 more iterations: 7h 4m 9s. -[2025-11-09 19:03:20,207][__main__][INFO] - Starting iteration 73. -[2025-11-09 19:03:20,677][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 7 and human policies 1. -[2025-11-09 19:03:20,678][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:03:25,415][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:03:35,875][__main__][INFO] - Number of regex retries in iteration 73: 1 -[2025-11-09 19:03:35,876][__main__][INFO] - agents played in iteration 73 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:03:36,677][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:03:36,705][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:03:36,731][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:03:36,754][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:03:36,755][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:03:36,756][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:03:37,412][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:03:37,866][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:03:38,370][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:03:38,864][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:03:39,363][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:03:39,856][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:03:40,352][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:03:40,852][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:03:41,348][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:03:41,844][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:03:42,338][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:03:42,832][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:03:43,328][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:03:43,822][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:03:44,319][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:03:44,821][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:03:45,318][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:03:45,815][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:03:46,328][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:03:46,825][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:03:47,326][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:03:47,822][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:03:48,316][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:03:48,817][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:03:49,313][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:03:49,808][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:03:50,302][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:03:50,796][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:03:51,294][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:03:51,793][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:03:52,290][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:03:52,799][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:03:53,294][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:03:53,790][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:03:54,296][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:03:54,792][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:03:55,292][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:03:55,790][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:03:56,288][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:03:56,799][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:03:57,299][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:03:57,797][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:03:58,327][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:03:58,829][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:03:59,331][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:03:59,829][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:04:00,327][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:04:00,828][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:04:01,328][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:04:01,831][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:04:02,329][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:04:02,826][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:04:03,331][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:04:03,831][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:04:04,332][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:04:04,833][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:04:05,337][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:04:05,841][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:04:06,342][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:04:06,841][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:04:07,348][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:04:07,846][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:04:08,348][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:04:08,849][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:04:09,346][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10608 tokens. -[2025-11-09 19:04:10,022][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 55.70%, Block Peak % of device VRAM: 62.20%, ΔTime: 00:00:32 -[2025-11-09 19:04:10,800][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:04:10,802][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:04:10,804][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:04:11,873][__main__][INFO] - Iteration 74 took 51s (29.69% Gen, 68.22% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 34m 34s. Estimated total time: 42h 39m 49s. Time estimates for 10 more iterations: 8m 31s, 100 more iterations: 1h 25m 19s, 500 more iterations: 7h 6m 38s. -[2025-11-09 19:04:11,875][__main__][INFO] - Starting iteration 74. -[2025-11-09 19:04:12,371][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 7 and human policies 1. -[2025-11-09 19:04:12,372][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:04:27,869][__main__][INFO] - Number of regex retries in iteration 74: 0 -[2025-11-09 19:04:27,869][__main__][INFO] - agents played in iteration 74 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:04:28,699][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:04:28,726][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:04:28,752][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:04:28,775][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:04:28,775][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:04:28,776][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:04:29,454][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:04:29,909][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:04:30,413][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:04:30,920][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:04:31,416][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:04:31,911][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:04:32,413][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:04:32,909][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:04:33,402][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:04:33,898][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:04:34,392][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:04:34,888][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:04:35,383][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:04:35,877][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:04:36,375][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:04:36,870][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:04:37,366][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:04:37,863][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:04:38,356][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:04:38,855][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:04:39,350][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:04:39,843][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:04:40,339][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:04:40,831][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:04:41,328][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:04:41,824][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:04:42,319][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:04:42,831][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:04:43,332][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:04:43,825][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:04:44,338][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:04:44,833][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:04:45,329][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:04:45,827][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:04:46,320][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:04:46,816][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:04:47,311][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:04:47,807][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:04:48,299][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:04:48,807][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:04:49,301][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:04:49,794][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:04:50,287][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:04:50,778][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:04:51,274][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:04:51,771][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:04:52,268][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:04:52,763][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:04:53,260][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:04:53,755][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:04:54,249][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:04:54,744][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:04:55,248][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:04:55,749][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:04:56,249][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:04:56,746][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:04:57,245][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:04:57,750][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:04:58,247][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:04:58,773][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:04:59,277][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:04:59,779][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:05:00,273][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:05:00,770][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:05:01,270][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10545 tokens. -[2025-11-09 19:05:01,972][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.28%, Current % of VRAM taken: 55.77%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 -[2025-11-09 19:05:02,760][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:05:02,762][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:05:02,764][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:05:04,108][__main__][INFO] - Iteration 75 took 51s (29.95% Gen, 67.45% Train). Generation: 15s, Training: 34s. Estimated remaining time: 42h 0m 44s. Estimated total time: 43h 6m 51s. Time estimates for 10 more iterations: 8m 37s, 100 more iterations: 1h 26m 13s, 500 more iterations: 7h 11m 8s. -[2025-11-09 19:05:04,110][__main__][INFO] - Starting iteration 75. -[2025-11-09 19:05:04,593][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 7 and human policies 1. -[2025-11-09 19:05:04,594][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:05:13,368][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:05:20,119][__main__][INFO] - Number of regex retries in iteration 75: 1 -[2025-11-09 19:05:20,120][__main__][INFO] - agents played in iteration 75 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:05:20,926][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:05:20,953][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:05:20,980][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:05:21,003][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:05:21,004][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:05:21,004][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:05:21,655][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:05:22,111][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:05:22,614][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:05:23,114][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:05:23,612][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:05:24,110][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:05:24,609][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:05:25,108][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:05:25,607][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:05:26,103][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:05:26,596][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:05:27,092][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:05:27,589][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:05:28,086][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:05:28,582][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:05:29,075][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:05:29,573][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:05:30,070][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:05:30,571][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:05:31,073][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:05:31,571][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:05:32,064][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:05:32,560][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:05:33,058][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:05:33,554][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:05:34,051][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:05:34,554][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:05:35,050][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:05:35,546][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:05:36,040][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:05:36,536][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:05:37,047][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:05:37,540][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:05:38,033][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:05:38,531][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:05:39,028][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:05:39,520][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:05:40,016][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:05:40,511][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:05:41,010][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:05:41,507][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:05:42,003][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:05:42,506][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:05:43,002][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:05:43,499][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:05:43,996][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:05:44,493][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:05:44,992][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:05:45,490][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:05:45,985][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:05:46,483][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:05:46,980][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:05:47,483][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:05:47,982][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:05:48,478][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:05:48,980][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:05:49,480][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:05:49,978][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:05:50,477][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:05:50,976][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:05:51,475][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:05:51,975][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:05:52,472][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:05:52,968][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:05:53,466][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10619 tokens. -[2025-11-09 19:05:54,162][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 55.71%, Block Peak % of device VRAM: 62.39%, ΔTime: 00:00:32 -[2025-11-09 19:05:54,937][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:05:54,938][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:05:54,940][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:05:55,920][__main__][INFO] - Iteration 76 took 51s (30.25% Gen, 67.84% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 39m 21s. Estimated total time: 42h 46m 20s. Time estimates for 10 more iterations: 8m 33s, 100 more iterations: 1h 25m 32s, 500 more iterations: 7h 7m 43s. -[2025-11-09 19:05:55,922][__main__][INFO] - Starting iteration 76. -[2025-11-09 19:05:56,410][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 7 and human policies 1. -[2025-11-09 19:05:56,411][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:06:11,791][__main__][INFO] - Number of regex retries in iteration 76: 0 -[2025-11-09 19:06:11,792][__main__][INFO] - agents played in iteration 76 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:06:12,608][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:06:12,631][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:06:12,653][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:06:12,675][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:06:12,676][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:06:12,678][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:06:13,315][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:06:13,776][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:06:14,275][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:06:14,769][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:06:15,268][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:06:15,766][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:06:16,260][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:06:16,754][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:06:17,250][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:06:17,750][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:06:18,244][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:06:18,739][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:06:19,234][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:06:19,730][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:06:20,234][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:06:20,729][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:06:21,233][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:06:21,748][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:06:22,243][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:06:22,741][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:06:23,237][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:06:23,731][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:06:24,241][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:06:24,736][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:06:25,233][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:06:25,729][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:06:26,225][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:06:26,719][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:06:27,215][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:06:27,708][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:06:28,206][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:06:28,702][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:06:29,197][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:06:29,691][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:06:30,184][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:06:30,681][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:06:31,173][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:06:31,666][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:06:32,160][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:06:32,653][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:06:33,146][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:06:33,642][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:06:34,134][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:06:34,627][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:06:35,119][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:06:35,611][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:06:36,104][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:06:36,596][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:06:37,092][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:06:37,586][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:06:38,085][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:06:38,585][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:06:39,081][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:06:39,583][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:06:40,084][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:06:40,581][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:06:41,079][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:06:41,574][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:06:42,070][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:06:42,565][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:06:43,063][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:06:43,559][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:06:44,055][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:06:44,551][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:06:45,044][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10396 tokens. -[2025-11-09 19:06:45,711][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 55.57%, Block Peak % of device VRAM: 62.20%, ΔTime: 00:00:32 -[2025-11-09 19:06:46,497][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:06:46,499][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:06:46,501][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:06:47,431][__main__][INFO] - Iteration 77 took 51s (30.15% Gen, 68.03% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 23m 14s. Estimated total time: 42h 31m 5s. Time estimates for 10 more iterations: 8m 30s, 100 more iterations: 1h 25m 2s, 500 more iterations: 7h 5m 10s. -[2025-11-09 19:06:47,433][__main__][INFO] - Starting iteration 77. -[2025-11-09 19:06:47,931][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 7 and human policies 1. -[2025-11-09 19:06:47,932][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:06:50,214][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:07:02,507][__main__][INFO] - Number of regex retries in iteration 77: 1 -[2025-11-09 19:07:02,508][__main__][INFO] - agents played in iteration 77 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:07:03,394][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:07:03,419][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:07:03,444][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:07:03,467][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:07:03,467][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:07:03,468][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:07:04,061][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:07:04,517][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:07:05,014][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:07:05,514][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:07:06,018][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:07:06,518][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:07:07,019][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:07:07,519][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:07:08,016][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:07:08,521][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:07:09,020][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:07:09,516][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:07:10,014][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:07:10,511][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:07:11,008][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:07:11,507][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:07:12,005][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:07:12,501][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:07:12,996][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:07:13,493][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:07:13,986][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:07:14,483][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:07:14,978][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:07:15,483][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:07:15,979][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:07:16,471][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:07:16,966][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:07:17,459][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:07:17,973][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:07:18,469][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:07:18,962][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:07:19,456][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:07:19,950][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:07:20,448][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:07:20,944][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:07:21,438][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:07:21,932][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:07:22,427][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:07:22,924][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:07:23,418][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:07:23,912][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:07:24,408][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:07:24,903][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:07:25,397][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:07:25,893][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:07:26,388][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:07:26,881][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:07:27,374][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:07:27,871][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:07:28,367][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:07:28,865][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:07:29,362][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:07:29,862][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:07:30,357][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:07:30,855][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:07:31,355][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:07:31,856][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:07:32,378][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:07:32,876][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:07:33,375][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:07:33,873][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:07:34,371][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:07:34,872][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:07:35,371][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:07:35,870][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10575 tokens. -[2025-11-09 19:07:36,519][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 55.67%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:32 -[2025-11-09 19:07:37,298][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:07:37,304][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:07:37,306][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:07:38,265][__main__][INFO] - Iteration 78 took 50s (28.96% Gen, 69.13% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 48m 2s. Estimated total time: 41h 56m 43s. Time estimates for 10 more iterations: 8m 23s, 100 more iterations: 1h 23m 53s, 500 more iterations: 6h 59m 27s. -[2025-11-09 19:07:38,267][__main__][INFO] - Starting iteration 78. -[2025-11-09 19:07:38,745][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 7 and human policies 1. -[2025-11-09 19:07:38,746][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:07:54,685][__main__][INFO] - Number of regex retries in iteration 78: 0 -[2025-11-09 19:07:54,686][__main__][INFO] - agents played in iteration 78 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:07:55,468][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:07:55,496][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:07:55,522][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:07:55,545][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:07:55,545][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:07:55,546][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:07:56,192][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:07:56,645][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:07:57,148][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:07:57,642][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:07:58,138][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:07:58,641][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:07:59,140][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:07:59,633][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:08:00,130][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:08:00,626][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:08:01,119][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:08:01,614][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:08:02,109][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:08:02,603][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:08:03,097][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:08:03,592][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:08:04,093][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:08:04,588][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:08:05,084][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:08:05,581][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:08:06,074][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:08:06,572][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:08:07,069][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:08:07,562][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:08:08,058][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:08:08,554][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:08:09,076][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:08:09,576][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:08:10,075][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:08:10,574][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:08:11,071][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:08:11,569][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:08:12,069][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:08:12,563][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:08:13,056][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:08:13,551][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:08:14,049][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:08:14,553][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:08:15,048][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:08:15,540][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:08:16,036][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:08:16,534][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:08:17,029][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:08:17,524][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:08:18,016][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:08:18,510][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:08:19,006][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:08:19,502][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:08:20,002][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:08:20,497][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:08:20,995][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:08:21,492][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:08:21,988][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:08:22,488][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:08:22,985][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:08:23,486][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:08:23,990][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:08:24,493][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:08:24,993][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:08:25,493][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:08:25,990][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:08:26,495][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:08:26,993][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:08:27,496][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:08:27,994][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10585 tokens. -[2025-11-09 19:08:28,623][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 55.66%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 -[2025-11-09 19:08:29,419][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:08:29,423][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:08:29,425][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:08:30,369][__main__][INFO] - Iteration 79 took 51s (30.88% Gen, 67.29% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 51m 41s. Estimated total time: 43h 1m 15s. Time estimates for 10 more iterations: 8m 36s, 100 more iterations: 1h 26m 2s, 500 more iterations: 7h 10m 12s. -[2025-11-09 19:08:30,372][__main__][INFO] - Starting iteration 79. -[2025-11-09 19:08:30,834][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 7 and human policies 1. -[2025-11-09 19:08:30,835][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:08:45,964][__main__][INFO] - Number of regex retries in iteration 79: 0 -[2025-11-09 19:08:45,965][__main__][INFO] - agents played in iteration 79 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:08:46,742][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:08:46,769][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:08:46,796][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:08:46,819][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:08:46,819][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:08:46,820][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:08:47,440][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:08:47,894][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:08:48,399][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:08:48,897][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:08:49,393][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:08:49,897][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:08:50,395][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:08:50,893][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:08:51,388][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:08:51,889][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:08:52,386][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:08:52,879][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:08:53,376][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:08:53,886][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:08:54,402][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:08:54,902][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:08:55,396][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:08:55,889][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:08:56,389][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:08:56,882][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:08:57,374][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:08:57,866][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:08:58,362][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:08:58,858][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:08:59,353][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:08:59,848][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:09:00,362][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:09:00,859][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:09:01,355][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:09:01,851][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:09:02,347][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:09:02,842][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:09:03,337][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:09:03,831][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:09:04,327][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:09:04,825][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:09:05,322][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:09:05,818][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:09:06,314][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:09:06,811][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:09:07,306][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:09:07,799][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:09:08,294][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:09:08,790][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:09:09,285][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:09:09,779][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:09:10,272][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:09:10,786][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:09:11,280][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:09:11,775][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:09:12,271][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:09:12,768][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:09:13,274][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:09:13,771][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:09:14,270][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:09:14,770][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:09:15,274][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:09:15,772][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:09:16,267][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:09:16,763][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:09:17,278][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:09:17,776][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:09:18,276][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:09:18,778][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:09:19,275][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10599 tokens. -[2025-11-09 19:09:19,931][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 55.66%, Block Peak % of device VRAM: 62.28%, ΔTime: 00:00:32 -[2025-11-09 19:09:20,704][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:09:20,705][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:09:20,707][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:09:21,625][__main__][INFO] - Iteration 80 took 50s (29.79% Gen, 68.40% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 9m 8s. Estimated total time: 42h 19m 33s. Time estimates for 10 more iterations: 8m 27s, 100 more iterations: 1h 24m 39s, 500 more iterations: 7h 3m 15s. -[2025-11-09 19:09:21,627][__main__][INFO] - Starting iteration 80. -[2025-11-09 19:09:22,090][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 7 and human policies 1. -[2025-11-09 19:09:22,091][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:09:37,036][__main__][INFO] - Number of regex retries in iteration 80: 0 -[2025-11-09 19:09:37,036][__main__][INFO] - agents played in iteration 80 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:09:37,826][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:09:37,852][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:09:37,877][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:09:37,911][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:09:37,912][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:09:37,913][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:09:38,567][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:09:39,022][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:09:39,523][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:09:40,023][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:09:40,529][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:09:41,029][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:09:41,534][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:09:42,040][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:09:42,537][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:09:43,035][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:09:43,534][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:09:44,032][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:09:44,534][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:09:45,034][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:09:45,532][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:09:46,033][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:09:46,534][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:09:47,032][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:09:47,531][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:09:48,026][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:09:48,529][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:09:49,025][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:09:49,522][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:09:50,020][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:09:50,518][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:09:51,013][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:09:51,510][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:09:52,009][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:09:52,511][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:09:53,007][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:09:53,505][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:09:54,006][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:09:54,502][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:09:55,009][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:09:55,504][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:09:56,004][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:09:56,503][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:09:56,997][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:09:57,500][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:09:57,998][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:09:58,496][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:09:58,997][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:09:59,494][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:09:59,994][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:10:00,492][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:10:00,989][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:10:01,499][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:10:01,997][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:10:02,499][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:10:03,017][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:10:03,515][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:10:04,013][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:10:04,516][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:10:05,015][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:10:05,514][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:10:06,015][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:10:06,517][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:10:07,016][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:10:07,512][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:10:08,015][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:10:08,515][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:10:09,014][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:10:09,514][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:10:10,014][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:10:10,512][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10638 tokens. -[2025-11-09 19:10:11,169][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 55.72%, Block Peak % of device VRAM: 62.49%, ΔTime: 00:00:32 -[2025-11-09 19:10:11,965][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:10:11,966][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:10:11,968][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:10:14,893][__main__][INFO] - Iteration 81 took 52s (28.30% Gen, 66.15% Train). Generation: 14s, Training: 34s. Estimated remaining time: 42h 48m 50s. Estimated total time: 44h 0m 8s. Time estimates for 10 more iterations: 8m 48s, 100 more iterations: 1h 28m 0s, 500 more iterations: 7h 20m 1s. -[2025-11-09 19:10:14,895][__main__][INFO] - Starting iteration 81. -[2025-11-09 19:10:15,357][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 8 and human policies 1. -[2025-11-09 19:10:15,357][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:10:26,018][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:10:30,572][__main__][INFO] - Number of regex retries in iteration 81: 1 -[2025-11-09 19:10:30,573][__main__][INFO] - agents played in iteration 81 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:10:31,348][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:10:31,380][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:10:31,416][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:10:31,443][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:10:31,443][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:10:31,444][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:10:32,054][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:10:32,509][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:10:33,015][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:10:33,513][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:10:34,009][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:10:34,506][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:10:35,003][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:10:35,504][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:10:36,004][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:10:36,499][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:10:37,009][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:10:37,504][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:10:37,999][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:10:38,496][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:10:38,992][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:10:39,499][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:10:39,995][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:10:40,495][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:10:40,991][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:10:41,490][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:10:41,988][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:10:42,488][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:10:42,987][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:10:43,500][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:10:43,995][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:10:44,491][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:10:44,989][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:10:45,486][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:10:45,981][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:10:46,478][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:10:46,972][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:10:47,494][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:10:47,990][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:10:48,492][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:10:49,008][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:10:49,505][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:10:50,008][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:10:50,506][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:10:51,007][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:10:51,512][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:10:52,011][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:10:52,508][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:10:53,005][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:10:53,501][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:10:53,996][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:10:54,494][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:10:54,991][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:10:55,505][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:10:56,001][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:10:56,499][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:10:56,994][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:10:57,492][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:10:57,998][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:10:58,495][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:10:58,992][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:10:59,509][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:11:00,006][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:11:00,501][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:11:00,999][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:11:01,495][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:11:02,007][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:11:02,507][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:11:03,008][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:11:03,508][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:11:04,008][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10635 tokens. -[2025-11-09 19:11:04,671][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 55.63%, Block Peak % of device VRAM: 62.22%, ΔTime: 00:00:32 -[2025-11-09 19:11:05,445][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:11:05,447][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:11:05,449][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:11:06,324][__main__][INFO] - Iteration 82 took 50s (29.85% Gen, 68.43% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 16m 13s. Estimated total time: 42h 28m 22s. Time estimates for 10 more iterations: 8m 29s, 100 more iterations: 1h 24m 56s, 500 more iterations: 7h 4m 43s. -[2025-11-09 19:11:06,326][__main__][INFO] - Starting iteration 82. -[2025-11-09 19:11:06,786][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 8 and human policies 1. -[2025-11-09 19:11:06,787][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:11:12,244][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:11:20,377][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:11:21,647][__main__][INFO] - Number of regex retries in iteration 82: 2 -[2025-11-09 19:11:21,648][__main__][INFO] - agents played in iteration 82 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:11:22,467][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:11:22,493][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:11:22,518][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:11:22,541][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:11:22,542][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:11:22,542][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:11:23,219][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:11:23,674][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:11:24,179][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:11:24,676][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:11:25,176][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:11:25,698][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:11:26,196][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:11:26,693][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:11:27,194][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:11:27,697][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:11:28,195][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:11:28,694][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:11:29,189][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:11:29,686][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:11:30,180][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:11:30,679][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:11:31,177][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:11:31,694][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:11:32,190][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:11:32,690][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:11:33,186][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:11:33,686][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:11:34,183][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:11:34,687][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:11:35,189][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:11:35,689][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:11:36,202][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:11:36,697][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:11:37,194][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:11:37,692][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:11:38,188][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:11:38,684][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:11:39,189][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:11:39,684][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:11:40,185][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:11:40,680][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:11:41,174][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:11:41,672][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:11:42,167][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:11:42,664][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:11:43,159][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:11:43,656][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:11:44,154][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:11:44,651][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:11:45,148][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:11:45,646][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:11:46,142][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:11:46,639][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:11:47,139][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:11:47,634][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:11:48,133][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:11:48,629][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:11:49,126][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:11:49,627][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:11:50,122][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:11:50,619][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:11:51,120][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:11:51,617][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:11:52,113][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:11:52,610][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:11:53,107][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:11:53,622][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:11:54,122][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:11:54,621][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:11:55,119][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10792 tokens. -[2025-11-09 19:11:55,810][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 55.71%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:32 -[2025-11-09 19:11:56,590][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:11:56,591][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:11:56,593][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:11:57,525][__main__][INFO] - Iteration 83 took 50s (29.29% Gen, 68.87% Train). Generation: 14s, Training: 34s. Estimated remaining time: 41h 3m 58s. Estimated total time: 42h 16m 59s. Time estimates for 10 more iterations: 8m 27s, 100 more iterations: 1h 24m 33s, 500 more iterations: 7h 2m 49s. -[2025-11-09 19:11:57,528][__main__][INFO] - Starting iteration 83. -[2025-11-09 19:11:58,035][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 8 and human policies 1. -[2025-11-09 19:11:58,035][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:12:06,264][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:12:13,117][__main__][INFO] - Number of regex retries in iteration 83: 1 -[2025-11-09 19:12:13,118][__main__][INFO] - agents played in iteration 83 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:12:13,918][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:12:13,946][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:12:13,974][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:12:13,997][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:12:13,998][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:12:13,999][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:12:14,664][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:12:15,121][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:12:15,631][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:12:16,129][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:12:16,633][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:12:17,136][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:12:17,635][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:12:18,162][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:12:18,664][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:12:19,166][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:12:19,669][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:12:20,169][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:12:20,668][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:12:21,168][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:12:21,667][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:12:22,167][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:12:22,662][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:12:23,159][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:12:23,659][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:12:24,156][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:12:24,654][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:12:25,155][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:12:25,652][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:12:26,150][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:12:26,648][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:12:27,147][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:12:27,644][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:12:28,140][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:12:28,634][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:12:29,131][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:12:29,628][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:12:30,132][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:12:30,631][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:12:31,127][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:12:31,624][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:12:32,124][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:12:32,625][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:12:33,120][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:12:33,616][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:12:34,126][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:12:34,621][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:12:35,119][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:12:35,616][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:12:36,111][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:12:36,621][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:12:37,118][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:12:37,614][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:12:38,112][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:12:38,609][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:12:39,106][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:12:39,606][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:12:40,101][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:12:40,599][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:12:41,095][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:12:41,594][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:12:42,107][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:12:42,601][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:12:43,098][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:12:43,594][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:12:44,089][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:12:44,587][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:12:45,082][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:12:45,577][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:12:46,074][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:12:46,575][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10659 tokens. -[2025-11-09 19:12:47,245][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 55.68%, Block Peak % of device VRAM: 62.35%, ΔTime: 00:00:32 -[2025-11-09 19:12:48,012][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:12:48,014][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:12:48,015][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:12:48,966][__main__][INFO] - Iteration 84 took 50s (29.61% Gen, 68.52% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 12m 42s. Estimated total time: 42h 26m 34s. Time estimates for 10 more iterations: 8m 29s, 100 more iterations: 1h 24m 53s, 500 more iterations: 7h 4m 25s. -[2025-11-09 19:12:48,968][__main__][INFO] - Starting iteration 84. -[2025-11-09 19:12:49,528][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 8 and human policies 1. -[2025-11-09 19:12:49,529][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:13:04,355][__main__][INFO] - Number of regex retries in iteration 84: 0 -[2025-11-09 19:13:04,356][__main__][INFO] - agents played in iteration 84 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:13:05,153][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:13:05,180][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:13:05,207][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:13:05,230][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:13:05,230][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:13:05,231][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:13:05,913][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:13:06,374][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:13:06,880][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:13:07,379][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:13:07,883][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:13:08,385][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:13:08,885][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:13:09,383][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:13:09,891][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:13:10,391][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:13:10,890][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:13:11,391][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:13:11,889][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:13:12,386][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:13:12,880][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:13:13,378][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:13:13,873][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:13:14,375][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:13:14,873][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:13:15,374][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:13:15,872][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:13:16,372][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:13:16,871][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:13:17,370][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:13:17,869][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:13:18,366][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:13:18,865][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:13:19,360][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:13:19,858][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:13:20,355][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:13:20,852][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:13:21,351][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:13:21,849][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:13:22,350][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:13:22,851][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:13:23,349][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:13:23,847][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:13:24,346][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:13:24,844][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:13:25,341][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:13:25,838][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:13:26,334][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:13:26,837][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:13:27,334][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:13:27,830][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:13:28,342][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:13:28,838][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:13:29,334][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:13:29,830][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:13:30,324][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:13:30,836][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:13:31,334][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:13:31,828][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:13:32,326][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:13:32,823][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:13:33,319][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:13:33,815][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:13:34,311][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:13:34,811][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:13:35,309][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:13:35,807][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:13:36,305][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:13:36,801][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:13:37,295][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:13:37,793][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10678 tokens. -[2025-11-09 19:13:38,422][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:32 -[2025-11-09 19:13:39,169][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:13:39,171][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:13:39,172][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:13:40,185][__main__][INFO] - Iteration 85 took 50s (29.27% Gen, 68.73% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 58m 7s. Estimated total time: 42h 12m 50s. Time estimates for 10 more iterations: 8m 26s, 100 more iterations: 1h 24m 25s, 500 more iterations: 7h 2m 8s. -[2025-11-09 19:13:40,187][__main__][INFO] - Starting iteration 85. -[2025-11-09 19:13:40,653][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 8 and human policies 1. -[2025-11-09 19:13:40,654][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:13:55,698][__main__][INFO] - Number of regex retries in iteration 85: 0 -[2025-11-09 19:13:55,698][__main__][INFO] - agents played in iteration 85 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:13:56,583][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:13:56,606][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:13:56,629][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:13:56,651][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:13:56,652][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:13:56,653][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:13:57,328][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:13:57,784][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:13:58,289][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:13:58,788][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:13:59,288][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:13:59,788][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:14:00,292][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:14:00,791][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:14:01,291][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:14:01,792][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:14:02,294][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:14:02,795][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:14:03,295][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:14:03,793][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:14:04,289][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:14:04,786][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:14:05,286][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:14:05,780][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:14:06,276][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:14:06,773][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:14:07,266][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:14:07,763][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:14:08,259][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:14:08,753][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:14:09,250][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:14:09,746][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:14:10,244][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:14:10,739][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:14:11,234][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:14:11,737][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:14:12,232][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:14:12,724][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:14:13,220][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:14:13,716][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:14:14,238][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:14:14,736][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:14:15,231][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:14:15,743][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:14:16,241][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:14:16,739][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:14:17,251][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:14:17,745][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:14:18,241][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:14:18,737][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:14:19,231][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:14:19,745][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:14:20,239][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:14:20,734][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:14:21,232][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:14:21,733][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:14:22,241][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:14:22,739][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:14:23,234][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:14:23,747][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:14:24,245][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:14:24,742][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:14:25,239][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:14:25,735][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:14:26,228][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:14:26,722][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:14:27,218][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:14:27,712][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:14:28,208][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:14:28,704][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:14:29,207][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10682 tokens. -[2025-11-09 19:14:29,899][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 55.66%, Block Peak % of device VRAM: 62.44%, ΔTime: 00:00:32 -[2025-11-09 19:14:30,679][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:14:30,680][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:14:30,683][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:14:31,624][__main__][INFO] - Iteration 86 took 50s (29.52% Gen, 68.64% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 12m 58s. Estimated total time: 42h 28m 33s. Time estimates for 10 more iterations: 8m 29s, 100 more iterations: 1h 24m 57s, 500 more iterations: 7h 4m 45s. -[2025-11-09 19:14:31,626][__main__][INFO] - Starting iteration 86. -[2025-11-09 19:14:32,142][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 8 and human policies 1. -[2025-11-09 19:14:32,142][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:14:46,236][__main__][INFO] - Number of regex retries in iteration 86: 0 -[2025-11-09 19:14:46,237][__main__][INFO] - agents played in iteration 86 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:14:47,118][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:14:47,145][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:14:47,172][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:14:47,194][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:14:47,195][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:14:47,196][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:14:47,809][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:14:48,261][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:14:48,777][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:14:49,278][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:14:49,778][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:14:50,279][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:14:50,775][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:14:51,282][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:14:51,782][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:14:52,294][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:14:52,795][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:14:53,294][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:14:53,796][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:14:54,294][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:14:54,792][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:14:55,291][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:14:55,788][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:14:56,290][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:14:56,789][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:14:57,287][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:14:57,785][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:14:58,282][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:14:58,779][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:14:59,279][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:14:59,776][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:15:00,274][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:15:00,789][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:15:01,287][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:15:01,787][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:15:02,284][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:15:02,961][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:15:03,460][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:15:03,956][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:15:04,469][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:15:04,964][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:15:05,459][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:15:05,954][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:15:06,451][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:15:06,959][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:15:07,455][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:15:07,950][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:15:08,451][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:15:08,949][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:15:09,445][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:15:09,942][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:15:10,438][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:15:10,935][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:15:11,432][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:15:11,929][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:15:12,428][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:15:12,922][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:15:13,416][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:15:13,914][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:15:14,409][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:15:14,906][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:15:15,398][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:15:15,893][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:15:16,391][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:15:16,887][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:15:17,385][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:15:17,879][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:15:18,374][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:15:18,876][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:15:19,372][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:15:19,868][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10761 tokens. -[2025-11-09 19:15:20,503][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 55.71%, Block Peak % of device VRAM: 62.37%, ΔTime: 00:00:32 -[2025-11-09 19:15:21,263][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:15:21,264][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:15:21,266][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:15:22,776][__main__][INFO] - Iteration 87 took 50s (27.84% Gen, 69.18% Train). Generation: 14s, Training: 35s. Estimated remaining time: 40h 55m 19s. Estimated total time: 42h 11m 45s. Time estimates for 10 more iterations: 8m 26s, 100 more iterations: 1h 24m 23s, 500 more iterations: 7h 1m 57s. -[2025-11-09 19:15:22,779][__main__][INFO] - Starting iteration 87. -[2025-11-09 19:15:23,243][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 8 and human policies 1. -[2025-11-09 19:15:23,243][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:15:37,398][__main__][INFO] - Number of regex retries in iteration 87: 0 -[2025-11-09 19:15:37,399][__main__][INFO] - agents played in iteration 87 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:15:38,196][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:15:38,224][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:15:38,250][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:15:38,273][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:15:38,273][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:15:38,274][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:15:38,894][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:15:39,351][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:15:39,856][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:15:40,355][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:15:40,858][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:15:41,363][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:15:41,865][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:15:42,377][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:15:42,878][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:15:43,376][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:15:43,878][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:15:44,381][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:15:44,882][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:15:45,382][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:15:45,881][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:15:46,389][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:15:46,885][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:15:47,383][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:15:47,880][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:15:48,376][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:15:48,873][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:15:49,370][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:15:49,868][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:15:50,379][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:15:50,875][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:15:51,372][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:15:51,876][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:15:52,372][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:15:52,879][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:15:53,377][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:15:53,874][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:15:54,372][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:15:54,871][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:15:55,370][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:15:55,867][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:15:56,364][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:15:56,875][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:15:57,371][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:15:57,870][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:15:58,371][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:15:58,868][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:15:59,374][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:15:59,869][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:16:00,366][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:16:00,863][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:16:01,359][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:16:01,855][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:16:02,354][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:16:02,850][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:16:03,348][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:16:03,844][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:16:04,339][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:16:04,847][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:16:05,344][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:16:05,840][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:16:06,336][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:16:06,832][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:16:07,330][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:16:07,827][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:16:08,323][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:16:08,818][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:16:09,315][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:16:09,811][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:16:10,307][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:16:10,804][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10718 tokens. -[2025-11-09 19:16:11,418][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 55.67%, Block Peak % of device VRAM: 62.41%, ΔTime: 00:00:32 -[2025-11-09 19:16:12,170][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:16:12,172][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:16:12,174][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:16:13,238][__main__][INFO] - Iteration 88 took 49s (28.31% Gen, 69.56% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 22m 31s. Estimated total time: 41h 39m 47s. Time estimates for 10 more iterations: 8m 19s, 100 more iterations: 1h 23m 19s, 500 more iterations: 6h 56m 37s. -[2025-11-09 19:16:13,240][__main__][INFO] - Starting iteration 88. -[2025-11-09 19:16:13,721][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 8 and human policies 1. -[2025-11-09 19:16:13,721][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:16:16,067][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:16:16,167][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:16:18,252][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:16:27,098][__main__][INFO] - Number of regex retries in iteration 88: 3 -[2025-11-09 19:16:27,099][__main__][INFO] - agents played in iteration 88 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:16:27,881][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:16:27,909][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:16:27,936][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:16:27,960][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:16:27,960][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:16:27,961][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:16:28,612][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:16:29,071][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:16:29,594][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:16:30,098][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:16:30,599][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:16:31,106][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:16:31,607][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:16:32,106][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:16:32,600][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:16:33,098][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:16:33,603][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:16:34,102][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:16:34,602][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:16:35,099][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:16:35,600][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:16:36,102][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:16:36,603][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:16:37,102][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:16:37,600][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:16:38,099][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:16:38,616][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:16:39,117][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:16:39,613][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:16:40,109][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:16:40,606][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:16:41,101][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:16:41,599][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:16:42,096][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:16:42,596][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:16:43,089][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:16:43,585][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:16:44,093][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:16:44,589][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:16:45,086][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:16:45,583][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:16:46,080][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:16:46,584][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:16:47,080][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:16:47,577][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:16:48,090][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:16:48,586][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:16:49,082][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:16:49,579][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:16:50,075][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:16:50,580][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:16:51,077][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:16:51,573][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:16:52,071][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:16:52,567][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:16:53,063][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:16:53,559][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:16:54,052][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:16:54,549][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:16:55,058][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:16:55,554][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:16:56,063][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:16:56,559][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:16:57,055][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:16:57,559][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:16:58,055][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:16:58,556][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:16:59,052][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:16:59,552][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:17:00,051][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:17:00,551][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10812 tokens. -[2025-11-09 19:17:01,185][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 55.71%, Block Peak % of device VRAM: 62.37%, ΔTime: 00:00:32 -[2025-11-09 19:17:01,947][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:17:01,948][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:17:01,950][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:17:03,005][__main__][INFO] - Iteration 89 took 49s (27.14% Gen, 70.71% Train). Generation: 13s, Training: 34s. Estimated remaining time: 39h 46m 9s. Estimated total time: 41h 4m 15s. Time estimates for 10 more iterations: 8m 12s, 100 more iterations: 1h 22m 8s, 500 more iterations: 6h 50m 42s. -[2025-11-09 19:17:03,007][__main__][INFO] - Starting iteration 89. -[2025-11-09 19:17:03,487][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 8 and human policies 1. -[2025-11-09 19:17:03,488][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:17:05,747][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:17:14,280][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:17:18,155][__main__][INFO] - Number of regex retries in iteration 89: 2 -[2025-11-09 19:17:18,155][__main__][INFO] - agents played in iteration 89 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:17:18,955][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:17:18,984][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:17:19,011][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:17:19,035][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:17:19,035][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:17:19,036][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:17:19,661][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:17:20,117][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:17:20,620][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:17:21,125][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:17:21,625][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:17:22,126][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:17:22,626][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:17:23,123][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:17:23,626][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:17:24,123][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:17:24,620][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:17:25,122][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:17:25,619][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:17:26,116][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:17:26,617][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:17:27,117][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:17:27,618][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:17:28,116][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:17:28,614][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:17:29,114][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:17:29,615][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:17:30,112][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:17:30,607][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:17:31,101][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:17:31,602][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:17:32,097][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:17:32,593][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:17:33,091][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:17:33,587][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:17:34,083][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:17:34,580][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:17:35,077][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:17:35,595][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:17:36,090][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:17:36,587][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:17:37,088][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:17:37,581][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:17:38,077][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:17:38,573][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:17:39,068][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:17:39,580][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:17:40,076][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:17:40,573][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:17:41,078][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:17:41,574][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:17:42,071][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:17:42,568][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:17:43,064][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:17:43,560][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:17:44,057][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:17:44,554][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:17:45,064][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:17:45,561][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:17:46,061][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:17:46,561][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:17:47,059][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:17:47,557][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:17:48,055][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:17:48,551][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:17:49,050][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:17:49,550][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:17:50,050][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:17:50,546][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:17:51,042][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:17:51,539][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10832 tokens. -[2025-11-09 19:17:52,165][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 55.68%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 -[2025-11-09 19:17:52,929][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:17:52,930][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:17:52,932][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:17:53,933][__main__][INFO] - Iteration 90 took 50s (29.08% Gen, 68.94% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 43m 23s. Estimated total time: 42h 2m 20s. Time estimates for 10 more iterations: 8m 24s, 100 more iterations: 1h 24m 4s, 500 more iterations: 7h 0m 23s. -[2025-11-09 19:17:53,935][__main__][INFO] - Starting iteration 90. -[2025-11-09 19:17:54,416][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 8 and human policies 1. -[2025-11-09 19:17:54,417][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:18:09,207][__main__][INFO] - Number of regex retries in iteration 90: 0 -[2025-11-09 19:18:09,208][__main__][INFO] - agents played in iteration 90 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:18:09,989][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:18:10,016][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:18:10,042][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:18:10,065][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:18:10,065][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:18:10,066][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:18:10,689][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:18:11,145][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:18:11,651][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:18:12,153][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:18:12,661][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:18:13,160][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:18:13,657][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:18:14,157][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:18:14,655][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:18:15,157][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:18:15,654][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:18:16,152][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:18:16,651][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:18:17,148][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:18:17,648][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:18:18,145][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:18:18,643][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:18:19,145][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:18:19,642][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:18:20,141][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:18:20,640][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:18:21,138][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:18:21,642][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:18:22,141][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:18:22,639][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:18:23,137][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:18:23,634][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:18:24,132][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:18:24,628][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:18:25,124][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:18:25,620][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:18:26,118][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:18:26,615][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:18:27,111][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:18:27,608][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:18:28,112][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:18:28,609][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:18:29,106][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:18:29,607][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:18:30,102][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:18:30,602][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:18:31,096][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:18:31,591][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:18:32,103][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:18:32,598][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:18:33,093][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:18:33,592][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:18:34,088][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:18:34,596][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:18:35,095][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:18:35,591][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:18:36,090][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:18:36,588][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:18:37,086][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:18:37,583][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:18:38,079][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:18:38,599][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:18:39,093][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:18:39,590][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:18:40,091][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:18:40,588][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:18:41,099][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:18:41,594][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:18:42,091][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:18:42,606][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10776 tokens. -[2025-11-09 19:18:43,241][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 55.72%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 -[2025-11-09 19:18:44,008][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:18:44,010][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:18:44,012][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:18:45,827][__main__][INFO] - Iteration 91 took 51s (28.77% Gen, 67.70% Train). Generation: 14s, Training: 34s. Estimated remaining time: 41h 30m 45s. Estimated total time: 42h 50m 34s. Time estimates for 10 more iterations: 8m 34s, 100 more iterations: 1h 25m 41s, 500 more iterations: 7h 8m 25s. -[2025-11-09 19:18:45,829][__main__][INFO] - Starting iteration 91. -[2025-11-09 19:18:46,297][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 9 and human policies 1. -[2025-11-09 19:18:46,297][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:19:00,851][__main__][INFO] - Number of regex retries in iteration 91: 0 -[2025-11-09 19:19:00,852][__main__][INFO] - agents played in iteration 91 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:19:01,637][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:19:01,664][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:19:01,691][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:19:01,714][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:19:01,714][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:19:01,715][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:19:02,333][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:19:02,918][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:19:03,428][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:19:03,931][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:19:04,427][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:19:04,927][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:19:05,428][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:19:05,924][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:19:06,424][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:19:06,921][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:19:07,425][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:19:07,923][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:19:08,423][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:19:08,919][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:19:09,418][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:19:09,921][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:19:10,423][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:19:10,926][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:19:11,427][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:19:11,929][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:19:12,434][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:19:12,934][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:19:13,435][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:19:13,939][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:19:14,438][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:19:14,940][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:19:15,440][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:19:15,938][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:19:16,433][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:19:16,927][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:19:17,423][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:19:17,917][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:19:18,413][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:19:18,917][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:19:19,412][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:19:19,909][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:19:20,407][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:19:20,903][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:19:21,399][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:19:21,896][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:19:22,392][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:19:22,903][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:19:23,398][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:19:23,895][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:19:24,397][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:19:24,892][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:19:25,393][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:19:25,889][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:19:26,384][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:19:26,880][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:19:27,373][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:19:27,871][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:19:28,370][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:19:28,868][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:19:29,367][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:19:29,863][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:19:30,359][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:19:30,857][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:19:31,357][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:19:31,854][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:19:32,352][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:19:32,851][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:19:33,350][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:19:33,847][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:19:34,346][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10790 tokens. -[2025-11-09 19:19:34,976][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.26%, Current % of VRAM taken: 55.76%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 -[2025-11-09 19:19:35,730][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:19:35,731][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:19:35,733][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:19:36,903][__main__][INFO] - Iteration 92 took 50s (28.76% Gen, 68.93% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 49m 40s. Estimated total time: 42h 10m 20s. Time estimates for 10 more iterations: 8m 26s, 100 more iterations: 1h 24m 20s, 500 more iterations: 7h 1m 43s. -[2025-11-09 19:19:36,905][__main__][INFO] - Starting iteration 92. -[2025-11-09 19:19:37,376][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 9 and human policies 1. -[2025-11-09 19:19:37,376][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:19:43,458][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:19:51,112][__main__][INFO] - Number of regex retries in iteration 92: 1 -[2025-11-09 19:19:51,113][__main__][INFO] - agents played in iteration 92 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:19:51,917][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:19:51,939][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:19:51,962][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:19:51,984][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:19:51,985][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:19:51,986][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:19:52,624][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:19:53,088][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:19:53,594][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:19:54,096][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:19:54,613][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:19:55,112][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:19:55,622][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:19:56,121][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:19:56,621][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:19:57,134][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:19:57,634][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:19:58,137][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:19:58,638][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:19:59,136][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:19:59,640][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:20:00,139][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:20:00,640][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:20:01,141][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:20:01,641][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:20:02,139][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:20:02,637][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:20:03,136][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:20:03,640][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:20:04,140][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:20:04,639][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:20:05,140][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:20:05,639][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:20:06,141][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:20:06,639][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:20:07,137][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:20:07,642][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:20:08,138][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:20:08,634][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:20:09,131][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:20:09,625][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:20:10,130][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:20:10,625][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:20:11,118][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:20:11,614][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:20:12,110][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:20:12,605][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:20:13,102][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:20:13,597][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:20:14,097][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:20:14,594][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:20:15,091][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:20:15,589][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:20:16,085][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:20:16,583][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:20:17,076][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:20:17,573][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:20:18,070][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:20:18,567][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:20:19,064][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:20:19,564][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:20:20,060][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:20:20,555][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:20:21,053][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:20:21,551][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:20:22,047][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:20:22,544][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:20:23,038][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:20:23,547][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:20:24,043][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:20:24,541][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10863 tokens. -[2025-11-09 19:20:25,157][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.26%, Current % of VRAM taken: 55.76%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 -[2025-11-09 19:20:25,900][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:20:25,902][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:20:25,904][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:20:27,234][__main__][INFO] - Iteration 93 took 49s (27.55% Gen, 69.78% Train). Generation: 13s, Training: 34s. Estimated remaining time: 40h 11m 28s. Estimated total time: 41h 32m 58s. Time estimates for 10 more iterations: 8m 18s, 100 more iterations: 1h 23m 5s, 500 more iterations: 6h 55m 29s. -[2025-11-09 19:20:27,236][__main__][INFO] - Starting iteration 93. -[2025-11-09 19:20:27,713][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 9 and human policies 1. -[2025-11-09 19:20:27,714][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:20:29,651][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:20:40,808][__main__][INFO] - Number of regex retries in iteration 93: 1 -[2025-11-09 19:20:40,809][__main__][INFO] - agents played in iteration 93 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:20:41,641][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:20:41,663][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:20:41,686][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:20:41,708][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:20:41,709][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:20:41,709][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:20:42,388][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:20:42,851][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:20:43,357][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:20:43,856][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:20:44,358][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:20:44,859][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:20:45,363][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:20:45,868][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:20:46,366][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:20:46,867][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:20:47,365][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:20:47,862][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:20:48,362][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:20:48,859][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:20:49,356][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:20:49,853][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:20:50,352][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:20:50,853][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:20:51,351][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:20:51,850][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:20:52,349][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:20:52,847][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:20:53,348][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:20:53,848][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:20:54,347][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:20:54,846][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:20:55,353][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:20:55,853][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:20:56,354][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:20:56,853][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:20:57,349][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:20:57,859][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:20:58,362][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:20:58,861][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:20:59,357][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:20:59,854][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:21:00,352][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:21:00,850][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:21:01,347][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:21:01,848][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:21:02,345][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:21:02,840][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:21:03,335][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:21:03,829][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:21:04,338][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:21:04,833][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:21:05,328][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:21:05,825][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:21:06,319][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:21:06,823][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:21:07,319][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:21:07,814][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:21:08,324][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:21:08,819][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:21:09,316][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:21:09,827][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:21:10,322][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:21:10,821][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:21:11,316][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:21:11,812][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:21:12,308][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:21:12,806][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:21:13,299][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:21:13,793][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:21:14,289][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10827 tokens. -[2025-11-09 19:21:14,937][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 55.65%, Block Peak % of device VRAM: 62.27%, ΔTime: 00:00:32 -[2025-11-09 19:21:15,697][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:21:15,698][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:21:15,700][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:21:16,634][__main__][INFO] - Iteration 94 took 48s (26.77% Gen, 71.32% Train). Generation: 13s, Training: 34s. Estimated remaining time: 39h 23m 42s. Estimated total time: 40h 46m 2s. Time estimates for 10 more iterations: 8m 9s, 100 more iterations: 1h 21m 32s, 500 more iterations: 6h 47m 40s. -[2025-11-09 19:21:16,636][__main__][INFO] - Starting iteration 94. -[2025-11-09 19:21:17,098][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 9 and human policies 1. -[2025-11-09 19:21:17,098][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:21:19,611][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:21:32,014][__main__][INFO] - Number of regex retries in iteration 94: 1 -[2025-11-09 19:21:32,015][__main__][INFO] - agents played in iteration 94 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:21:32,829][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:21:32,851][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:21:32,883][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:21:32,911][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:21:32,912][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:21:32,913][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:21:33,591][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:21:34,051][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:21:34,561][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:21:35,063][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:21:35,576][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:21:36,078][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:21:36,579][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:21:37,082][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:21:37,581][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:21:38,079][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:21:38,574][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:21:39,071][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:21:39,576][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:21:40,075][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:21:40,572][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:21:41,068][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:21:41,566][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:21:42,072][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:21:42,572][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:21:43,071][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:21:43,580][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:21:44,080][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:21:44,582][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:21:45,079][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:21:45,580][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:21:46,082][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:21:46,580][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:21:47,078][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:21:47,572][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:21:48,073][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:21:48,579][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:21:49,077][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:21:49,574][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:21:50,073][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:21:50,574][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:21:51,074][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:21:51,572][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:21:52,070][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:21:52,582][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:21:53,081][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:21:53,580][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:21:54,074][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:21:54,571][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:21:55,075][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:21:55,572][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:21:56,068][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:21:56,578][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:21:57,074][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:21:57,571][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:21:58,068][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:21:58,563][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:21:59,068][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:21:59,563][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:22:00,058][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:22:00,554][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:22:01,051][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:22:01,547][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:22:02,046][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:22:02,543][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:22:03,053][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:22:03,548][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:22:04,043][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:22:04,549][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:22:05,043][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:22:05,540][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10741 tokens. -[2025-11-09 19:22:06,173][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 55.66%, Block Peak % of device VRAM: 62.34%, ΔTime: 00:00:32 -[2025-11-09 19:22:06,954][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:22:06,956][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:22:06,958][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:22:07,899][__main__][INFO] - Iteration 95 took 50s (29.36% Gen, 68.78% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 56m 54s. Estimated total time: 42h 20m 5s. Time estimates for 10 more iterations: 8m 28s, 100 more iterations: 1h 24m 40s, 500 more iterations: 7h 3m 20s. -[2025-11-09 19:22:07,901][__main__][INFO] - Starting iteration 95. -[2025-11-09 19:22:08,370][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 9 and human policies 1. -[2025-11-09 19:22:08,371][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:22:23,021][__main__][INFO] - Number of regex retries in iteration 95: 0 -[2025-11-09 19:22:23,022][__main__][INFO] - agents played in iteration 95 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:22:23,857][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:22:23,879][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:22:23,902][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:22:23,925][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:22:23,925][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:22:23,926][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:22:24,614][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:22:25,074][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:22:25,580][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:22:26,082][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:22:26,593][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:22:27,095][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:22:27,597][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:22:28,097][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:22:28,597][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:22:29,097][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:22:29,594][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:22:30,093][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:22:30,596][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:22:31,093][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:22:31,593][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:22:32,088][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:22:32,588][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:22:33,089][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:22:33,586][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:22:34,081][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:22:34,578][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:22:35,075][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:22:35,572][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:22:36,069][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:22:36,568][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:22:37,074][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:22:37,572][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:22:38,074][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:22:38,572][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:22:39,070][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:22:39,578][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:22:40,077][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:22:40,579][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:22:41,080][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:22:41,577][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:22:42,073][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:22:42,572][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:22:43,071][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:22:43,582][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:22:44,077][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:22:44,573][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:22:45,072][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:22:45,568][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:22:46,065][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:22:46,560][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:22:47,057][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:22:47,556][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:22:48,053][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:22:48,548][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:22:49,048][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:22:49,545][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:22:50,043][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:22:50,538][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:22:51,033][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:22:51,530][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:22:52,026][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:22:52,521][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:22:53,016][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:22:53,511][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:22:54,004][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:22:54,497][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:22:54,992][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:22:55,490][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:22:55,984][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:22:56,478][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10777 tokens. -[2025-11-09 19:22:57,099][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 55.64%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 -[2025-11-09 19:22:57,852][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:22:57,854][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:22:57,856][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:22:58,811][__main__][INFO] - Iteration 96 took 50s (29.04% Gen, 69.06% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 38m 3s. Estimated total time: 42h 2m 5s. Time estimates for 10 more iterations: 8m 24s, 100 more iterations: 1h 24m 4s, 500 more iterations: 7h 0m 20s. -[2025-11-09 19:22:58,813][__main__][INFO] - Starting iteration 96. -[2025-11-09 19:22:59,283][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 9 and human policies 1. -[2025-11-09 19:22:59,285][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:23:13,218][__main__][INFO] - Number of regex retries in iteration 96: 0 -[2025-11-09 19:23:13,218][__main__][INFO] - agents played in iteration 96 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:23:14,010][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:23:14,038][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:23:14,064][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:23:14,087][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:23:14,088][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:23:14,089][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:23:14,740][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:23:15,210][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:23:15,715][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:23:16,218][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:23:16,721][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:23:17,222][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:23:17,725][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:23:18,224][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:23:18,729][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:23:19,230][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:23:19,731][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:23:20,231][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:23:20,733][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:23:21,233][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:23:21,735][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:23:22,232][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:23:22,732][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:23:23,232][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:23:23,730][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:23:24,229][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:23:24,728][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:23:25,227][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:23:25,729][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:23:26,228][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:23:26,727][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:23:27,224][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:23:27,722][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:23:28,221][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:23:28,719][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:23:29,218][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:23:29,717][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:23:30,215][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:23:30,714][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:23:31,217][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:23:31,717][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:23:32,217][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:23:32,717][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:23:33,216][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:23:33,715][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:23:34,213][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:23:34,711][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:23:35,210][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:23:35,710][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:23:36,209][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:23:36,707][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:23:37,203][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:23:37,698][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:23:38,195][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:23:38,694][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:23:39,191][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:23:39,689][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:23:40,185][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:23:40,681][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:23:41,185][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:23:41,681][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:23:42,178][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:23:42,677][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:23:43,178][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:23:43,674][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:23:44,171][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:23:44,667][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:23:45,165][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:23:45,664][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:23:46,160][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:23:46,656][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10855 tokens. -[2025-11-09 19:23:47,273][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 55.72%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:32 -[2025-11-09 19:23:48,041][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:23:48,043][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:23:48,045][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:23:48,982][__main__][INFO] - Iteration 97 took 49s (28.03% Gen, 70.07% Train). Generation: 13s, Training: 34s. Estimated remaining time: 40h 0m 5s. Estimated total time: 41h 24m 57s. Time estimates for 10 more iterations: 8m 16s, 100 more iterations: 1h 22m 49s, 500 more iterations: 6h 54m 9s. -[2025-11-09 19:23:48,984][__main__][INFO] - Starting iteration 97. -[2025-11-09 19:23:49,444][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 9 and human policies 1. -[2025-11-09 19:23:49,445][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:24:03,515][__main__][INFO] - Number of regex retries in iteration 97: 0 -[2025-11-09 19:24:03,516][__main__][INFO] - agents played in iteration 97 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:24:04,301][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:24:04,330][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:24:04,357][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:24:04,380][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:24:04,381][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:24:04,382][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:24:05,046][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:24:05,502][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:24:06,013][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:24:06,515][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:24:07,017][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:24:07,521][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:24:08,021][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:24:08,519][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:24:09,021][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:24:09,522][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:24:10,023][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:24:10,521][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:24:11,023][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:24:11,525][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:24:12,025][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:24:12,526][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:24:13,025][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:24:13,523][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:24:14,023][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:24:14,520][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:24:15,018][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:24:15,515][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:24:16,014][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:24:16,512][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:24:17,011][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:24:17,509][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:24:18,011][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:24:18,509][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:24:19,006][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:24:19,505][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:24:20,003][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:24:20,501][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:24:21,000][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:24:21,496][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:24:22,005][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:24:22,502][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:24:23,002][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:24:23,501][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:24:23,997][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:24:24,500][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:24:24,997][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:24:25,494][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:24:26,001][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:24:26,498][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:24:26,998][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:24:27,508][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:24:28,007][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:24:28,503][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:24:29,999][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:24:29,493][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:24:30,008][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:24:30,502][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:24:30,995][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:24:31,492][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:24:31,985][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:24:32,492][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:24:32,984][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:24:33,480][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:24:33,991][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:24:34,490][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:24:34,987][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:24:35,489][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:24:35,984][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:24:36,480][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:24:36,976][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10750 tokens. -[2025-11-09 19:24:37,595][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.25%, Current % of VRAM taken: 55.74%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 -[2025-11-09 19:24:38,361][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:24:38,363][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:24:38,365][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:24:39,318][__main__][INFO] - Iteration 98 took 49s (28.21% Gen, 69.87% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 7m 59s. Estimated total time: 41h 33m 41s. Time estimates for 10 more iterations: 8m 18s, 100 more iterations: 1h 23m 7s, 500 more iterations: 6h 55m 36s. -[2025-11-09 19:24:39,320][__main__][INFO] - Starting iteration 98. -[2025-11-09 19:24:39,786][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 9 and human policies 1. -[2025-11-09 19:24:39,787][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:24:53,807][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 ballots did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:24:54,584][__main__][INFO] - Number of regex retries in iteration 98: 1 -[2025-11-09 19:24:54,585][__main__][INFO] - agents played in iteration 98 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:24:55,360][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:24:55,387][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:24:55,414][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:24:55,437][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:24:55,437][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:24:55,439][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:24:56,108][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:24:56,576][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:24:57,083][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:24:57,598][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:24:58,101][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:24:58,603][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:24:59,105][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:24:59,607][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:25:00,108][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:25:00,610][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:25:01,111][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:25:01,619][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:25:02,118][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:25:02,618][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:25:03,120][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:25:03,618][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:25:04,122][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:25:04,619][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:25:05,118][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:25:05,612][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:25:06,109][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:25:06,613][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:25:07,110][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:25:07,610][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:25:08,109][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:25:08,605][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:25:09,102][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:25:09,598][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:25:10,093][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:25:10,607][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:25:11,104][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:25:11,606][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:25:12,103][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:25:12,601][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:25:13,110][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:25:13,609][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:25:14,113][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:25:14,614][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:25:15,113][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:25:15,611][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:25:16,109][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:25:16,608][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:25:17,118][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:25:17,617][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:25:18,116][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:25:18,614][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:25:19,113][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:25:19,617][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:25:20,119][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:25:20,620][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:25:21,127][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:25:21,623][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:25:22,125][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:25:22,624][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:25:23,124][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:25:23,631][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:25:24,132][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:25:24,627][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:25:25,122][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:25:25,616][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:25:26,116][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:25:26,612][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:25:27,110][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:25:27,607][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:25:28,101][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10777 tokens. -[2025-11-09 19:25:28,752][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 55.61%, Block Peak % of device VRAM: 62.28%, ΔTime: 00:00:32 -[2025-11-09 19:25:29,518][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:25:29,519][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:25:29,521][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:25:30,460][__main__][INFO] - Iteration 99 took 50s (29.20% Gen, 68.94% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 47m 8s. Estimated total time: 42h 13m 41s. Time estimates for 10 more iterations: 8m 26s, 100 more iterations: 1h 24m 27s, 500 more iterations: 7h 2m 16s. -[2025-11-09 19:25:30,462][__main__][INFO] - Starting iteration 99. -[2025-11-09 19:25:30,938][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 9 and human policies 1. -[2025-11-09 19:25:30,939][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:25:45,417][__main__][INFO] - Number of regex retries in iteration 99: 0 -[2025-11-09 19:25:45,418][__main__][INFO] - agents played in iteration 99 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:25:46,210][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:25:46,232][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:25:46,254][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:25:46,276][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:25:46,277][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:25:46,278][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:25:46,882][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:25:47,346][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:25:47,853][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:25:48,363][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:25:48,870][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:25:49,374][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:25:49,880][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:25:50,382][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:25:50,884][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:25:51,389][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:25:51,888][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:25:52,389][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:25:52,892][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:25:53,393][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:25:53,905][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:25:54,405][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:25:54,903][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:25:55,404][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:25:55,906][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:25:56,419][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:25:56,919][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:25:57,416][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:25:57,913][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:25:58,412][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:25:58,925][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:25:59,424][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:25:59,923][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:26:00,422][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:26:00,920][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:26:01,419][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:26:01,920][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:26:02,419][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:26:02,927][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:26:03,428][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:26:03,928][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:26:04,430][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:26:04,929][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:26:05,429][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:26:05,929][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:26:06,431][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:26:06,941][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:26:07,437][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:26:07,934][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:26:08,441][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:26:08,938][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:26:09,438][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:26:09,935][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:26:10,433][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:26:10,941][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:26:11,439][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:26:11,938][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:26:12,438][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:26:12,939][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:26:13,442][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:26:13,939][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:26:14,436][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:26:14,943][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:26:15,437][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:26:15,930][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:26:16,430][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:26:16,926][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:26:17,431][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:26:17,927][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:26:18,424][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:26:18,924][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10836 tokens. -[2025-11-09 19:26:19,536][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 55.72%, Block Peak % of device VRAM: 62.42%, ΔTime: 00:00:32 -[2025-11-09 19:26:20,296][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:26:20,298][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:26:20,299][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:26:21,224][__main__][INFO] - Iteration 100 took 50s (28.79% Gen, 69.36% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 26m 55s. Estimated total time: 41h 54m 19s. Time estimates for 10 more iterations: 8m 22s, 100 more iterations: 1h 23m 48s, 500 more iterations: 6h 59m 3s. -[2025-11-09 19:26:21,226][__main__][INFO] - Starting iteration 100. -[2025-11-09 19:26:21,694][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 9 and human policies 1. -[2025-11-09 19:26:21,695][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:26:35,477][__main__][INFO] - Number of regex retries in iteration 100: 0 -[2025-11-09 19:26:35,477][__main__][INFO] - agents played in iteration 100 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:26:36,253][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:26:36,277][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:26:36,302][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:26:36,324][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:26:36,325][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:26:36,326][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:26:36,944][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:26:37,412][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:26:37,912][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:26:38,417][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:26:38,917][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:26:39,417][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:26:39,936][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:26:40,436][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:26:40,935][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:26:41,437][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:26:41,936][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:26:42,444][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:26:42,941][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:26:43,439][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:26:43,951][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:26:44,449][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:26:44,946][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:26:45,452][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:26:45,954][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:26:46,455][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:26:46,956][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:26:47,456][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:26:47,955][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:26:48,455][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:26:48,957][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:26:49,458][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:26:49,958][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:26:50,459][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:26:50,956][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:26:51,454][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:26:51,955][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:26:52,453][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:26:52,953][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:26:53,447][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:26:53,942][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:26:54,444][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:26:54,939][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:26:55,438][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:26:55,951][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:26:56,447][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:26:56,946][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:26:57,445][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:26:57,946][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:26:58,453][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:26:58,952][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:26:59,452][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:26:59,950][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:27:00,447][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:27:00,951][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:27:01,447][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:27:01,943][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:27:02,442][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:27:02,937][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:27:03,433][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:27:03,932][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:27:04,430][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:27:04,938][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:27:05,441][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:27:05,938][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:27:06,434][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:27:06,927][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:27:07,422][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:27:07,915][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:27:08,410][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:27:08,906][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10804 tokens. -[2025-11-09 19:27:09,531][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 55.62%, Block Peak % of device VRAM: 62.28%, ΔTime: 00:00:32 -[2025-11-09 19:27:10,309][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:27:10,311][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:27:10,312][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:27:12,778][__main__][INFO] - Iteration 101 took 51s (26.98% Gen, 68.19% Train). Generation: 13s, Training: 34s. Estimated remaining time: 41h 5m 57s. Estimated total time: 42h 34m 13s. Time estimates for 10 more iterations: 8m 30s, 100 more iterations: 1h 25m 8s, 500 more iterations: 7h 5m 42s. -[2025-11-09 19:27:12,780][__main__][INFO] - Starting iteration 101. -[2025-11-09 19:27:13,248][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 10 and human policies 1. -[2025-11-09 19:27:13,249][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:27:16,238][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:27:28,806][__main__][INFO] - Number of regex retries in iteration 101: 1 -[2025-11-09 19:27:28,807][__main__][INFO] - agents played in iteration 101 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:27:29,686][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:27:29,710][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:27:29,733][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:27:29,755][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:27:29,755][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:27:29,756][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:27:30,362][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:27:30,823][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:27:31,334][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:27:31,836][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:27:32,341][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:27:32,848][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:27:33,348][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:27:33,859][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:27:34,358][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:27:34,861][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:27:35,358][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:27:35,860][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:27:36,361][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:27:36,858][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:27:37,357][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:27:37,868][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:27:38,366][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:27:38,865][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:27:39,367][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:27:39,871][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:27:40,375][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:27:40,879][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:27:41,379][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:27:41,881][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:27:42,386][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:27:42,886][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:27:43,385][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:27:43,886][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:27:44,383][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:27:44,883][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:27:45,381][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:27:45,878][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:27:46,378][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:27:46,876][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:27:47,375][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:27:47,878][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:27:48,378][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:27:48,876][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:27:49,376][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:27:49,874][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:27:50,376][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:27:50,876][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:27:51,375][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:27:51,875][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:27:52,374][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:27:52,873][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:27:53,372][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:27:53,871][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:27:54,370][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:27:54,899][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:27:55,401][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:27:55,901][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:27:56,404][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:27:56,903][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:27:57,401][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:27:57,904][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:27:58,406][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:27:58,905][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:27:59,405][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:27:59,901][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:28:00,399][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:28:00,896][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:28:01,390][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:28:01,885][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:28:02,380][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10814 tokens. -[2025-11-09 19:28:03,037][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 55.69%, Block Peak % of device VRAM: 62.40%, ΔTime: 00:00:32 -[2025-11-09 19:28:03,803][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:28:03,804][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:28:03,806][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:28:04,812][__main__][INFO] - Iteration 102 took 51s (30.17% Gen, 67.87% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 29m 5s. Estimated total time: 42h 58m 12s. Time estimates for 10 more iterations: 8m 35s, 100 more iterations: 1h 25m 56s, 500 more iterations: 7h 9m 42s. -[2025-11-09 19:28:04,814][__main__][INFO] - Starting iteration 102. -[2025-11-09 19:28:05,294][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 10 and human policies 1. -[2025-11-09 19:28:05,295][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:28:08,985][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 30 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:28:12,153][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:28:19,491][__main__][INFO] - Number of regex retries in iteration 102: 2 -[2025-11-09 19:28:19,492][__main__][INFO] - agents played in iteration 102 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:28:20,270][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:28:20,293][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:28:20,315][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:28:20,337][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:28:20,337][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:28:20,338][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:28:20,992][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:28:21,446][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:28:21,950][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:28:22,449][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:28:22,950][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:28:23,451][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:28:23,953][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:28:24,452][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:28:24,951][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:28:25,449][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:28:25,945][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:28:26,463][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:28:26,965][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:28:27,461][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:28:27,958][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:28:28,454][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:28:28,960][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:28:29,461][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:28:29,957][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:28:30,456][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:28:30,954][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:28:31,454][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:28:31,954][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:28:32,454][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:28:32,957][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:28:33,486][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:28:33,986][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:28:34,488][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:28:34,989][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:28:35,491][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:28:35,993][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:28:36,490][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:28:36,994][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:28:37,496][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:28:37,996][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:28:38,497][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:28:38,995][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:28:39,495][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:28:39,995][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:28:40,498][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:28:40,999][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:28:41,499][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:28:41,996][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:28:42,494][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:28:42,993][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:28:43,493][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:28:43,993][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:28:44,493][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:28:44,992][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:28:45,490][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:28:45,987][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:28:46,485][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:28:46,985][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:28:47,491][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:28:47,995][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:28:48,494][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:28:48,991][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:28:49,489][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:28:49,997][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:28:50,498][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:28:50,995][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:28:51,494][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:28:51,989][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:28:52,488][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:28:52,985][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10802 tokens. -[2025-11-09 19:28:53,626][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 55.72%, Block Peak % of device VRAM: 62.35%, ΔTime: 00:00:32 -[2025-11-09 19:28:54,385][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:28:54,386][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:28:54,388][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:28:55,331][__main__][INFO] - Iteration 103 took 50s (28.37% Gen, 69.74% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 11m 52s. Estimated total time: 41h 41m 50s. Time estimates for 10 more iterations: 8m 20s, 100 more iterations: 1h 23m 23s, 500 more iterations: 6h 56m 58s. -[2025-11-09 19:28:55,333][__main__][INFO] - Starting iteration 103. -[2025-11-09 19:28:55,800][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 10 and human policies 1. -[2025-11-09 19:28:55,800][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:29:03,335][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:29:10,389][__main__][INFO] - Number of regex retries in iteration 103: 1 -[2025-11-09 19:29:10,390][__main__][INFO] - agents played in iteration 103 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:29:11,188][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:29:11,216][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:29:11,242][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:29:11,264][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:29:11,265][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:29:11,266][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:29:11,891][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:29:12,364][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:29:12,870][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:29:13,372][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:29:13,876][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:29:14,376][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:29:14,881][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:29:15,381][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:29:15,881][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:29:16,386][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:29:16,891][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:29:17,391][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:29:17,892][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:29:18,392][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:29:18,891][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:29:19,390][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:29:19,889][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:29:20,387][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:29:20,885][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:29:21,388][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:29:21,885][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:29:22,388][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:29:22,886][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:29:23,388][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:29:23,889][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:29:24,390][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:29:24,890][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:29:25,395][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:29:25,897][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:29:26,397][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:29:26,896][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:29:27,394][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:29:27,894][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:29:28,411][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:29:28,911][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:29:29,411][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:29:29,912][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:29:30,411][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:29:30,910][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:29:31,408][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:29:31,908][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:29:32,409][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:29:32,911][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:29:33,411][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:29:33,912][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:29:34,412][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:29:34,910][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:29:35,409][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:29:35,910][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:29:36,410][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:29:36,909][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:29:37,408][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:29:37,906][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:29:38,408][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:29:38,906][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:29:39,407][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:29:39,910][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:29:40,404][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:29:40,903][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:29:41,399][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:29:41,898][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:29:42,400][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:29:42,897][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:29:43,393][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:29:43,891][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10848 tokens. -[2025-11-09 19:29:44,514][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 55.70%, Block Peak % of device VRAM: 62.42%, ΔTime: 00:00:32 -[2025-11-09 19:29:45,291][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:29:45,293][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:29:45,295][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:29:46,236][__main__][INFO] - Iteration 104 took 50s (28.93% Gen, 69.21% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 31m 0s. Estimated total time: 42h 1m 49s. Time estimates for 10 more iterations: 8m 24s, 100 more iterations: 1h 24m 3s, 500 more iterations: 7h 0m 18s. -[2025-11-09 19:29:46,238][__main__][INFO] - Starting iteration 104. -[2025-11-09 19:29:46,706][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 10 and human policies 1. -[2025-11-09 19:29:46,706][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:30:01,820][__main__][INFO] - Number of regex retries in iteration 104: 0 -[2025-11-09 19:30:01,821][__main__][INFO] - agents played in iteration 104 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:30:02,624][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:30:02,647][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:30:02,669][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:30:02,690][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:30:02,691][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:30:02,692][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:30:03,341][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:30:03,800][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:30:04,306][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:30:04,802][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:30:05,305][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:30:05,805][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:30:06,308][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:30:06,810][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:30:07,308][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:30:07,808][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:30:08,308][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:30:08,811][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:30:09,312][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:30:09,814][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:30:10,311][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:30:10,808][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:30:11,304][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:30:11,801][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:30:12,297][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:30:12,792][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:30:13,286][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:30:13,784][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:30:14,285][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:30:14,785][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:30:15,280][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:30:15,784][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:30:16,285][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:30:16,796][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:30:17,297][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:30:17,798][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:30:18,330][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:30:18,840][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:30:19,342][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:30:19,841][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:30:20,338][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:30:20,838][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:30:21,337][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:30:21,833][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:30:22,332][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:30:22,832][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:30:23,331][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:30:23,834][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:30:24,331][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:30:24,830][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:30:25,329][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:30:25,827][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:30:26,324][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:30:26,822][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:30:27,320][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:30:27,815][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:30:28,312][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:30:28,807][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:30:29,300][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:30:29,803][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:30:30,297][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:30:30,794][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:30:31,293][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:30:31,792][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:30:32,289][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:30:32,782][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:30:33,278][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:30:33,786][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:30:34,285][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:30:34,781][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:30:35,283][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10762 tokens. -[2025-11-09 19:30:35,907][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.52%, ΔTime: 00:00:32 -[2025-11-09 19:30:36,686][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:30:36,688][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:30:36,690][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:30:37,626][__main__][INFO] - Iteration 105 took 50s (29.68% Gen, 68.48% Train). Generation: 15s, Training: 34s. Estimated remaining time: 40h 54m 19s. Estimated total time: 42h 26m 0s. Time estimates for 10 more iterations: 8m 29s, 100 more iterations: 1h 24m 52s, 500 more iterations: 7h 4m 20s. -[2025-11-09 19:30:37,628][__main__][INFO] - Starting iteration 105. -[2025-11-09 19:30:38,095][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 10 and human policies 1. -[2025-11-09 19:30:38,096][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:30:52,573][__main__][INFO] - Number of regex retries in iteration 105: 0 -[2025-11-09 19:30:52,573][__main__][INFO] - agents played in iteration 105 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:30:53,382][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.67%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:30:53,409][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.67%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:30:53,435][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.67%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:30:53,458][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.67%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:30:53,458][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:30:53,459][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:30:54,114][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:30:54,570][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:30:55,073][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:30:55,573][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:30:56,067][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:30:56,569][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:30:57,070][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:30:57,584][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:30:58,092][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:30:58,605][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:30:59,116][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:30:59,616][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:31:00,117][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:31:00,616][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:31:01,116][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:31:01,618][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:31:02,114][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:31:02,611][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:31:03,111][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:31:03,608][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:31:04,107][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:31:04,605][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:31:05,102][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:31:05,603][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:31:06,101][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:31:06,599][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:31:07,098][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:31:07,596][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:31:08,131][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:31:08,635][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:31:09,135][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:31:09,636][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:31:10,140][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:31:10,640][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:31:11,143][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:31:11,645][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:31:12,148][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:31:12,653][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:31:13,153][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:31:13,654][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:31:14,154][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:31:14,653][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:31:15,153][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:31:15,650][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:31:16,151][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:31:16,648][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:31:17,144][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:31:17,639][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:31:18,135][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:31:18,632][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:31:19,131][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:31:19,629][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:31:20,127][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:31:20,625][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:31:21,122][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:31:21,620][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:31:22,118][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:31:22,619][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:31:23,117][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:31:23,614][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:31:24,114][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:31:24,613][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:31:25,110][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:31:25,607][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:31:26,102][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10804 tokens. -[2025-11-09 19:31:26,786][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 55.69%, Block Peak % of device VRAM: 62.35%, ΔTime: 00:00:32 -[2025-11-09 19:31:27,551][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:31:27,553][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:31:27,555][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:31:28,503][__main__][INFO] - Iteration 106 took 50s (28.72% Gen, 69.39% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 27m 54s. Estimated total time: 42h 0m 25s. Time estimates for 10 more iterations: 8m 24s, 100 more iterations: 1h 24m 0s, 500 more iterations: 7h 0m 4s. -[2025-11-09 19:31:28,505][__main__][INFO] - Starting iteration 106. -[2025-11-09 19:31:28,969][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 10 and human policies 1. -[2025-11-09 19:31:28,969][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:31:31,829][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:31:43,177][__main__][INFO] - Number of regex retries in iteration 106: 1 -[2025-11-09 19:31:43,177][__main__][INFO] - agents played in iteration 106 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:31:44,018][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:31:44,046][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:31:44,074][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:31:44,097][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:31:44,098][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:31:44,098][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:31:44,728][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:31:45,188][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:31:45,692][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:31:46,190][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:31:46,688][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:31:47,185][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:31:47,697][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:31:48,197][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:31:48,699][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:31:49,198][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:31:49,695][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:31:50,196][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:31:50,697][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:31:51,196][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:31:51,709][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:31:52,208][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:31:52,706][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:31:53,203][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:31:53,700][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:31:54,210][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:31:54,707][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:31:55,205][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:31:55,704][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:31:56,200][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:31:56,715][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:31:57,213][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:31:57,710][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:31:58,218][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:31:58,716][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:31:59,214][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:31:59,716][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:32:00,215][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:32:00,722][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:32:01,218][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:32:01,719][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:32:02,216][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:32:02,716][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:32:03,222][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:32:03,721][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:32:04,224][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:32:04,726][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:32:05,234][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:32:05,735][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:32:06,235][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:32:06,734][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:32:07,231][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:32:07,724][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:32:08,223][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:32:08,719][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:32:09,218][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:32:09,722][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:32:10,220][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:32:10,718][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:32:11,226][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:32:11,723][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:32:12,220][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:32:12,720][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:32:13,219][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:32:13,719][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:32:14,216][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:32:14,713][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:32:15,221][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:32:15,718][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:32:16,218][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:32:16,718][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10836 tokens. -[2025-11-09 19:32:17,391][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 55.69%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 -[2025-11-09 19:32:18,162][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:32:18,163][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:32:18,166][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:32:19,291][__main__][INFO] - Iteration 107 took 50s (28.23% Gen, 69.53% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 22m 45s. Estimated total time: 41h 56m 7s. Time estimates for 10 more iterations: 8m 23s, 100 more iterations: 1h 23m 52s, 500 more iterations: 6h 59m 21s. -[2025-11-09 19:32:19,293][__main__][INFO] - Starting iteration 107. -[2025-11-09 19:32:19,809][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 10 and human policies 1. -[2025-11-09 19:32:19,810][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:32:22,681][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:32:34,402][__main__][INFO] - Number of regex retries in iteration 107: 1 -[2025-11-09 19:32:34,402][__main__][INFO] - agents played in iteration 107 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:32:35,180][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:32:35,207][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:32:35,233][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:32:35,255][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:32:35,256][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:32:35,256][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:32:35,856][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:32:36,311][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:32:36,819][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:32:37,319][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:32:37,816][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:32:38,311][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:32:38,809][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:32:39,316][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:32:39,816][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:32:40,336][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:32:40,835][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:32:41,335][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:32:41,837][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:32:42,337][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:32:42,841][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:32:43,340][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:32:43,839][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:32:44,356][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:32:44,865][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:32:45,365][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:32:45,862][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:32:46,360][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:32:46,863][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:32:47,361][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:32:47,858][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:32:48,356][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:32:48,853][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:32:49,351][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:32:49,849][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:32:50,345][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:32:50,844][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:32:51,343][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:32:51,841][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:32:52,347][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:32:52,848][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:32:53,348][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:32:53,845][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:32:54,344][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:32:54,846][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:32:55,348][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:32:55,849][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:32:56,347][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:32:56,845][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:32:57,346][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:32:57,846][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:32:58,346][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:32:58,845][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:32:59,345][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:32:59,846][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:33:00,344][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:33:00,842][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:33:01,341][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:33:01,840][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:33:02,338][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:33:02,842][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:33:03,340][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:33:03,840][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:33:04,342][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:33:04,844][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:33:05,347][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:33:05,852][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:33:06,353][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:33:06,858][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:33:07,360][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:33:07,864][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10861 tokens. -[2025-11-09 19:33:08,538][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.30%, Current % of VRAM taken: 55.80%, Block Peak % of device VRAM: 62.40%, ΔTime: 00:00:32 -[2025-11-09 19:33:09,306][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:33:09,308][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:33:09,310][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:33:10,184][__main__][INFO] - Iteration 108 took 50s (28.97% Gen, 69.30% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 24m 31s. Estimated total time: 41h 58m 45s. Time estimates for 10 more iterations: 8m 23s, 100 more iterations: 1h 23m 57s, 500 more iterations: 6h 59m 47s. -[2025-11-09 19:33:10,186][__main__][INFO] - Starting iteration 108. -[2025-11-09 19:33:10,664][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 10 and human policies 1. -[2025-11-09 19:33:10,665][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:33:13,460][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:33:17,590][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:33:24,323][__main__][INFO] - Number of regex retries in iteration 108: 2 -[2025-11-09 19:33:24,324][__main__][INFO] - agents played in iteration 108 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:33:25,115][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:33:25,141][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:33:25,164][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:33:25,187][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:33:25,187][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:33:25,188][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:33:25,789][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:33:26,245][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:33:26,742][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:33:27,238][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:33:27,737][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:33:28,236][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:33:28,733][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:33:29,238][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:33:29,734][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:33:30,234][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:33:30,736][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:33:31,234][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:33:31,742][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:33:32,240][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:33:32,737][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:33:33,238][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:33:33,737][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:33:34,240][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:33:34,738][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:33:35,235][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:33:35,735][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:33:36,232][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:33:36,725][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:33:37,224][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:33:37,720][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:33:38,238][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:33:38,736][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:33:39,233][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:33:39,733][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:33:40,232][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:33:40,736][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:33:41,235][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:33:41,735][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:33:42,233][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:33:42,732][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:33:43,229][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:33:43,730][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:33:44,230][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:33:44,743][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:33:45,243][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:33:45,746][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:33:46,243][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:33:46,745][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:33:47,253][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:33:47,757][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:33:48,258][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:33:48,757][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:33:49,255][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:33:49,753][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:33:50,250][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:33:50,745][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:33:51,243][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:33:51,739][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:33:52,235][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:33:52,734][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:33:53,231][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:33:53,728][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:33:54,229][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:33:54,727][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:33:55,224][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:33:55,723][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:33:56,220][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:33:56,727][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:33:57,232][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:33:57,731][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10863 tokens. -[2025-11-09 19:33:58,414][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:32 -[2025-11-09 19:33:59,168][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:33:59,170][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:33:59,173][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:34:00,110][__main__][INFO] - Iteration 109 took 49s (27.62% Gen, 70.48% Train). Generation: 13s, Training: 34s. Estimated remaining time: 39h 37m 14s. Estimated total time: 41h 12m 18s. Time estimates for 10 more iterations: 8m 14s, 100 more iterations: 1h 22m 24s, 500 more iterations: 6h 52m 3s. -[2025-11-09 19:34:00,112][__main__][INFO] - Starting iteration 109. -[2025-11-09 19:34:00,586][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 10 and human policies 1. -[2025-11-09 19:34:00,587][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:34:14,678][__main__][INFO] - Number of regex retries in iteration 109: 0 -[2025-11-09 19:34:14,679][__main__][INFO] - agents played in iteration 109 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:34:15,447][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:34:15,474][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:34:15,502][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:34:15,526][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:34:15,526][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:34:15,527][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:34:16,138][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:34:16,594][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:34:17,095][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:34:17,594][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:34:18,091][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:34:18,593][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:34:19,090][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:34:19,587][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:34:20,080][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:34:20,578][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:34:21,084][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:34:21,585][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:34:22,086][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:34:22,590][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:34:23,089][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:34:23,589][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:34:24,092][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:34:24,593][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:34:25,093][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:34:25,592][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:34:26,090][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:34:26,586][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:34:27,080][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:34:27,577][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:34:28,075][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:34:28,573][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:34:29,075][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:34:29,577][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:34:30,075][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:34:30,577][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:34:31,075][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:34:31,573][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:34:32,072][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:34:32,570][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:34:33,070][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:34:33,569][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:34:34,067][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:34:34,568][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:34:35,065][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:34:35,567][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:34:36,068][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:34:36,567][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:34:37,070][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:34:37,571][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:34:38,071][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:34:38,572][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:34:39,073][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:34:39,571][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:34:40,070][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:34:40,566][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:34:41,062][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:34:41,557][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:34:42,054][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:34:42,550][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:34:43,044][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:34:43,537][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:34:44,034][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:34:44,534][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:34:45,042][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:34:45,540][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:34:46,037][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:34:46,540][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:34:47,039][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:34:47,540][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:34:48,040][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10809 tokens. -[2025-11-09 19:34:48,730][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 55.70%, Block Peak % of device VRAM: 62.34%, ΔTime: 00:00:32 -[2025-11-09 19:34:49,495][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:34:49,497][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:34:49,499][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:34:50,528][__main__][INFO] - Iteration 110 took 49s (28.21% Gen, 69.72% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 1m 12s. Estimated total time: 41h 37m 6s. Time estimates for 10 more iterations: 8m 19s, 100 more iterations: 1h 23m 14s, 500 more iterations: 6h 56m 11s. -[2025-11-09 19:34:50,530][__main__][INFO] - Starting iteration 110. -[2025-11-09 19:34:50,996][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 10 and human policies 1. -[2025-11-09 19:34:50,997][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:34:55,544][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:34:59,681][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:35:04,739][__main__][INFO] - Number of regex retries in iteration 110: 2 -[2025-11-09 19:35:04,739][__main__][INFO] - agents played in iteration 110 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:35:05,505][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:35:05,534][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:35:05,560][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:35:05,583][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:35:05,583][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:35:05,584][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:35:06,188][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:35:06,644][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:35:07,146][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:35:07,645][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:35:08,143][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:35:08,648][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:35:09,144][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:35:09,640][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:35:10,140][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:35:10,637][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:35:11,136][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:35:11,636][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:35:12,134][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:35:12,642][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:35:13,139][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:35:13,638][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:35:14,139][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:35:14,638][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:35:15,136][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:35:15,635][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:35:16,133][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:35:16,643][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:35:17,140][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:35:17,640][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:35:18,141][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:35:18,640][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:35:19,149][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:35:19,645][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:35:20,142][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:35:20,641][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:35:21,138][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:35:21,638][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:35:22,134][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:35:22,631][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:35:23,128][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:35:23,625][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:35:24,121][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:35:24,618][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:35:25,115][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:35:25,612][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:35:26,113][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:35:26,611][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:35:27,113][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:35:27,612][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:35:28,110][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:35:28,616][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:35:29,115][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:35:29,611][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:35:30,119][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:35:30,617][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:35:31,118][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:35:31,616][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:35:32,113][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:35:32,616][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:35:33,114][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:35:33,610][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:35:34,111][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:35:34,612][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:35:35,116][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:35:35,617][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:35:36,118][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:35:36,627][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:35:37,127][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:35:37,644][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:35:38,143][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10845 tokens. -[2025-11-09 19:35:38,805][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.27%, Current % of VRAM taken: 55.76%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:32 -[2025-11-09 19:35:39,565][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:35:39,566][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:35:39,568][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:35:41,486][__main__][INFO] - Iteration 111 took 50s (27.22% Gen, 68.98% Train). Generation: 13s, Training: 34s. Estimated remaining time: 40h 27m 46s. Estimated total time: 42h 4m 30s. Time estimates for 10 more iterations: 8m 24s, 100 more iterations: 1h 24m 9s, 500 more iterations: 7h 0m 45s. -[2025-11-09 19:35:41,489][__main__][INFO] - Starting iteration 111. -[2025-11-09 19:35:41,958][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 11 and human policies 1. -[2025-11-09 19:35:41,959][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:35:45,004][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:35:56,373][__main__][INFO] - Number of regex retries in iteration 111: 1 -[2025-11-09 19:35:56,374][__main__][INFO] - agents played in iteration 111 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:35:57,139][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:35:57,167][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:35:57,193][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:35:57,216][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:35:57,216][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:35:57,217][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:35:57,821][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:35:58,274][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:35:58,776][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:35:59,275][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:35:59,775][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:36:00,276][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:36:00,774][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:36:01,277][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:36:01,778][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:36:02,282][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:36:02,828][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:36:03,325][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:36:03,828][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:36:04,341][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:36:04,843][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:36:05,355][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:36:05,855][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:36:06,352][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:36:06,864][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:36:07,362][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:36:07,861][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:36:08,357][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:36:08,853][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:36:09,359][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:36:09,856][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:36:10,352][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:36:10,849][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:36:11,345][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:36:11,841][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:36:12,338][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:36:12,835][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:36:13,342][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:36:13,839][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:36:14,336][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:36:14,839][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:36:15,338][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:36:15,836][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:36:16,334][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:36:16,833][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:36:17,344][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:36:17,839][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:36:18,336][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:36:18,842][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:36:19,341][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:36:19,840][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:36:20,338][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:36:20,838][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:36:21,337][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:36:21,836][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:36:22,337][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:36:22,847][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:36:23,345][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:36:23,868][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:36:24,371][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:36:24,873][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:36:25,373][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:36:25,873][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:36:26,376][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:36:26,872][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:36:27,366][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:36:27,860][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:36:28,358][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:36:28,855][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:36:29,356][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:36:29,855][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10838 tokens. -[2025-11-09 19:36:30,539][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 55.66%, Block Peak % of device VRAM: 62.41%, ΔTime: 00:00:32 -[2025-11-09 19:36:31,274][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:36:31,275][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:36:31,277][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:36:32,169][__main__][INFO] - Iteration 112 took 50s (28.71% Gen, 69.51% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 12m 58s. Estimated total time: 41h 50m 33s. Time estimates for 10 more iterations: 8m 22s, 100 more iterations: 1h 23m 41s, 500 more iterations: 6h 58m 25s. -[2025-11-09 19:36:32,171][__main__][INFO] - Starting iteration 112. -[2025-11-09 19:36:32,641][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 11 and human policies 1. -[2025-11-09 19:36:32,642][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:36:36,064][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:36:36,952][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:36:41,417][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:36:45,243][__main__][INFO] - Number of regex retries in iteration 112: 3 -[2025-11-09 19:36:45,244][__main__][INFO] - agents played in iteration 112 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:36:46,202][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:36:46,224][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:36:46,247][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:36:46,269][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:36:46,270][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:36:46,270][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:36:46,941][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:36:47,400][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:36:47,902][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:36:48,400][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:36:48,898][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:36:49,395][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:36:49,894][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:36:50,395][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:36:50,893][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:36:51,391][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:36:51,892][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:36:52,390][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:36:52,897][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:36:53,396][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:36:53,895][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:36:54,396][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:36:54,897][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:36:55,398][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:36:55,896][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:36:56,395][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:36:56,897][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:36:57,393][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:36:57,891][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:36:58,390][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:36:58,888][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:36:59,388][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:36:59,887][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:37:00,386][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:37:00,889][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:37:01,387][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:37:01,884][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:37:02,380][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:37:02,879][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:37:03,378][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:37:03,879][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:37:04,380][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:37:04,878][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:37:05,375][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:37:05,874][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:37:06,373][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:37:06,874][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:37:07,379][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:37:07,876][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:37:08,374][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:37:08,888][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:37:09,388][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:37:09,888][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:37:10,383][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:37:10,879][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:37:11,387][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:37:11,887][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:37:12,388][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:37:12,884][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:37:13,383][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:37:13,888][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:37:14,386][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:37:14,887][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:37:15,393][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:37:15,893][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:37:16,396][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:37:16,897][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:37:17,396][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:37:17,899][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:37:18,397][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:37:18,894][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10865 tokens. -[2025-11-09 19:37:19,607][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 55.72%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:32 -[2025-11-09 19:37:20,383][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:37:20,385][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:37:20,386][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:37:21,339][__main__][INFO] - Iteration 113 took 48s (25.88% Gen, 72.16% Train). Generation: 12s, Training: 35s. Estimated remaining time: 38h 56m 30s. Estimated total time: 40h 34m 54s. Time estimates for 10 more iterations: 8m 6s, 100 more iterations: 1h 21m 9s, 500 more iterations: 6h 45m 49s. -[2025-11-09 19:37:21,341][__main__][INFO] - Starting iteration 113. -[2025-11-09 19:37:21,890][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 11 and human policies 1. -[2025-11-09 19:37:21,890][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:37:35,547][__main__][INFO] - Number of regex retries in iteration 113: 0 -[2025-11-09 19:37:35,548][__main__][INFO] - agents played in iteration 113 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:37:36,351][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:37:36,378][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:37:36,404][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:37:36,427][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:37:36,428][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:37:36,429][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:37:37,030][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:37:37,479][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:37:37,981][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:37:38,474][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:37:38,985][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:37:39,481][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:37:39,976][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:37:40,475][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:37:40,972][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:37:41,471][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:37:41,975][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:37:42,474][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:37:42,986][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:37:43,486][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:37:43,985][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:37:44,496][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:37:44,997][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:37:45,499][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:37:46,001][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:37:46,500][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:37:47,001][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:37:47,498][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:37:47,998][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:37:48,498][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:37:48,994][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:37:49,502][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:37:50,015][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:37:50,515][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:37:51,016][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:37:51,514][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:37:52,013][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:37:52,512][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:37:53,011][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:37:53,508][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:37:54,002][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:37:54,497][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:37:55,001][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:37:55,495][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:37:55,992][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:37:56,488][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:37:56,984][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:37:57,493][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:37:57,991][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:37:58,488][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:37:59,002][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:37:59,500][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:38:00,004][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:38:00,501][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:38:00,998][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:38:01,504][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:38:02,000][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:38:02,498][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:38:02,994][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:38:03,492][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:38:03,996][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:38:04,492][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:38:04,989][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:38:05,504][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:38:06,005][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:38:06,506][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:38:07,007][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:38:07,507][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:38:08,006][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:38:08,504][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:38:09,004][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10851 tokens. -[2025-11-09 19:38:09,708][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 -[2025-11-09 19:38:10,436][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:38:10,437][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:38:10,439][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:38:11,405][__main__][INFO] - Iteration 114 took 49s (27.58% Gen, 70.46% Train). Generation: 13s, Training: 34s. Estimated remaining time: 39h 36m 31s. Estimated total time: 41h 15m 46s. Time estimates for 10 more iterations: 8m 15s, 100 more iterations: 1h 22m 31s, 500 more iterations: 6h 52m 37s. -[2025-11-09 19:38:11,407][__main__][INFO] - Starting iteration 114. -[2025-11-09 19:38:11,885][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 11 and human policies 1. -[2025-11-09 19:38:11,886][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:38:26,555][__main__][INFO] - Number of regex retries in iteration 114: 0 -[2025-11-09 19:38:26,556][__main__][INFO] - agents played in iteration 114 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:38:27,324][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:38:27,347][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:38:27,369][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:38:27,391][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:38:27,392][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:38:27,393][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:38:27,995][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:38:28,449][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:38:28,947][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:38:29,444][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:38:29,941][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:38:30,442][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:38:30,938][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:38:31,434][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:38:31,930][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:38:32,427][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:38:32,935][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:38:33,434][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:38:33,931][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:38:34,430][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:38:34,929][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:38:35,426][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:38:35,923][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:38:36,421][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:38:36,924][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:38:37,421][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:38:37,919][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:38:38,417][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:38:38,913][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:38:39,408][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:38:39,904][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:38:40,422][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:38:40,922][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:38:41,419][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:38:41,919][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:38:42,414][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:38:42,912][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:38:43,415][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:38:43,913][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:38:44,414][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:38:44,915][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:38:45,413][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:38:45,913][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:38:46,412][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:38:46,907][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:38:47,414][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:38:47,912][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:38:48,414][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:38:48,909][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:38:49,407][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:38:49,913][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:38:50,412][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:38:50,912][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:38:51,413][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:38:51,910][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:38:52,408][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:38:52,903][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:38:53,399][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:38:53,896][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:38:54,391][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:38:54,885][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:38:55,403][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:38:55,898][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:38:56,395][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:38:56,895][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:38:57,391][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:38:57,905][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:38:58,406][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:38:58,908][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:38:59,413][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:38:59,916][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10793 tokens. -[2025-11-09 19:39:00,607][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.27%, Current % of VRAM taken: 55.77%, Block Peak % of device VRAM: 62.42%, ΔTime: 00:00:32 -[2025-11-09 19:39:01,331][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:39:01,332][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:39:01,334][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:39:02,390][__main__][INFO] - Iteration 115 took 50s (29.05% Gen, 68.86% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 25m 11s. Estimated total time: 42h 5m 17s. Time estimates for 10 more iterations: 8m 25s, 100 more iterations: 1h 24m 10s, 500 more iterations: 7h 0m 52s. -[2025-11-09 19:39:02,392][__main__][INFO] - Starting iteration 115. -[2025-11-09 19:39:02,973][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 11 and human policies 1. -[2025-11-09 19:39:02,973][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:39:14,804][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:39:17,245][__main__][INFO] - Number of regex retries in iteration 115: 1 -[2025-11-09 19:39:17,245][__main__][INFO] - agents played in iteration 115 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:39:18,028][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:39:18,055][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:39:18,082][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:39:18,105][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:39:18,106][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:39:18,106][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:39:18,725][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:39:19,181][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:39:19,683][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:39:20,180][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:39:20,677][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:39:21,174][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:39:21,673][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:39:22,173][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:39:22,674][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:39:23,177][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:39:23,679][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:39:24,182][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:39:24,683][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:39:25,184][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:39:25,684][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:39:26,185][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:39:26,692][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:39:27,193][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:39:27,704][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:39:28,203][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:39:28,702][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:39:29,200][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:39:29,697][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:39:30,194][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:39:30,692][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:39:31,190][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:39:31,712][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:39:32,210][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:39:32,707][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:39:33,207][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:39:33,705][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:39:34,213][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:39:34,710][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:39:35,207][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:39:35,721][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:39:36,217][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:39:36,715][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:39:37,214][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:39:37,711][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:39:38,218][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:39:38,716][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:39:39,215][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:39:39,726][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:39:40,224][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:39:40,721][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:39:41,220][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:39:41,720][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:39:42,222][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:39:42,721][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:39:43,217][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:39:43,718][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:39:44,215][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:39:44,717][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:39:45,214][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:39:45,709][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:39:46,223][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:39:46,721][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:39:47,222][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:39:47,721][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:39:48,221][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:39:48,731][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:39:49,229][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:39:49,729][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:39:50,230][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:39:50,730][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10871 tokens. -[2025-11-09 19:39:51,435][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 55.66%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 -[2025-11-09 19:39:52,197][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:39:52,198][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:39:52,204][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:39:53,154][__main__][INFO] - Iteration 116 took 50s (28.44% Gen, 69.67% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 8m 7s. Estimated total time: 41h 49m 4s. Time estimates for 10 more iterations: 8m 21s, 100 more iterations: 1h 23m 38s, 500 more iterations: 6h 58m 10s. -[2025-11-09 19:39:53,156][__main__][INFO] - Starting iteration 116. -[2025-11-09 19:39:53,652][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 11 and human policies 1. -[2025-11-09 19:39:53,653][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:40:08,470][__main__][INFO] - Number of regex retries in iteration 116: 0 -[2025-11-09 19:40:08,471][__main__][INFO] - agents played in iteration 116 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:40:09,254][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:40:09,282][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:40:09,309][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:40:09,332][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:40:09,333][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:40:09,334][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:40:10,017][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:40:10,476][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:40:10,980][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:40:11,480][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:40:11,976][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:40:12,476][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:40:12,976][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:40:13,474][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:40:13,971][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:40:14,473][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:40:14,974][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:40:15,475][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:40:15,975][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:40:16,476][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:40:16,974][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:40:17,475][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:40:17,975][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:40:18,475][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:40:18,977][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:40:19,500][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:40:19,997][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:40:20,496][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:40:20,995][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:40:21,492][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:40:21,989][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:40:22,486][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:40:22,995][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:40:23,490][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:40:23,988][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:40:24,486][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:40:24,983][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:40:25,483][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:40:25,979][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:40:26,477][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:40:26,979][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:40:27,478][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:40:27,977][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:40:28,475][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:40:28,975][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:40:29,475][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:40:29,976][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:40:30,475][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:40:30,973][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:40:31,471][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:40:31,969][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:40:32,468][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:40:32,971][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:40:33,472][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:40:33,969][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:40:34,465][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:40:34,963][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:40:35,458][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:40:35,958][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:40:36,458][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:40:36,958][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:40:37,458][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:40:37,958][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:40:38,458][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:40:38,954][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:40:39,452][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:40:39,951][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:40:40,446][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:40:40,941][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:40:41,436][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:40:41,931][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10857 tokens. -[2025-11-09 19:40:42,585][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 55.65%, Block Peak % of device VRAM: 62.24%, ΔTime: 00:00:32 -[2025-11-09 19:40:43,331][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:40:43,333][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:40:43,335][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:40:44,313][__main__][INFO] - Iteration 117 took 50s (29.25% Gen, 68.82% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 31m 17s. Estimated total time: 42h 13m 4s. Time estimates for 10 more iterations: 8m 26s, 100 more iterations: 1h 24m 26s, 500 more iterations: 7h 2m 10s. -[2025-11-09 19:40:44,315][__main__][INFO] - Starting iteration 117. -[2025-11-09 19:40:44,797][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 11 and human policies 1. -[2025-11-09 19:40:44,797][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:40:47,788][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:40:59,241][__main__][INFO] - Number of regex retries in iteration 117: 1 -[2025-11-09 19:40:59,242][__main__][INFO] - agents played in iteration 117 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:41:00,036][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:41:00,063][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:41:00,089][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:41:00,112][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:41:00,112][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:41:00,114][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:41:00,739][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:41:01,203][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:41:01,708][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:41:02,218][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:41:02,793][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:41:03,292][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:41:03,787][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:41:04,285][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:41:04,781][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:41:05,281][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:41:05,781][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:41:06,280][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:41:06,780][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:41:07,279][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:41:07,776][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:41:08,277][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:41:08,776][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:41:09,274][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:41:09,779][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:41:10,296][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:41:10,793][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:41:11,291][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:41:11,792][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:41:12,289][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:41:12,787][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:41:13,284][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:41:13,781][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:41:14,277][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:41:14,774][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:41:15,271][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:41:15,768][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:41:16,264][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:41:16,763][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:41:17,258][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:41:17,756][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:41:18,252][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:41:18,750][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:41:19,251][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:41:19,749][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:41:20,247][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:41:20,753][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:41:21,250][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:41:21,746][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:41:22,243][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:41:22,742][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:41:23,252][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:41:23,751][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:41:24,247][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:41:24,742][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:41:25,241][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:41:25,738][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:41:26,237][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:41:26,737][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:41:27,250][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:41:27,748][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:41:28,245][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:41:28,742][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:41:29,241][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:41:29,750][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:41:30,247][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:41:30,741][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:41:31,239][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:41:31,740][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:41:32,236][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:41:32,737][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10832 tokens. -[2025-11-09 19:41:33,373][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.18%, ΔTime: 00:00:32 -[2025-11-09 19:41:34,122][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:41:34,124][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:41:34,126][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:41:35,049][__main__][INFO] - Iteration 118 took 50s (28.74% Gen, 69.42% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 9m 59s. Estimated total time: 41h 52m 37s. Time estimates for 10 more iterations: 8m 22s, 100 more iterations: 1h 23m 45s, 500 more iterations: 6h 58m 46s. -[2025-11-09 19:41:35,051][__main__][INFO] - Starting iteration 118. -[2025-11-09 19:41:35,511][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 11 and human policies 1. -[2025-11-09 19:41:35,512][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:41:43,273][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:41:49,026][__main__][INFO] - Number of regex retries in iteration 118: 1 -[2025-11-09 19:41:49,027][__main__][INFO] - agents played in iteration 118 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:41:49,810][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:41:49,835][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:41:49,861][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:41:49,883][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:41:49,883][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:41:49,885][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:41:50,498][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:41:50,952][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:41:51,459][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:41:51,957][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:41:52,456][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:41:52,955][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:41:53,471][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:41:53,969][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:41:54,469][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:41:54,969][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:41:55,472][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:41:55,976][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:41:56,477][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:41:56,979][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:41:57,478][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:41:57,978][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:41:58,480][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:41:58,982][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:41:59,482][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:41:59,984][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:42:00,480][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:42:00,981][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:42:01,483][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:42:01,979][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:42:02,478][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:42:02,977][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:42:03,474][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:42:03,979][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:42:04,475][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:42:04,972][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:42:05,466][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:42:05,962][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:42:06,465][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:42:06,962][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:42:07,459][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:42:07,974][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:42:08,473][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:42:08,967][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:42:09,466][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:42:09,962][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:42:10,477][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:42:10,975][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:42:11,499][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:42:11,999][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:42:12,500][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:42:13,003][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:42:13,499][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:42:13,997][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:42:14,498][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:42:14,998][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:42:15,497][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:42:15,998][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:42:16,494][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:42:16,993][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:42:17,491][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:42:17,990][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:42:18,486][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:42:18,986][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:42:19,485][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:42:19,983][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:42:20,482][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:42:20,987][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:42:21,489][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:42:21,988][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:42:22,482][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10857 tokens. -[2025-11-09 19:42:23,122][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 55.65%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 -[2025-11-09 19:42:23,887][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:42:23,889][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:42:23,890][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:42:24,829][__main__][INFO] - Iteration 119 took 49s (27.40% Gen, 70.69% Train). Generation: 13s, Training: 34s. Estimated remaining time: 39h 22m 25s. Estimated total time: 41h 5m 53s. Time estimates for 10 more iterations: 8m 13s, 100 more iterations: 1h 22m 11s, 500 more iterations: 6h 50m 58s. -[2025-11-09 19:42:24,831][__main__][INFO] - Starting iteration 119. -[2025-11-09 19:42:25,312][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 11 and human policies 1. -[2025-11-09 19:42:25,313][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:42:33,245][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:42:36,824][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:42:39,567][__main__][INFO] - Number of regex retries in iteration 119: 2 -[2025-11-09 19:42:39,567][__main__][INFO] - agents played in iteration 119 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:42:40,345][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:42:40,370][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:42:40,395][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:42:40,417][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:42:40,418][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:42:40,418][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:42:41,026][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:42:41,481][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:42:41,985][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:42:42,483][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:42:42,981][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:42:43,486][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:42:43,983][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:42:44,482][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:42:44,978][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:42:45,479][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:42:45,980][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:42:46,481][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:42:46,981][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:42:47,480][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:42:47,975][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:42:48,479][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:42:48,978][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:42:49,498][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:42:50,007][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:42:50,502][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:42:51,006][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:42:51,505][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:42:52,005][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:42:52,505][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:42:53,005][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:42:53,507][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:42:54,008][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:42:54,506][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:42:55,007][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:42:55,504][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:42:55,999][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:42:56,497][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:42:56,997][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:42:57,495][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:42:57,994][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:42:58,493][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:42:58,990][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:42:59,488][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:42:59,987][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:43:00,503][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:43:01,001][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:43:01,498][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:43:01,997][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:43:02,495][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:43:02,997][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:43:03,499][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:43:03,998][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:43:04,505][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:43:04,998][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:43:05,498][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:43:05,996][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:43:06,494][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:43:07,000][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:43:07,498][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:43:07,994][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:43:08,501][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:43:08,999][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:43:09,499][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:43:10,001][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:43:10,498][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:43:11,003][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:43:11,500][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:43:11,998][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:43:12,499][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:43:12,996][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10868 tokens. -[2025-11-09 19:43:13,645][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 55.70%, Block Peak % of device VRAM: 62.20%, ΔTime: 00:00:32 -[2025-11-09 19:43:14,397][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:43:14,399][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:43:14,400][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:43:15,339][__main__][INFO] - Iteration 120 took 50s (28.49% Gen, 69.63% Train). Generation: 14s, Training: 34s. Estimated remaining time: 39h 57m 3s. Estimated total time: 41h 41m 21s. Time estimates for 10 more iterations: 8m 20s, 100 more iterations: 1h 23m 22s, 500 more iterations: 6h 56m 53s. -[2025-11-09 19:43:15,341][__main__][INFO] - Starting iteration 120. -[2025-11-09 19:43:15,816][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 11 and human policies 1. -[2025-11-09 19:43:15,816][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:43:18,329][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:43:28,481][__main__][INFO] - Number of regex retries in iteration 120: 1 -[2025-11-09 19:43:28,481][__main__][INFO] - agents played in iteration 120 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:43:29,255][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:43:29,283][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:43:29,320][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:43:29,347][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:43:29,348][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:43:29,348][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:43:29,967][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:43:30,427][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:43:30,934][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:43:31,438][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:43:31,939][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:43:32,443][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:43:32,940][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:43:33,443][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:43:33,939][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:43:34,438][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:43:34,940][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:43:35,444][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:43:35,949][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:43:36,453][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:43:36,959][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:43:37,463][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:43:37,963][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:43:38,459][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:43:38,953][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:43:39,450][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:43:39,947][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:43:40,441][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:43:40,935][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:43:41,429][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:43:41,926][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:43:42,428][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:43:42,927][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:43:43,427][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:43:43,930][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:43:44,431][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:43:44,929][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:43:45,431][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:43:45,933][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:43:46,432][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:43:46,933][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:43:47,435][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:43:47,936][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:43:48,435][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:43:48,934][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:43:49,435][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:43:49,936][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:43:50,437][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:43:50,935][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:43:51,438][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:43:51,939][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:43:52,440][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:43:52,940][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:43:53,439][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:43:53,937][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:43:54,434][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:43:54,930][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:43:55,426][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:43:55,921][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:43:56,416][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:43:56,914][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:43:57,413][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:43:57,911][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:43:58,419][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:43:58,919][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:43:59,419][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:43:59,917][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:44:00,417][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:44:00,916][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:44:01,415][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:44:01,916][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10855 tokens. -[2025-11-09 19:44:02,614][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.53%, ΔTime: 00:00:32 -[2025-11-09 19:44:03,338][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:44:03,339][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:44:03,342][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:44:05,131][__main__][INFO] - Iteration 121 took 49s (25.68% Gen, 70.69% Train). Generation: 12s, Training: 34s. Estimated remaining time: 39h 20m 39s. Estimated total time: 41h 5m 47s. Time estimates for 10 more iterations: 8m 13s, 100 more iterations: 1h 22m 11s, 500 more iterations: 6h 50m 57s. -[2025-11-09 19:44:05,133][__main__][INFO] - Starting iteration 121. -[2025-11-09 19:44:05,619][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 12 and human policies 1. -[2025-11-09 19:44:05,620][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:44:12,315][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:44:20,238][__main__][INFO] - Number of regex retries in iteration 121: 1 -[2025-11-09 19:44:20,238][__main__][INFO] - agents played in iteration 121 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:44:21,014][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:44:21,039][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:44:21,064][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:44:21,086][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:44:21,087][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:44:21,088][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:44:21,711][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:44:22,171][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:44:22,675][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:44:23,174][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:44:23,671][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:44:24,173][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:44:24,689][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:44:25,194][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:44:25,699][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:44:26,199][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:44:26,698][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:44:27,203][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:44:27,703][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:44:28,208][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:44:28,716][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:44:29,220][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:44:29,719][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:44:30,221][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:44:30,720][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:44:31,224][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:44:31,724][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:44:32,224][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:44:32,724][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:44:33,226][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:44:33,724][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:44:34,222][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:44:34,720][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:44:35,221][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:44:35,718][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:44:36,216][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:44:36,715][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:44:37,210][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:44:37,707][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:44:38,206][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:44:38,704][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:44:39,213][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:44:39,709][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:44:40,206][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:44:40,705][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:44:41,202][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:44:41,701][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:44:42,200][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:44:42,698][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:44:43,198][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:44:43,696][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:44:44,197][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:44:44,702][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:44:45,201][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:44:45,705][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:44:46,203][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:44:46,699][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:44:47,198][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:44:47,695][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:44:48,193][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:44:48,690][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:44:49,192][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:44:49,708][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:44:50,211][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:44:50,711][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:44:51,216][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:44:51,722][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:44:52,224][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:44:52,727][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:44:53,226][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:44:53,727][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10861 tokens. -[2025-11-09 19:44:54,380][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.33%, ΔTime: 00:00:32 -[2025-11-09 19:44:55,132][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:44:55,134][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:44:55,135][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:44:56,071][__main__][INFO] - Iteration 122 took 50s (28.98% Gen, 69.17% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 16m 36s. Estimated total time: 42h 2m 35s. Time estimates for 10 more iterations: 8m 24s, 100 more iterations: 1h 24m 5s, 500 more iterations: 7h 0m 25s. -[2025-11-09 19:44:56,073][__main__][INFO] - Starting iteration 122. -[2025-11-09 19:44:56,535][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 12 and human policies 1. -[2025-11-09 19:44:56,536][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:45:03,029][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:45:04,942][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:45:10,769][__main__][INFO] - Number of regex retries in iteration 122: 2 -[2025-11-09 19:45:10,770][__main__][INFO] - agents played in iteration 122 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:45:11,540][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:45:11,567][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:45:11,593][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:45:11,616][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:45:11,616][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:45:11,618][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:45:12,240][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:45:12,694][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:45:13,199][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:45:13,697][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:45:14,195][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:45:14,696][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:45:15,195][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:45:15,694][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:45:16,197][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:45:16,696][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:45:17,212][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:45:17,713][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:45:18,212][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:45:18,714][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:45:19,217][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:45:19,717][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:45:20,219][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:45:20,718][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:45:21,229][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:45:21,732][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:45:22,234][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:45:22,736][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:45:23,236][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:45:23,734][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:45:24,231][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:45:24,729][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:45:25,262][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:45:25,760][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:45:26,259][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:45:26,757][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:45:27,256][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:45:27,753][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:45:28,253][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:45:28,752][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:45:29,252][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:45:29,753][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:45:30,253][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:45:30,753][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:45:31,252][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:45:31,752][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:45:32,252][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:45:32,752][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:45:33,255][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:45:33,753][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:45:34,252][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:45:34,752][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:45:35,253][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:45:35,753][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:45:36,251][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:45:36,748][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:45:37,250][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:45:37,749][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:45:38,248][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:45:38,750][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:45:39,250][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:45:39,749][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:45:40,252][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:45:40,753][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:45:41,257][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:45:41,758][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:45:42,259][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:45:42,760][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:45:43,260][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:45:43,759][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:45:44,257][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10867 tokens. -[2025-11-09 19:45:44,910][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.26%, Current % of VRAM taken: 55.76%, Block Peak % of device VRAM: 62.39%, ΔTime: 00:00:32 -[2025-11-09 19:45:45,652][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:45:45,654][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:45:45,656][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:45:46,581][__main__][INFO] - Iteration 123 took 50s (28.44% Gen, 69.71% Train). Generation: 14s, Training: 34s. Estimated remaining time: 39h 55m 29s. Estimated total time: 41h 42m 18s. Time estimates for 10 more iterations: 8m 20s, 100 more iterations: 1h 23m 24s, 500 more iterations: 6h 57m 3s. -[2025-11-09 19:45:46,583][__main__][INFO] - Starting iteration 123. -[2025-11-09 19:45:47,048][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 12 and human policies 1. -[2025-11-09 19:45:47,049][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:45:57,381][mllm.models.large_language_model_local][WARNING] - Response Given the per-item values, the optimal strategy would be to propose a distribution that maximizes your value while ensuring that the total quantities of proposed items do not exceed the available items. +Given the per-item values, I (Alice) have a significantly higher value for hats compared to Bob. Books and balls have a lower value for me, which, combined with Bob's higher values for these items, make them less desirable to claim fully. By proposing to take all 10 hats, I hope to maximize my points for this round, as the allocation will be proportional if my proposals exceed the item quantity or fully allocated if they do not. did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:16:37,602][__main__][INFO] - Number of regex retries in iteration 81: 2 +[2026-03-25 22:16:37,603][__main__][INFO] - agents played in iteration 81 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:16:38,503][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:16:38,530][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:16:38,557][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:16:38,584][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:16:38,585][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:16:38,586][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:16:39,417][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:16:39,886][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:16:40,408][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:16:40,920][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:16:41,431][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:16:41,943][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:16:42,454][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:16:42,967][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:16:43,480][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:16:43,991][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:16:44,501][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:16:45,017][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:16:45,526][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:16:46,046][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:16:46,561][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:16:47,072][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:16:47,587][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:16:48,099][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:16:48,614][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:16:49,126][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:16:49,647][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:16:50,161][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:16:50,671][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:16:51,181][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:16:51,697][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:16:52,208][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:16:52,721][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:16:53,232][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:16:53,744][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:16:54,256][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:16:54,769][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:16:55,280][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:16:55,790][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:16:56,303][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:16:56,835][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:16:57,353][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:16:57,867][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:16:58,382][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:16:58,891][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:16:59,400][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:16:59,922][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:17:00,437][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:17:00,952][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:17:01,469][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:17:01,982][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:17:02,495][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:17:03,010][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:17:03,524][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:17:04,042][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:17:04,555][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:17:05,065][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:17:05,574][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:17:06,088][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:17:06,601][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:17:07,112][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:17:07,626][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:17:08,141][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:17:08,652][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:17:09,166][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:17:09,681][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:17:10,196][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:17:10,709][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:17:11,223][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:17:11,735][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:17:12,250][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10298 tokens. +[2026-03-25 22:17:13,128][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 60.57%, Block Peak % of device VRAM: 62.07%, ΔTime: 00:00:33 +[2026-03-25 22:17:13,937][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:17:13,939][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:17:13,941][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:17:14,681][__main__][INFO] - Iteration 82 took 53s (30.73% Gen, 67.89% Train). Generation: 16s, Training: 36s. Estimated remaining time: 43h 24m 44s. Estimated total time: 44h 36m 18s. Time estimates for 10 more iterations: 8m 55s, 100 more iterations: 1h 29m 12s, 500 more iterations: 7h 26m 3s. +[2026-03-25 22:17:14,684][__main__][INFO] - Starting iteration 82. +[2026-03-25 22:17:15,096][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 8 and human policies 1. +[2026-03-25 22:17:15,097][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:17:32,046][__main__][INFO] - Number of regex retries in iteration 82: 0 +[2026-03-25 22:17:32,047][__main__][INFO] - agents played in iteration 82 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:17:32,997][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:17:33,023][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:17:33,049][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:17:33,073][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:17:33,075][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:17:33,075][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:17:33,876][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:17:34,346][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:17:34,861][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:17:35,371][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:17:35,884][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:17:36,397][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:17:36,911][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:17:37,423][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:17:37,934][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:17:38,446][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:17:38,961][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:17:39,478][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:17:39,988][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:17:40,502][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:17:41,012][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:17:41,524][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:17:42,041][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:17:42,562][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:17:43,073][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:17:43,589][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:17:44,098][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:17:44,612][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:17:45,129][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:17:45,643][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:17:46,159][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:17:46,673][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:17:47,189][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:17:47,699][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:17:48,208][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:17:48,720][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:17:49,230][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:17:49,739][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:17:50,250][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:17:50,761][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:17:51,274][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:17:51,786][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:17:52,296][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:17:52,808][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:17:53,322][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:17:53,832][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:17:54,343][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:17:54,855][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:17:55,368][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:17:55,881][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:17:56,394][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:17:56,908][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:17:57,422][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:17:57,933][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:17:58,448][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:17:58,961][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:17:59,479][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:17:59,998][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:18:00,509][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:18:01,022][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:18:01,533][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:18:02,047][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:18:02,559][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:18:03,070][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:18:03,585][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:18:04,096][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:18:04,610][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:18:05,124][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:18:05,634][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:18:06,146][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:18:06,659][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10472 tokens. +[2026-03-25 22:18:07,536][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:33 +[2026-03-25 22:18:08,338][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:18:08,340][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:18:08,342][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:18:09,080][__main__][INFO] - Iteration 83 took 53s (31.40% Gen, 67.23% Train). Generation: 16s, Training: 36s. Estimated remaining time: 43h 46m 47s. Estimated total time: 44h 59m 16s. Time estimates for 10 more iterations: 8m 59s, 100 more iterations: 1h 29m 58s, 500 more iterations: 7h 29m 52s. +[2026-03-25 22:18:09,083][__main__][INFO] - Starting iteration 83. +[2026-03-25 22:18:09,498][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 8 and human policies 1. +[2026-03-25 22:18:09,498][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:18:22,262][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:18:25,909][__main__][INFO] - Number of regex retries in iteration 83: 1 +[2026-03-25 22:18:25,910][__main__][INFO] - agents played in iteration 83 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:18:26,793][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:18:26,818][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:18:26,843][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:18:26,867][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:18:26,868][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:18:26,869][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:18:27,639][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:18:28,106][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:18:28,622][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:18:29,139][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:18:29,650][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:18:30,174][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:18:30,682][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:18:31,190][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:18:31,699][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:18:32,206][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:18:32,715][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:18:33,224][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:18:33,733][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:18:34,246][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:18:34,756][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:18:35,273][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:18:35,782][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:18:36,296][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:18:36,809][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:18:37,324][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:18:37,839][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:18:38,349][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:18:38,862][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:18:39,370][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:18:39,886][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:18:40,394][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:18:40,907][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:18:41,415][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:18:41,924][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:18:42,441][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:18:42,951][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:18:43,461][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:18:43,973][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:18:44,478][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:18:44,985][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:18:45,496][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:18:46,005][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:18:46,512][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:18:47,019][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:18:47,528][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:18:48,036][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:18:48,552][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:18:49,059][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:18:49,568][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:18:50,080][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:18:50,591][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:18:51,098][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:18:51,605][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:18:52,115][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:18:52,624][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:18:53,135][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:18:53,644][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:18:54,153][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:18:54,658][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:18:55,165][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:18:55,672][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:18:56,178][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:18:56,689][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:18:57,197][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:18:57,706][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:18:58,212][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:18:58,720][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:18:59,227][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:18:59,732][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:19:00,238][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10290 tokens. +[2026-03-25 22:19:01,006][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 60.53%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:33 +[2026-03-25 22:19:01,773][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:19:01,775][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:19:01,778][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:19:02,477][__main__][INFO] - Iteration 84 took 52s (30.98% Gen, 67.70% Train). Generation: 16s, Training: 35s. Estimated remaining time: 42h 55m 38s. Estimated total time: 44h 9m 0s. Time estimates for 10 more iterations: 8m 49s, 100 more iterations: 1h 28m 18s, 500 more iterations: 7h 21m 30s. +[2026-03-25 22:19:02,480][__main__][INFO] - Starting iteration 84. +[2026-03-25 22:19:02,884][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 8 and human policies 1. +[2026-03-25 22:19:02,884][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:19:13,345][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:19:18,632][__main__][INFO] - Number of regex retries in iteration 84: 1 +[2026-03-25 22:19:18,633][__main__][INFO] - agents played in iteration 84 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:19:19,488][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:19:19,512][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:19:19,539][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:19:19,564][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:19:19,565][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:19:19,566][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:19:20,266][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:19:20,730][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:19:21,245][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:19:21,749][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:19:22,260][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:19:22,766][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:19:23,271][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:19:23,782][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:19:24,286][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:19:24,792][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:19:25,297][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:19:25,802][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:19:26,308][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:19:26,813][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:19:27,320][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:19:27,825][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:19:28,332][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:19:28,865][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:19:29,375][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:19:29,885][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:19:30,396][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:19:30,906][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:19:31,413][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:19:31,922][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:19:32,430][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:19:32,935][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:19:33,445][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:19:33,958][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:19:34,468][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:19:34,977][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:19:35,487][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:19:35,996][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:19:36,504][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:19:37,012][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:19:37,523][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:19:38,030][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:19:38,538][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:19:39,044][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:19:39,557][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:19:40,064][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:19:40,571][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:19:41,079][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:19:41,584][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:19:42,090][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:19:42,598][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:19:43,106][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:19:43,616][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:19:44,122][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:19:44,629][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:19:45,133][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:19:45,642][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:19:46,146][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:19:46,649][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:19:47,152][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:19:47,655][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:19:48,158][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:19:48,661][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:19:49,166][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:19:49,668][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:19:50,170][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:19:50,672][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:19:51,178][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:19:51,681][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:19:52,187][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:19:52,689][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10369 tokens. +[2026-03-25 22:19:53,381][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.21%, ΔTime: 00:00:33 +[2026-03-25 22:19:54,115][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:19:54,117][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:19:54,119][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:19:54,805][__main__][INFO] - Iteration 85 took 51s (30.33% Gen, 68.35% Train). Generation: 15s, Training: 35s. Estimated remaining time: 42h 1m 49s. Estimated total time: 43h 16m 4s. Time estimates for 10 more iterations: 8m 39s, 100 more iterations: 1h 26m 32s, 500 more iterations: 7h 12m 40s. +[2026-03-25 22:19:54,807][__main__][INFO] - Starting iteration 85. +[2026-03-25 22:19:55,211][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 8 and human policies 1. +[2026-03-25 22:19:55,212][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:20:11,164][__main__][INFO] - Number of regex retries in iteration 85: 0 +[2026-03-25 22:20:11,165][__main__][INFO] - agents played in iteration 85 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:20:11,979][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:20:12,003][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:20:12,028][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:20:12,053][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:20:12,055][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:20:12,055][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:20:12,715][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:20:13,180][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:20:13,692][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:20:14,198][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:20:14,708][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:20:15,215][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:20:15,722][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:20:16,225][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:20:16,727][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:20:17,232][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:20:17,729][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:20:18,228][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:20:18,728][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:20:19,228][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:20:19,727][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:20:20,226][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:20:20,726][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:20:21,222][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:20:21,720][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:20:22,216][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:20:22,712][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:20:23,210][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:20:23,708][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:20:24,204][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:20:24,702][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:20:25,202][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:20:25,699][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:20:26,197][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:20:26,701][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:20:27,199][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:20:27,697][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:20:28,195][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:20:28,693][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:20:29,195][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:20:29,697][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:20:30,199][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:20:30,700][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:20:31,199][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:20:31,700][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:20:32,223][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:20:32,724][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:20:33,223][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:20:33,723][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:20:34,223][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:20:34,720][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:20:35,220][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:20:35,719][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:20:36,219][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:20:36,719][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:20:37,220][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:20:37,720][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:20:38,222][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:20:38,721][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:20:39,221][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:20:39,723][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:20:40,222][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:20:40,723][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:20:41,219][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:20:41,717][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:20:42,216][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:20:42,714][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:20:43,215][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:20:43,715][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:20:44,214][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:20:44,713][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10465 tokens. +[2026-03-25 22:20:45,367][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 +[2026-03-25 22:20:46,126][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:20:46,128][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:20:46,130][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:20:46,808][__main__][INFO] - Iteration 86 took 51s (30.92% Gen, 67.76% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 44m 46s. Estimated total time: 42h 59m 52s. Time estimates for 10 more iterations: 8m 35s, 100 more iterations: 1h 25m 59s, 500 more iterations: 7h 9m 58s. +[2026-03-25 22:20:46,811][__main__][INFO] - Starting iteration 86. +[2026-03-25 22:20:47,211][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 8 and human policies 1. +[2026-03-25 22:20:47,211][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:21:02,949][__main__][INFO] - Number of regex retries in iteration 86: 0 +[2026-03-25 22:21:02,950][__main__][INFO] - agents played in iteration 86 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:21:03,724][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:21:03,747][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:21:03,771][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:21:03,794][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:21:03,794][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:21:03,795][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:21:04,417][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:21:04,871][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:21:05,376][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:21:05,874][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:21:06,372][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:21:06,870][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:21:07,367][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:21:07,865][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:21:08,364][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:21:08,865][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:21:09,362][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:21:09,859][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:21:10,360][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:21:10,857][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:21:11,356][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:21:11,855][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:21:12,354][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:21:12,854][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:21:13,353][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:21:13,852][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:21:14,349][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:21:14,846][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:21:15,345][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:21:15,844][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:21:16,343][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:21:16,843][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:21:17,343][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:21:17,842][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:21:18,338][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:21:18,837][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:21:19,340][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:21:19,839][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:21:20,339][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:21:20,839][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:21:21,337][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:21:21,833][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:21:22,333][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:21:22,832][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:21:23,330][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:21:23,828][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:21:24,323][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:21:24,823][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:21:25,323][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:21:25,822][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:21:26,320][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:21:26,820][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:21:27,319][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:21:27,817][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:21:28,317][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:21:28,814][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:21:29,314][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:21:29,814][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:21:30,310][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:21:30,807][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:21:31,305][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:21:31,803][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:21:32,300][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:21:32,799][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:21:33,298][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:21:33,797][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:21:34,296][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:21:34,795][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:21:35,290][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:21:35,789][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:21:36,288][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10371 tokens. +[2026-03-25 22:21:36,945][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 60.62%, Block Peak % of device VRAM: 62.18%, ΔTime: 00:00:32 +[2026-03-25 22:21:37,662][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:21:37,664][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:21:37,666][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:21:38,351][__main__][INFO] - Iteration 87 took 51s (30.78% Gen, 67.88% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 21m 5s. Estimated total time: 42h 37m 3s. Time estimates for 10 more iterations: 8m 31s, 100 more iterations: 1h 25m 14s, 500 more iterations: 7h 6m 10s. +[2026-03-25 22:21:38,354][__main__][INFO] - Starting iteration 87. +[2026-03-25 22:21:38,756][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 8 and human policies 1. +[2026-03-25 22:21:38,757][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:21:41,196][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:21:50,068][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:21:54,301][__main__][INFO] - Number of regex retries in iteration 87: 2 +[2026-03-25 22:21:54,302][__main__][INFO] - agents played in iteration 87 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:21:55,067][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:21:55,090][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:21:55,113][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:21:55,136][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:21:55,137][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:21:55,138][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:21:55,785][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:21:56,245][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:21:56,751][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:21:57,248][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:21:57,754][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:21:58,250][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:21:58,752][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:21:59,248][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:21:59,744][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:22:00,244][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:22:00,746][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:22:01,248][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:22:01,749][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:22:02,249][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:22:02,748][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:22:03,247][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:22:03,747][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:22:04,248][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:22:04,750][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:22:05,250][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:22:05,750][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:22:06,249][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:22:06,751][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:22:07,251][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:22:07,751][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:22:08,248][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:22:08,749][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:22:09,249][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:22:09,749][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:22:10,249][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:22:10,750][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:22:11,249][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:22:11,749][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:22:12,250][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:22:12,748][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:22:13,248][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:22:13,749][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:22:14,249][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:22:14,748][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:22:15,248][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:22:15,748][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:22:16,244][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:22:16,742][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:22:17,241][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:22:17,740][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:22:18,238][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:22:18,736][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:22:19,234][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:22:19,731][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:22:20,228][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:22:20,725][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:22:21,224][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:22:21,723][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:22:22,222][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:22:22,722][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:22:23,221][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:22:23,718][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:22:24,218][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:22:24,717][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:22:25,217][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:22:25,715][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:22:26,216][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:22:26,714][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:22:27,214][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:22:27,714][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10446 tokens. +[2026-03-25 22:22:28,391][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.15%, ΔTime: 00:00:32 +[2026-03-25 22:22:29,108][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:22:29,111][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:22:29,112][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:22:29,789][__main__][INFO] - Iteration 88 took 51s (30.46% Gen, 68.21% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 14m 50s. Estimated total time: 42h 31m 39s. Time estimates for 10 more iterations: 8m 30s, 100 more iterations: 1h 25m 3s, 500 more iterations: 7h 5m 16s. +[2026-03-25 22:22:29,792][__main__][INFO] - Starting iteration 88. +[2026-03-25 22:22:30,193][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 8 and human policies 1. +[2026-03-25 22:22:30,194][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:22:45,431][__main__][INFO] - Number of regex retries in iteration 88: 0 +[2026-03-25 22:22:45,432][__main__][INFO] - agents played in iteration 88 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:22:46,196][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:22:46,219][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:22:46,243][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:22:46,266][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:22:46,266][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:22:46,267][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:22:46,899][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:22:47,351][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:22:47,857][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:22:48,353][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:22:48,855][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:22:49,355][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:22:49,851][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:22:50,348][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:22:50,845][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:22:51,346][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:22:51,846][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:22:52,350][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:22:52,850][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:22:53,369][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:22:53,872][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:22:54,369][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:22:54,870][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:22:55,369][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:22:55,870][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:22:56,370][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:22:56,869][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:22:57,371][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:22:57,872][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:22:58,372][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:22:58,872][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:22:59,368][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:22:59,867][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:23:00,365][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:23:00,861][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:23:01,359][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:23:01,858][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:23:02,358][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:23:02,854][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:23:03,354][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:23:03,854][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:23:04,354][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:23:04,854][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:23:05,355][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:23:05,856][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:23:06,356][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:23:06,855][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:23:07,354][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:23:07,854][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:23:08,355][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:23:08,854][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:23:09,354][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:23:09,854][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:23:10,352][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:23:10,850][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:23:11,348][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:23:11,846][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:23:12,342][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:23:12,840][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:23:13,341][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:23:13,841][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:23:14,339][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:23:14,839][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:23:15,336][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:23:15,834][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:23:16,335][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:23:16,835][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:23:17,334][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:23:17,830][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:23:18,329][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:23:18,825][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10585 tokens. +[2026-03-25 22:23:19,468][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.58%, Block Peak % of device VRAM: 62.36%, ΔTime: 00:00:32 +[2026-03-25 22:23:20,187][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:23:20,190][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:23:20,191][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:23:20,873][__main__][INFO] - Iteration 89 took 50s (30.07% Gen, 68.59% Train). Generation: 15s, Training: 34s. Estimated remaining time: 40h 56m 19s. Estimated total time: 42h 14m 0s. Time estimates for 10 more iterations: 8m 26s, 100 more iterations: 1h 24m 28s, 500 more iterations: 7h 2m 20s. +[2026-03-25 22:23:20,875][__main__][INFO] - Starting iteration 89. +[2026-03-25 22:23:21,275][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 8 and human policies 1. +[2026-03-25 22:23:21,275][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:23:37,044][__main__][INFO] - Number of regex retries in iteration 89: 0 +[2026-03-25 22:23:37,045][__main__][INFO] - agents played in iteration 89 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:23:37,825][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:23:37,848][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:23:37,871][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:23:37,894][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:23:37,895][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:23:37,895][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:23:38,550][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:23:39,044][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:23:39,553][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:23:40,053][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:23:40,552][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:23:41,054][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:23:41,556][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:23:42,057][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:23:42,560][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:23:43,058][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:23:43,555][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:23:44,065][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:23:44,589][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:23:45,085][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:23:45,582][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:23:46,081][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:23:46,578][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:23:47,076][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:23:47,577][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:23:48,076][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:23:48,572][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:23:49,070][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:23:49,567][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:23:50,066][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:23:50,563][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:23:51,063][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:23:51,568][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:23:52,070][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:23:52,570][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:23:53,071][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:23:53,568][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:23:54,069][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:23:54,570][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:23:55,069][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:23:55,564][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:23:56,061][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:23:56,558][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:23:57,055][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:23:57,551][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:23:58,047][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:23:58,546][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:23:59,046][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:23:59,547][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:24:00,047][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:24:00,546][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:24:01,044][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:24:01,544][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:24:02,045][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:24:02,544][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:24:03,044][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:24:03,547][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:24:04,047][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:24:04,548][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:24:05,047][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:24:05,548][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:24:06,049][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:24:06,549][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:24:07,048][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:24:07,546][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:24:08,043][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:24:08,542][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:24:09,040][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:24:09,542][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:24:10,043][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:24:10,542][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10590 tokens. +[2026-03-25 22:24:11,204][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.57%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:32 +[2026-03-25 22:24:11,926][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:24:11,929][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:24:11,930][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:24:12,616][__main__][INFO] - Iteration 90 took 51s (30.72% Gen, 67.95% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 28m 32s. Estimated total time: 42h 47m 4s. Time estimates for 10 more iterations: 8m 33s, 100 more iterations: 1h 25m 34s, 500 more iterations: 7h 7m 50s. +[2026-03-25 22:24:12,619][__main__][INFO] - Starting iteration 90. +[2026-03-25 22:24:13,021][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 8 and human policies 1. +[2026-03-25 22:24:13,022][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:24:28,256][__main__][INFO] - Number of regex retries in iteration 90: 0 +[2026-03-25 22:24:28,258][__main__][INFO] - agents played in iteration 90 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:24:29,024][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:24:29,048][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:24:29,071][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:24:29,094][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:24:29,095][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:24:29,096][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:24:29,706][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:24:30,160][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:24:30,663][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:24:31,164][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:24:31,664][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:24:32,162][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:24:32,661][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:24:33,161][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:24:33,661][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:24:34,158][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:24:34,654][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:24:35,151][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:24:35,648][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:24:36,146][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:24:36,646][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:24:37,143][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:24:37,640][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:24:38,138][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:24:38,638][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:24:39,136][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:24:39,636][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:24:40,135][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:24:40,630][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:24:41,133][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:24:41,632][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:24:42,132][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:24:42,628][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:24:43,128][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:24:43,628][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:24:44,127][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:24:44,623][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:24:45,125][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:24:45,628][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:24:46,129][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:24:46,628][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:24:47,129][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:24:47,631][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:24:48,134][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:24:48,634][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:24:49,133][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:24:49,629][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:24:50,126][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:24:50,624][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:24:51,123][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:24:51,620][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:24:52,119][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:24:52,618][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:24:53,116][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:24:53,613][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:24:54,116][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:24:54,618][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:24:55,117][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:24:55,617][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:24:56,116][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:24:56,615][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:24:57,113][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:24:57,611][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:24:58,107][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:24:58,603][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:24:59,102][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:24:59,600][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:25:00,099][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:25:00,597][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:25:01,093][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:25:01,589][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10565 tokens. +[2026-03-25 22:25:02,235][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 60.56%, Block Peak % of device VRAM: 62.12%, ΔTime: 00:00:32 +[2026-03-25 22:25:02,956][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:25:02,958][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:25:02,959][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:25:04,231][__main__][INFO] - Iteration 91 took 51s (29.75% Gen, 67.76% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 21m 6s. Estimated total time: 42h 40m 30s. Time estimates for 10 more iterations: 8m 32s, 100 more iterations: 1h 25m 21s, 500 more iterations: 7h 6m 45s. +[2026-03-25 22:25:04,234][__main__][INFO] - Starting iteration 91. +[2026-03-25 22:25:04,634][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 9 and human policies 1. +[2026-03-25 22:25:04,635][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:25:20,073][__main__][INFO] - Number of regex retries in iteration 91: 0 +[2026-03-25 22:25:20,075][__main__][INFO] - agents played in iteration 91 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:25:20,844][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:25:20,868][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:25:20,891][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:25:20,915][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:25:20,915][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:25:20,916][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:25:21,547][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:25:22,270][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:25:22,756][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:25:23,256][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:25:23,758][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:25:24,256][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:25:24,757][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:25:25,253][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:25:25,752][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:25:26,251][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:25:26,751][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:25:27,255][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:25:27,759][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:25:28,263][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:25:28,766][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:25:29,263][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:25:29,764][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:25:30,264][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:25:30,765][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:25:31,264][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:25:31,764][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:25:32,264][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:25:32,765][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:25:33,263][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:25:33,762][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:25:34,261][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:25:34,757][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:25:35,255][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:25:35,750][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:25:36,247][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:25:36,746][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:25:37,245][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:25:37,747][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:25:38,248][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:25:38,744][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:25:39,242][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:25:39,739][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:25:40,239][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:25:40,737][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:25:41,237][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:25:41,736][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:25:42,234][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:25:42,734][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:25:43,231][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:25:43,728][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:25:44,229][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:25:44,727][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:25:45,225][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:25:45,723][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:25:46,220][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:25:46,716][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:25:47,215][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:25:47,710][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:25:48,205][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:25:48,702][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:25:49,198][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:25:49,695][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:25:50,197][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:25:50,696][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:25:51,195][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:25:51,694][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:25:52,197][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:25:52,697][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:25:53,197][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:25:53,696][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10423 tokens. +[2026-03-25 22:25:54,356][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 60.61%, Block Peak % of device VRAM: 62.23%, ΔTime: 00:00:32 +[2026-03-25 22:25:55,084][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:25:55,086][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:25:55,088][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:25:55,768][__main__][INFO] - Iteration 92 took 51s (30.19% Gen, 68.47% Train). Generation: 15s, Training: 35s. Estimated remaining time: 41h 16m 26s. Estimated total time: 42h 36m 41s. Time estimates for 10 more iterations: 8m 31s, 100 more iterations: 1h 25m 13s, 500 more iterations: 7h 6m 6s. +[2026-03-25 22:25:55,770][__main__][INFO] - Starting iteration 92. +[2026-03-25 22:25:56,171][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 9 and human policies 1. +[2026-03-25 22:25:56,172][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:26:04,744][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:26:11,403][__main__][INFO] - Number of regex retries in iteration 92: 1 +[2026-03-25 22:26:11,404][__main__][INFO] - agents played in iteration 92 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:26:12,166][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:26:12,190][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:26:12,213][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:26:12,236][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:26:12,236][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:26:12,237][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:26:12,863][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:26:13,318][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:26:13,823][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:26:14,322][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:26:14,823][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:26:15,322][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:26:15,821][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:26:16,318][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:26:16,818][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:26:17,318][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:26:17,817][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:26:18,319][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:26:18,817][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:26:19,317][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:26:19,813][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:26:20,314][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:26:20,813][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:26:21,309][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:26:21,803][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:26:22,302][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:26:22,798][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:26:23,297][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:26:23,794][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:26:24,295][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:26:24,791][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:26:25,290][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:26:25,789][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:26:26,288][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:26:26,787][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:26:27,286][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:26:27,782][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:26:28,283][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:26:28,782][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:26:29,282][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:26:29,779][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:26:30,275][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:26:30,773][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:26:31,268][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:26:31,958][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:26:32,441][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:26:32,936][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:26:33,437][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:26:33,937][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:26:34,436][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:26:34,936][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:26:35,436][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:26:35,935][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:26:36,433][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:26:36,932][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:26:37,427][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:26:37,925][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:26:38,423][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:26:38,922][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:26:39,419][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:26:39,918][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:26:40,417][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:26:40,915][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:26:41,416][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:26:41,917][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:26:42,416][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:26:42,916][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:26:43,415][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:26:43,914][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:26:44,414][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:26:44,912][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10622 tokens. +[2026-03-25 22:26:45,585][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 60.63%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:32 +[2026-03-25 22:26:46,304][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:26:46,307][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:26:46,308][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:26:47,023][__main__][INFO] - Iteration 93 took 50s (29.95% Gen, 68.64% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 1m 30s. Estimated total time: 42h 22m 37s. Time estimates for 10 more iterations: 8m 28s, 100 more iterations: 1h 24m 45s, 500 more iterations: 7h 3m 46s. +[2026-03-25 22:26:47,026][__main__][INFO] - Starting iteration 93. +[2026-03-25 22:26:47,426][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 9 and human policies 1. +[2026-03-25 22:26:47,426][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:26:56,433][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 1 y book, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:27:01,944][__main__][INFO] - Number of regex retries in iteration 93: 1 +[2026-03-25 22:27:01,945][__main__][INFO] - agents played in iteration 93 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:27:02,723][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:27:02,747][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:27:02,770][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:27:02,793][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:27:02,794][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:27:02,795][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:27:03,418][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:27:03,874][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:27:04,371][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:27:04,867][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:27:05,365][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:27:05,864][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:27:06,363][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:27:06,865][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:27:07,361][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:27:07,856][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:27:08,357][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:27:08,853][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:27:09,347][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:27:09,845][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:27:10,342][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:27:10,839][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:27:11,338][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:27:11,838][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:27:12,336][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:27:12,836][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:27:13,337][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:27:13,837][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:27:14,337][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:27:14,836][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:27:15,336][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:27:15,836][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:27:16,335][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:27:16,835][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:27:17,334][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:27:17,839][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:27:18,339][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:27:18,838][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:27:19,338][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:27:19,854][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:27:20,355][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:27:20,856][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:27:21,356][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:27:21,858][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:27:22,357][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:27:22,856][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:27:23,354][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:27:23,851][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:27:24,348][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:27:24,845][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:27:25,345][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:27:25,843][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:27:26,340][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:27:26,834][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:27:27,335][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:27:27,835][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:27:28,335][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:27:28,835][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:27:29,334][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:27:29,834][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:27:30,333][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:27:30,832][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:27:31,331][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:27:31,827][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:27:32,326][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:27:32,823][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:27:33,320][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:27:33,820][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:27:34,316][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:27:34,813][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:27:35,310][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10525 tokens. +[2026-03-25 22:27:35,965][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.24%, ΔTime: 00:00:32 +[2026-03-25 22:27:36,689][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:27:36,692][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:27:36,693][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:27:37,467][__main__][INFO] - Iteration 94 took 50s (29.01% Gen, 69.44% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 20m 10s. Estimated total time: 41h 42m 7s. Time estimates for 10 more iterations: 8m 20s, 100 more iterations: 1h 23m 24s, 500 more iterations: 6h 57m 1s. +[2026-03-25 22:27:37,470][__main__][INFO] - Starting iteration 94. +[2026-03-25 22:27:37,868][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 9 and human policies 1. +[2026-03-25 22:27:37,868][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:27:41,423][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:27:53,652][__main__][INFO] - Number of regex retries in iteration 94: 1 +[2026-03-25 22:27:53,653][__main__][INFO] - agents played in iteration 94 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:27:54,427][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:27:54,450][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:27:54,473][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:27:54,496][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:27:54,497][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:27:54,497][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:27:55,125][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:27:55,581][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:27:56,081][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:27:56,582][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:27:57,082][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:27:57,583][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:27:58,082][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:27:58,581][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:27:59,080][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:27:59,577][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:28:00,075][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:28:00,572][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:28:01,069][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:28:01,567][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:28:02,066][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:28:02,563][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:28:03,060][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:28:03,560][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:28:04,060][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:28:04,557][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:28:05,056][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:28:05,572][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:28:06,077][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:28:06,577][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:28:07,077][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:28:07,572][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:28:08,070][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:28:08,567][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:28:09,067][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:28:09,563][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:28:10,061][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:28:10,560][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:28:11,057][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:28:11,556][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:28:12,055][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:28:12,554][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:28:13,052][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:28:13,550][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:28:14,048][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:28:14,546][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:28:15,047][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:28:15,546][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:28:16,044][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:28:16,543][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:28:17,042][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:28:17,541][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:28:18,041][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:28:18,541][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:28:19,041][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:28:19,539][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:28:20,038][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:28:20,538][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:28:21,034][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:28:21,534][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:28:22,034][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:28:22,530][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:28:23,029][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:28:23,528][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:28:24,027][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:28:24,524][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:28:25,021][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:28:25,520][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:28:26,019][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:28:26,519][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:28:27,015][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10473 tokens. +[2026-03-25 22:28:27,647][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 60.54%, Block Peak % of device VRAM: 62.23%, ΔTime: 00:00:32 +[2026-03-25 22:28:28,366][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:28:28,368][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:28:28,370][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:28:29,148][__main__][INFO] - Iteration 95 took 51s (30.78% Gen, 67.70% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 21m 15s. Estimated total time: 42h 44m 3s. Time estimates for 10 more iterations: 8m 32s, 100 more iterations: 1h 25m 28s, 500 more iterations: 7h 7m 20s. +[2026-03-25 22:28:29,151][__main__][INFO] - Starting iteration 95. +[2026-03-25 22:28:29,550][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 9 and human policies 1. +[2026-03-25 22:28:29,550][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:28:45,773][__main__][INFO] - Number of regex retries in iteration 95: 0 +[2026-03-25 22:28:45,774][__main__][INFO] - agents played in iteration 95 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:28:46,571][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:28:46,595][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:28:46,618][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:28:46,641][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:28:46,642][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:28:46,642][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:28:47,291][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:28:47,747][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:28:48,251][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:28:48,749][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:28:49,247][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:28:49,748][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:28:50,247][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:28:50,748][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:28:51,246][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:28:51,744][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:28:52,241][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:28:52,741][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:28:53,240][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:28:53,740][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:28:54,239][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:28:54,735][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:28:55,235][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:28:55,731][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:28:56,227][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:28:56,723][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:28:57,221][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:28:57,718][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:28:58,215][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:28:58,712][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:28:59,211][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:28:59,709][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:29:00,206][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:29:00,705][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:29:01,204][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:29:01,701][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:29:02,203][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:29:02,703][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:29:03,202][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:29:03,701][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:29:04,200][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:29:04,699][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:29:05,193][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:29:05,688][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:29:06,187][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:29:06,680][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:29:07,178][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:29:07,676][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:29:08,176][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:29:08,677][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:29:09,174][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:29:09,670][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:29:10,170][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:29:10,667][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:29:11,166][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:29:11,662][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:29:12,161][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:29:12,659][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:29:13,159][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:29:13,655][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:29:14,153][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:29:14,652][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:29:15,153][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:29:15,652][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:29:16,147][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:29:16,645][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:29:17,140][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:29:17,636][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:29:18,134][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:29:18,630][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:29:19,125][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10471 tokens. +[2026-03-25 22:29:19,784][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 60.50%, Block Peak % of device VRAM: 62.13%, ΔTime: 00:00:32 +[2026-03-25 22:29:20,511][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:29:20,514][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:29:20,517][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:29:21,276][__main__][INFO] - Iteration 96 took 51s (31.37% Gen, 67.17% Train). Generation: 16s, Training: 34s. Estimated remaining time: 41h 42m 41s. Estimated total time: 43h 6m 22s. Time estimates for 10 more iterations: 8m 37s, 100 more iterations: 1h 26m 12s, 500 more iterations: 7h 11m 3s. +[2026-03-25 22:29:21,279][__main__][INFO] - Starting iteration 96. +[2026-03-25 22:29:21,681][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 9 and human policies 1. +[2026-03-25 22:29:21,681][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:29:24,270][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:29:26,502][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:29:33,047][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:29:36,541][mllm.models.large_language_model_local][WARNING] - Response Given the per-item values, it's clear that both you and Alice have very different valuations for the items. You value hats and books more, while Alice values books more. However, since you also value balls moderately, it might be beneficial to ensure you get some balls, even though the value is lower compared to your other choices. A balanced approach would ensure you don't miss out on items you value highly. -Your per-item values are: hats=10, books=1, balls=10. -Alice's per-item values are: hats=10, books=10, balls=1. +Proposal: 8 hats, 8 books, 4 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:29:37,919][__main__][INFO] - Number of regex retries in iteration 96: 4 +[2026-03-25 22:29:37,920][__main__][INFO] - agents played in iteration 96 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:29:38,695][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:29:38,718][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:29:38,742][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:29:38,765][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:29:38,765][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:29:38,766][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:29:39,384][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:29:39,841][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:29:40,344][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:29:40,846][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:29:41,345][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:29:41,846][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:29:42,347][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:29:42,847][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:29:43,345][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:29:43,842][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:29:44,342][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:29:44,843][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:29:45,344][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:29:45,844][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:29:46,343][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:29:46,842][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:29:47,342][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:29:47,843][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:29:48,343][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:29:48,843][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:29:49,342][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:29:49,841][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:29:50,342][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:29:50,842][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:29:51,343][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:29:51,843][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:29:52,342][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:29:52,841][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:29:53,340][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:29:53,839][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:29:54,339][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:29:54,839][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:29:55,338][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:29:55,837][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:29:56,337][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:29:56,836][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:29:57,336][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:29:57,835][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:29:58,335][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:29:58,836][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:29:59,335][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:29:59,836][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:30:00,336][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:30:00,834][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:30:01,330][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:30:01,830][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:30:02,329][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:30:02,827][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:30:03,323][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:30:03,823][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:30:04,322][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:30:04,819][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:30:05,317][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:30:05,818][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:30:06,318][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:30:06,817][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:30:07,316][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:30:07,816][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:30:08,311][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:30:08,810][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:30:09,306][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:30:09,804][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:30:10,302][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:30:10,798][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:30:11,298][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10683 tokens. +[2026-03-25 22:30:11,947][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 60.62%, Block Peak % of device VRAM: 62.21%, ΔTime: 00:00:32 +[2026-03-25 22:30:12,676][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:30:12,678][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:30:12,680][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:30:13,370][__main__][INFO] - Iteration 97 took 51s (31.42% Gen, 67.25% Train). Generation: 16s, Training: 34s. Estimated remaining time: 41h 39m 57s. Estimated total time: 43h 4m 30s. Time estimates for 10 more iterations: 8m 36s, 100 more iterations: 1h 26m 9s, 500 more iterations: 7h 10m 45s. +[2026-03-25 22:30:13,373][__main__][INFO] - Starting iteration 97. +[2026-03-25 22:30:13,776][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 9 and human policies 1. +[2026-03-25 22:30:13,776][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:30:24,488][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 11 books, 1 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:30:28,512][__main__][INFO] - Number of regex retries in iteration 97: 1 +[2026-03-25 22:30:28,512][__main__][INFO] - agents played in iteration 97 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:30:29,277][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:30:29,301][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:30:29,324][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:30:29,347][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:30:29,347][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:30:29,348][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:30:29,960][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:30:30,416][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:30:30,919][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:30:31,417][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:30:31,921][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:30:32,421][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:30:32,919][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:30:33,415][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:30:33,918][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:30:34,420][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:30:34,920][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:30:35,420][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:30:35,919][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:30:36,427][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:30:36,933][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:30:37,435][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:30:37,941][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:30:38,444][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:30:38,945][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:30:39,448][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:30:39,949][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:30:40,447][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:30:40,946][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:30:41,447][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:30:41,948][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:30:42,448][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:30:42,947][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:30:43,449][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:30:43,950][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:30:44,449][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:30:44,948][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:30:45,447][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:30:45,950][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:30:46,449][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:30:46,946][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:30:47,445][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:30:47,948][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:30:48,443][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:30:48,942][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:30:49,444][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:30:49,946][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:30:50,445][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:30:50,944][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:30:51,446][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:30:51,952][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:30:52,451][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:30:52,950][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:30:53,452][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:30:53,954][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:30:54,448][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:30:54,946][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:30:55,445][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:30:55,945][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:30:56,446][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:30:56,945][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:30:57,448][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:30:57,947][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:30:58,447][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:30:58,947][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:30:59,448][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:30:59,950][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:31:00,449][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:31:00,946][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:31:01,449][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:31:01,952][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10438 tokens. +[2026-03-25 22:31:02,605][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 60.66%, Block Peak % of device VRAM: 62.24%, ΔTime: 00:00:32 +[2026-03-25 22:31:03,338][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:31:03,341][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:31:03,342][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:31:04,069][__main__][INFO] - Iteration 98 took 50s (29.30% Gen, 69.25% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 29m 18s. Estimated total time: 41h 54m 42s. Time estimates for 10 more iterations: 8m 22s, 100 more iterations: 1h 23m 49s, 500 more iterations: 6h 59m 7s. +[2026-03-25 22:31:04,072][__main__][INFO] - Starting iteration 98. +[2026-03-25 22:31:04,473][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 9 and human policies 1. +[2026-03-25 22:31:04,474][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:31:06,899][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:31:19,657][__main__][INFO] - Number of regex retries in iteration 98: 1 +[2026-03-25 22:31:19,658][__main__][INFO] - agents played in iteration 98 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:31:20,423][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:31:20,446][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:31:20,469][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:31:20,492][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:31:20,493][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:31:20,493][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:31:21,099][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:31:21,558][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:31:22,065][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:31:22,566][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:31:23,067][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:31:23,570][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:31:24,071][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:31:24,570][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:31:25,071][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:31:25,572][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:31:26,075][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:31:26,573][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:31:27,070][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:31:27,569][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:31:28,069][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:31:28,568][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:31:29,066][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:31:29,564][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:31:30,066][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:31:30,563][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:31:31,063][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:31:31,563][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:31:32,065][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:31:32,561][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:31:33,061][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:31:33,561][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:31:34,060][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:31:34,558][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:31:35,055][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:31:35,554][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:31:36,052][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:31:36,548][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:31:37,048][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:31:37,549][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:31:38,049][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:31:38,548][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:31:39,048][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:31:39,546][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:31:40,046][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:31:40,548][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:31:41,048][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:31:41,548][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:31:42,047][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:31:42,546][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:31:43,046][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:31:43,545][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:31:44,048][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:31:44,547][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:31:45,049][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:31:45,550][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:31:46,051][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:31:46,553][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:31:47,056][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:31:47,556][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:31:48,058][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:31:48,560][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:31:49,059][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:31:49,556][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:31:50,055][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:31:50,555][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:31:51,053][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:31:51,549][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:31:52,048][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:31:52,548][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:31:53,046][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10520 tokens. +[2026-03-25 22:31:53,699][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.58%, Block Peak % of device VRAM: 62.27%, ΔTime: 00:00:32 +[2026-03-25 22:31:54,422][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:31:54,424][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:31:54,426][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:31:55,115][__main__][INFO] - Iteration 99 took 50s (29.98% Gen, 68.65% Train). Generation: 15s, Training: 34s. Estimated remaining time: 40h 45m 51s. Estimated total time: 42h 12m 6s. Time estimates for 10 more iterations: 8m 26s, 100 more iterations: 1h 24m 24s, 500 more iterations: 7h 2m 1s. +[2026-03-25 22:31:55,117][__main__][INFO] - Starting iteration 99. +[2026-03-25 22:31:55,515][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 9 and human policies 1. +[2026-03-25 22:31:55,515][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:32:11,277][__main__][INFO] - Number of regex retries in iteration 99: 0 +[2026-03-25 22:32:11,277][__main__][INFO] - agents played in iteration 99 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:32:12,056][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:32:12,080][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:32:12,104][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:32:12,127][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:32:12,128][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:32:12,128][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:32:12,751][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:32:13,209][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:32:13,712][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:32:14,217][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:32:14,722][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:32:15,224][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:32:15,724][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:32:16,230][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:32:16,733][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:32:17,234][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:32:17,732][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:32:18,231][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:32:18,732][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:32:19,232][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:32:19,729][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:32:20,232][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:32:20,729][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:32:21,228][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:32:21,730][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:32:22,231][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:32:22,730][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:32:23,228][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:32:23,725][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:32:24,226][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:32:24,725][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:32:25,226][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:32:25,727][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:32:26,227][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:32:26,726][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:32:27,225][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:32:27,726][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:32:28,227][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:32:28,729][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:32:29,228][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:32:29,726][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:32:30,226][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:32:30,726][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:32:31,223][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:32:31,724][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:32:32,223][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:32:32,722][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:32:33,221][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:32:33,721][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:32:34,221][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:32:34,722][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:32:35,222][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:32:35,719][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:32:36,218][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:32:36,719][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:32:37,219][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:32:37,721][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:32:38,222][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:32:38,724][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:32:39,225][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:32:39,722][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:32:40,222][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:32:40,725][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:32:41,222][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:32:41,724][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:32:42,223][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:32:42,721][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:32:43,222][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:32:43,723][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:32:44,223][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:32:44,722][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10582 tokens. +[2026-03-25 22:32:45,376][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.58%, Block Peak % of device VRAM: 62.27%, ΔTime: 00:00:32 +[2026-03-25 22:32:46,096][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:32:46,098][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:32:46,099][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:32:46,806][__main__][INFO] - Iteration 100 took 51s (30.73% Gen, 67.89% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 17m 27s. Estimated total time: 42h 44m 33s. Time estimates for 10 more iterations: 8m 32s, 100 more iterations: 1h 25m 29s, 500 more iterations: 7h 7m 25s. +[2026-03-25 22:32:46,808][__main__][INFO] - Starting iteration 100. +[2026-03-25 22:32:47,207][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 9 and human policies 1. +[2026-03-25 22:32:47,208][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:33:04,625][__main__][INFO] - Number of regex retries in iteration 100: 0 +[2026-03-25 22:33:04,626][__main__][INFO] - agents played in iteration 100 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:33:05,414][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:33:05,438][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:33:05,461][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:33:05,485][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:33:05,485][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:33:05,486][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:33:06,105][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:33:06,560][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:33:07,062][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:33:07,565][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:33:08,067][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:33:08,569][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:33:09,067][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:33:09,573][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:33:10,071][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:33:10,597][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:33:11,099][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:33:11,601][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:33:12,103][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:33:12,604][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:33:13,105][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:33:13,607][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:33:14,109][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:33:14,608][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:33:15,107][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:33:15,608][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:33:16,110][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:33:16,610][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:33:17,109][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:33:17,610][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:33:18,113][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:33:18,615][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:33:19,116][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:33:19,618][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:33:20,117][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:33:20,615][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:33:21,115][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:33:21,616][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:33:22,121][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:33:22,619][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:33:23,117][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:33:23,616][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:33:24,114][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:33:24,610][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:33:25,107][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:33:25,607][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:33:26,106][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:33:26,605][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:33:27,106][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:33:27,603][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:33:28,104][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:33:28,603][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:33:29,102][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:33:29,601][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:33:30,103][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:33:30,602][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:33:31,101][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:33:31,598][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:33:32,095][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:33:32,595][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:33:33,095][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:33:33,597][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:33:34,098][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:33:34,599][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:33:35,098][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:33:35,599][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:33:36,094][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:33:36,591][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:33:37,089][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:33:37,587][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:33:38,085][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10599 tokens. +[2026-03-25 22:33:38,734][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 60.53%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 +[2026-03-25 22:33:39,446][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:33:39,448][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:33:39,450][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:33:40,971][__main__][INFO] - Iteration 101 took 53s (32.40% Gen, 64.77% Train). Generation: 17s, Training: 34s. Estimated remaining time: 43h 20m 13s. Estimated total time: 44h 48m 14s. Time estimates for 10 more iterations: 8m 57s, 100 more iterations: 1h 29m 36s, 500 more iterations: 7h 28m 2s. +[2026-03-25 22:33:40,974][__main__][INFO] - Starting iteration 101. +[2026-03-25 22:33:41,377][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 10 and human policies 1. +[2026-03-25 22:33:41,378][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:33:56,768][mllm.models.large_language_model_local][WARNING] - Response Proposal: 11 hats, 10 books, 1 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:33:57,569][__main__][INFO] - Number of regex retries in iteration 101: 1 +[2026-03-25 22:33:57,570][__main__][INFO] - agents played in iteration 101 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:33:58,363][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:33:58,387][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:33:58,410][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:33:58,433][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:33:58,434][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:33:58,434][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:33:59,050][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:33:59,511][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:34:00,015][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:34:00,515][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:34:01,022][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:34:01,521][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:34:02,023][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:34:02,526][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:34:03,027][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:34:03,524][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:34:04,021][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:34:04,522][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:34:05,020][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:34:05,520][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:34:06,020][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:34:06,520][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:34:07,018][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:34:07,518][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:34:08,019][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:34:08,517][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:34:09,019][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:34:09,519][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:34:10,020][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:34:10,522][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:34:11,025][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:34:11,524][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:34:12,023][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:34:12,526][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:34:13,027][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:34:13,525][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:34:14,025][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:34:14,525][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:34:15,026][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:34:15,522][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:34:16,023][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:34:16,525][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:34:17,024][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:34:17,524][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:34:18,025][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:34:18,525][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:34:19,024][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:34:19,523][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:34:20,023][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:34:20,523][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:34:21,027][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:34:21,528][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:34:22,029][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:34:22,529][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:34:23,033][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:34:23,533][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:34:24,037][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:34:24,538][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:34:25,038][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:34:25,540][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:34:26,043][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:34:26,543][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:34:27,044][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:34:27,540][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:34:28,040][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:34:28,535][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:34:29,033][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:34:29,529][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:34:30,027][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:34:30,528][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:34:31,027][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10597 tokens. +[2026-03-25 22:34:31,701][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 60.56%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 +[2026-03-25 22:34:32,408][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:34:32,411][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:34:32,412][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:34:33,110][__main__][INFO] - Iteration 102 took 51s (31.30% Gen, 67.35% Train). Generation: 16s, Training: 34s. Estimated remaining time: 41h 37m 49s. Estimated total time: 43h 6m 41s. Time estimates for 10 more iterations: 8m 37s, 100 more iterations: 1h 26m 13s, 500 more iterations: 7h 11m 6s. +[2026-03-25 22:34:33,112][__main__][INFO] - Starting iteration 102. +[2026-03-25 22:34:33,517][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 10 and human policies 1. +[2026-03-25 22:34:33,518][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:34:36,140][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:34:41,812][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls -You should propose to take all the items that you value highly, and let Alice take the items she values more. +Given the per-item values, you have a significantly higher value for hats (10) compared to Bob's value (1). Books and balls have the same values for both of you, making them less attractive to go after compared to hats. Additionally, since the total number of items is such that if both of us take the entire quantity of hats, it doesn't exceed the item quantity, we can maximize our points by taking all the hats. did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:34:50,014][__main__][INFO] - Number of regex retries in iteration 102: 2 +[2026-03-25 22:34:50,015][__main__][INFO] - agents played in iteration 102 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:34:50,795][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:34:50,819][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:34:50,842][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:34:50,865][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:34:50,865][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:34:50,866][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:34:51,483][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:34:51,934][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:34:52,440][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:34:52,940][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:34:53,435][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:34:53,931][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:34:54,433][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:34:54,934][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:34:55,437][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:34:55,932][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:34:56,430][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:34:56,928][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:34:57,427][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:34:57,925][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:34:58,425][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:34:58,922][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:34:59,421][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:34:59,920][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:35:00,424][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:35:00,927][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:35:01,427][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:35:01,928][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:35:02,429][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:35:02,929][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:35:03,428][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:35:03,929][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:35:04,430][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:35:04,930][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:35:05,428][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:35:05,928][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:35:06,428][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:35:06,927][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:35:07,428][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:35:07,951][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:35:08,450][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:35:08,949][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:35:09,448][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:35:09,950][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:35:10,452][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:35:10,952][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:35:11,451][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:35:11,952][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:35:12,452][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:35:12,954][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:35:13,454][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:35:13,953][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:35:14,453][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:35:14,950][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:35:15,450][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:35:15,952][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:35:16,452][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:35:16,956][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:35:17,454][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:35:17,955][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:35:18,458][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:35:18,958][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:35:19,458][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:35:19,956][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:35:20,456][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:35:20,953][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:35:21,452][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:35:21,953][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:35:22,449][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:35:22,949][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:35:23,448][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10542 tokens. +[2026-03-25 22:35:24,087][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 60.60%, Block Peak % of device VRAM: 62.23%, ΔTime: 00:00:32 +[2026-03-25 22:35:24,783][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:35:24,785][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:35:24,787][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:35:25,480][__main__][INFO] - Iteration 103 took 51s (31.75% Gen, 66.92% Train). Generation: 16s, Training: 34s. Estimated remaining time: 41h 48m 24s. Estimated total time: 43h 18m 9s. Time estimates for 10 more iterations: 8m 39s, 100 more iterations: 1h 26m 36s, 500 more iterations: 7h 13m 1s. +[2026-03-25 22:35:25,482][__main__][INFO] - Starting iteration 103. +[2026-03-25 22:35:25,882][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 10 and human policies 1. +[2026-03-25 22:35:25,882][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:35:35,026][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 1 y book, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:35:41,892][__main__][INFO] - Number of regex retries in iteration 103: 1 +[2026-03-25 22:35:41,893][__main__][INFO] - agents played in iteration 103 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:35:42,670][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:35:42,693][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:35:42,716][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:35:42,739][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:35:42,739][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:35:42,740][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:35:43,341][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:35:43,793][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:35:44,299][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:35:44,798][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:35:45,296][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:35:45,799][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:35:46,298][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:35:46,796][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:35:47,290][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:35:47,795][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:35:48,298][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:35:48,798][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:35:49,297][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:35:49,797][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:35:50,298][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:35:50,799][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:35:51,300][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:35:51,802][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:35:52,300][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:35:52,800][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:35:53,300][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:35:53,801][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:35:54,302][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:35:54,801][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:35:55,300][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:35:55,802][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:35:56,304][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:35:56,800][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:35:57,302][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:35:57,803][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:35:58,302][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:35:58,799][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:35:59,299][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:35:59,801][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:36:00,304][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:36:00,804][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:36:01,305][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:36:01,807][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:36:02,309][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:36:02,833][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:36:03,337][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:36:03,839][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:36:04,339][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:36:04,839][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:36:05,345][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:36:05,850][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:36:06,350][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:36:06,852][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:36:07,355][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:36:07,856][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:36:08,355][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:36:08,859][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:36:09,360][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:36:09,862][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:36:10,362][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:36:10,863][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:36:11,366][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:36:11,863][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:36:12,362][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:36:12,860][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:36:13,360][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:36:13,856][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:36:14,354][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:36:14,853][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:36:15,352][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10535 tokens. +[2026-03-25 22:36:16,006][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.34%, ΔTime: 00:00:32 +[2026-03-25 22:36:16,730][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:36:16,732][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:36:16,734][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:36:17,427][__main__][INFO] - Iteration 104 took 51s (31.06% Gen, 67.59% Train). Generation: 16s, Training: 34s. Estimated remaining time: 41h 26m 41s. Estimated total time: 42h 57m 18s. Time estimates for 10 more iterations: 8m 35s, 100 more iterations: 1h 25m 54s, 500 more iterations: 7h 9m 33s. +[2026-03-25 22:36:17,430][__main__][INFO] - Starting iteration 104. +[2026-03-25 22:36:17,830][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 10 and human policies 1. +[2026-03-25 22:36:17,831][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:36:32,956][__main__][INFO] - Number of regex retries in iteration 104: 0 +[2026-03-25 22:36:32,957][__main__][INFO] - agents played in iteration 104 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:36:33,715][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:36:33,738][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:36:33,761][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:36:33,784][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:36:33,785][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:36:33,785][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:36:34,395][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:36:34,855][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:36:35,359][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:36:35,856][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:36:36,354][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:36:36,854][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:36:37,354][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:36:37,853][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:36:38,351][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:36:38,855][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:36:39,354][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:36:39,852][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:36:40,354][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:36:40,856][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:36:41,354][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:36:41,854][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:36:42,352][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:36:42,853][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:36:43,350][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:36:43,850][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:36:44,351][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:36:44,848][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:36:45,345][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:36:45,846][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:36:46,348][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:36:46,853][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:36:47,370][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:36:47,875][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:36:48,379][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:36:48,881][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:36:49,383][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:36:49,887][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:36:50,391][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:36:50,890][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:36:51,392][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:36:51,894][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:36:52,398][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:36:52,898][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:36:53,399][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:36:53,898][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:36:54,398][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:36:54,895][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:36:55,392][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:36:55,891][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:36:56,392][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:36:56,892][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:36:57,394][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:36:57,896][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:36:58,395][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:36:58,890][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:36:59,390][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:36:59,889][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:37:00,387][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:37:00,882][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:37:01,382][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:37:01,884][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:37:02,380][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:37:02,879][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:37:03,376][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:37:03,876][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:37:04,378][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:37:04,873][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:37:05,375][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:37:05,874][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:37:06,374][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10512 tokens. +[2026-03-25 22:37:07,005][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 60.61%, Block Peak % of device VRAM: 62.39%, ΔTime: 00:00:32 +[2026-03-25 22:37:07,726][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:37:07,728][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:37:07,730][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:37:08,420][__main__][INFO] - Iteration 105 took 50s (29.90% Gen, 68.73% Train). Generation: 15s, Training: 34s. Estimated remaining time: 40h 38m 4s. Estimated total time: 42h 9m 32s. Time estimates for 10 more iterations: 8m 25s, 100 more iterations: 1h 24m 19s, 500 more iterations: 7h 1m 35s. +[2026-03-25 22:37:08,423][__main__][INFO] - Starting iteration 105. +[2026-03-25 22:37:08,821][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 10 and human policies 1. +[2026-03-25 22:37:08,822][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:37:24,043][__main__][INFO] - Number of regex retries in iteration 105: 0 +[2026-03-25 22:37:24,044][__main__][INFO] - agents played in iteration 105 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:37:24,818][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:37:24,841][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:37:24,864][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:37:24,887][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:37:24,888][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:37:24,889][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:37:25,504][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:37:25,964][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:37:26,467][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:37:26,966][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:37:27,467][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:37:27,970][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:37:28,466][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:37:28,967][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:37:29,466][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:37:29,969][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:37:30,468][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:37:30,969][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:37:31,469][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:37:31,969][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:37:32,468][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:37:32,968][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:37:33,468][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:37:33,974][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:37:34,475][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:37:34,973][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:37:35,476][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:37:35,998][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:37:36,500][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:37:37,003][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:37:37,506][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:37:38,002][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:37:38,500][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:37:38,998][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:37:39,500][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:37:40,002][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:37:40,499][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:37:41,002][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:37:41,503][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:37:42,004][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:37:42,504][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:37:43,006][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:37:43,508][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:37:44,009][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:37:44,508][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:37:45,006][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:37:45,508][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:37:46,007][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:37:46,504][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:37:47,005][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:37:47,507][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:37:48,005][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:37:48,503][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:37:49,003][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:37:49,503][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:37:50,005][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:37:50,503][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:37:51,005][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:37:51,507][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:37:52,009][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:37:52,507][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:37:53,007][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:37:53,508][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:37:54,009][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:37:54,506][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:37:55,006][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:37:55,508][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:37:56,008][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:37:56,504][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:37:57,005][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:37:57,505][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10494 tokens. +[2026-03-25 22:37:58,138][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 60.56%, Block Peak % of device VRAM: 62.28%, ΔTime: 00:00:32 +[2026-03-25 22:37:58,864][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:37:58,866][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:37:58,868][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:37:59,564][__main__][INFO] - Iteration 106 took 50s (30.00% Gen, 68.63% Train). Generation: 15s, Training: 34s. Estimated remaining time: 40h 44m 52s. Estimated total time: 42h 17m 11s. Time estimates for 10 more iterations: 8m 27s, 100 more iterations: 1h 24m 34s, 500 more iterations: 7h 2m 51s. +[2026-03-25 22:37:59,567][__main__][INFO] - Starting iteration 106. +[2026-03-25 22:37:59,966][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 10 and human policies 1. +[2026-03-25 22:37:59,966][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:38:14,168][mllm.models.large_language_model_local][WARNING] - Response Given the updated values where you have a higher value for balls (10) compared to hats and books (1 each), and Bob values books the most, it's strategic to maximize the allocation of balls while also asserting some value for hats and books to avoid any proportionate loss. Here’s a balanced proposal: -Proposal: 10 hats, 10 balls, 0 books did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:46:00,882][__main__][INFO] - Number of regex retries in iteration 123: 1 -[2025-11-09 19:46:00,883][__main__][INFO] - agents played in iteration 123 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:46:01,665][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:46:01,688][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:46:01,711][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:46:01,732][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:46:01,733][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:46:01,733][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:46:02,355][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:46:02,959][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:46:03,466][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:46:03,966][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:46:04,466][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:46:04,967][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:46:05,466][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:46:05,966][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:46:06,464][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:46:06,962][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:46:07,460][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:46:07,962][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:46:08,464][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:46:08,966][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:46:09,468][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:46:09,972][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:46:10,473][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:46:10,977][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:46:11,480][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:46:11,981][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:46:12,482][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:46:12,983][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:46:13,484][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:46:13,985][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:46:14,487][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:46:14,991][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:46:15,489][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:46:15,987][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:46:16,497][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:46:16,995][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:46:17,497][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:46:17,995][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:46:18,494][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:46:18,998][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:46:19,499][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:46:19,998][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:46:20,498][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:46:20,999][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:46:21,512][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:46:22,011][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:46:22,513][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:46:23,019][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:46:23,517][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:46:24,015][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:46:24,518][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:46:25,019][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:46:25,527][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:46:26,024][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:46:26,525][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:46:27,035][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:46:27,534][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:46:28,034][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:46:28,533][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:46:29,033][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:46:29,539][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:46:30,038][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:46:30,537][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:46:31,038][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:46:31,538][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:46:32,042][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:46:32,541][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:46:33,041][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:46:33,540][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:46:34,037][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:46:34,538][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10866 tokens. -[2025-11-09 19:46:35,178][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.50%, ΔTime: 00:00:32 -[2025-11-09 19:46:35,929][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:46:35,931][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:46:35,933][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:46:36,883][__main__][INFO] - Iteration 124 took 49s (27.76% Gen, 70.33% Train). Generation: 13s, Training: 35s. Estimated remaining time: 39h 44m 5s. Estimated total time: 41h 31m 45s. Time estimates for 10 more iterations: 8m 18s, 100 more iterations: 1h 23m 3s, 500 more iterations: 6h 55m 17s. -[2025-11-09 19:46:36,885][__main__][INFO] - Starting iteration 124. -[2025-11-09 19:46:37,373][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 12 and human policies 1. -[2025-11-09 19:46:37,373][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:46:40,219][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:46:46,969][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:46:49,098][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:46:51,272][__main__][INFO] - Number of regex retries in iteration 124: 3 -[2025-11-09 19:46:51,273][__main__][INFO] - agents played in iteration 124 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:46:52,054][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:46:52,076][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:46:52,098][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:46:52,121][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:46:52,122][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:46:52,122][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:46:52,770][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:46:53,227][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:46:53,732][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:46:54,238][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:46:54,735][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:46:55,235][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:46:55,735][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:46:56,234][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:46:56,736][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:46:57,238][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:46:57,742][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:46:58,250][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:46:58,750][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:46:59,252][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:46:59,753][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:47:00,254][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:47:00,757][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:47:01,257][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:47:01,757][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:47:02,258][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:47:02,758][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:47:03,259][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:47:03,761][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:47:04,259][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:47:04,760][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:47:05,258][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:47:05,754][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:47:06,251][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:47:06,749][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:47:07,251][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:47:07,749][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:47:08,246][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:47:08,751][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:47:09,248][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:47:09,748][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:47:10,243][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:47:10,740][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:47:11,242][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:47:11,740][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:47:12,237][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:47:12,736][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:47:13,235][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:47:13,732][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:47:14,229][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:47:14,730][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:47:15,245][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:47:15,743][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:47:16,243][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:47:16,743][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:47:17,240][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:47:17,743][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:47:18,243][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:47:18,741][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:47:19,247][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:47:19,747][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:47:20,246][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:47:20,745][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:47:21,242][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:47:21,739][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:47:22,236][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:47:22,734][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:47:23,244][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:47:23,738][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:47:24,235][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:47:24,734][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10866 tokens. -[2025-11-09 19:47:25,398][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 55.70%, Block Peak % of device VRAM: 62.34%, ΔTime: 00:00:32 -[2025-11-09 19:47:26,141][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:47:26,142][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:47:26,144][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:47:27,090][__main__][INFO] - Iteration 125 took 49s (27.96% Gen, 70.14% Train). Generation: 13s, Training: 34s. Estimated remaining time: 39h 37m 24s. Estimated total time: 41h 25m 54s. Time estimates for 10 more iterations: 8m 17s, 100 more iterations: 1h 22m 51s, 500 more iterations: 6h 54m 19s. -[2025-11-09 19:47:27,092][__main__][INFO] - Starting iteration 125. -[2025-11-09 19:47:27,571][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 12 and human policies 1. -[2025-11-09 19:47:27,571][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:47:41,125][__main__][INFO] - Number of regex retries in iteration 125: 0 -[2025-11-09 19:47:41,126][__main__][INFO] - agents played in iteration 125 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:47:41,904][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:47:41,929][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:47:41,953][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:47:41,975][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:47:41,976][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:47:41,977][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:47:42,634][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:47:43,102][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:47:43,608][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:47:44,111][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:47:44,611][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:47:45,110][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:47:45,618][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:47:46,116][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:47:46,615][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:47:47,123][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:47:47,623][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:47:48,130][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:47:48,630][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:47:49,132][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:47:49,653][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:47:50,156][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:47:50,661][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:47:51,161][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:47:51,660][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:47:52,164][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:47:52,664][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:47:53,162][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:47:53,664][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:47:54,163][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:47:54,660][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:47:55,158][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:47:55,657][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:47:56,159][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:47:56,657][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:47:57,154][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:47:57,654][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:47:58,153][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:47:58,654][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:47:59,153][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:47:59,652][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:48:00,157][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:48:00,655][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:48:01,155][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:48:01,653][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:48:02,151][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:48:02,771][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:48:03,269][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:48:03,770][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:48:04,268][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:48:04,770][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:48:05,273][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:48:05,774][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:48:06,271][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:48:06,774][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:48:07,276][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:48:07,780][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:48:08,283][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:48:08,783][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:48:09,287][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:48:09,792][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:48:10,293][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:48:10,793][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:48:11,292][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:48:11,790][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:48:12,287][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:48:12,784][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:48:13,285][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:48:13,782][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:48:14,278][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:48:14,775][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10869 tokens. -[2025-11-09 19:48:15,425][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 55.70%, Block Peak % of device VRAM: 62.43%, ΔTime: 00:00:32 -[2025-11-09 19:48:16,182][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:48:16,183][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:48:16,185][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:48:17,180][__main__][INFO] - Iteration 126 took 49s (27.32% Gen, 70.67% Train). Generation: 13s, Training: 35s. Estimated remaining time: 39h 31m 9s. Estimated total time: 41h 20m 29s. Time estimates for 10 more iterations: 8m 16s, 100 more iterations: 1h 22m 40s, 500 more iterations: 6h 53m 24s. -[2025-11-09 19:48:17,182][__main__][INFO] - Starting iteration 126. -[2025-11-09 19:48:17,664][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 12 and human policies 1. -[2025-11-09 19:48:17,664][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:48:31,965][__main__][INFO] - Number of regex retries in iteration 126: 0 -[2025-11-09 19:48:31,966][__main__][INFO] - agents played in iteration 126 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:48:32,759][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:48:32,782][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:48:32,807][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:48:32,830][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:48:32,830][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:48:32,831][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:48:33,446][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:48:33,910][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:48:34,415][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:48:34,915][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:48:35,418][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:48:35,920][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:48:36,420][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:48:36,928][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:48:37,429][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:48:37,929][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:48:38,428][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:48:38,926][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:48:39,441][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:48:39,941][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:48:40,442][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:48:40,942][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:48:41,441][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:48:41,954][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:48:42,454][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:48:42,954][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:48:43,455][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:48:43,955][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:48:44,460][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:48:44,960][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:48:45,460][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:48:45,960][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:48:46,460][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:48:46,959][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:48:47,458][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:48:47,957][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:48:48,464][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:48:48,963][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:48:49,463][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:48:49,970][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:48:50,467][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:48:50,966][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:48:51,464][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:48:51,965][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:48:52,469][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:48:52,968][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:48:53,467][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:48:53,969][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:48:54,471][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:48:54,975][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:48:55,480][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:48:55,981][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:48:56,482][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:48:56,982][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:48:57,478][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:48:57,979][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:48:58,479][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:48:58,978][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:48:59,477][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:48:59,976][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:49:00,474][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:49:00,972][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:49:01,471][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:49:01,970][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:49:02,470][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:49:02,973][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:49:03,473][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:49:03,972][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:49:04,472][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:49:04,972][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:49:05,472][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10874 tokens. -[2025-11-09 19:49:06,131][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 -[2025-11-09 19:49:06,888][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:49:06,890][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:49:06,892][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:49:07,842][__main__][INFO] - Iteration 127 took 50s (28.50% Gen, 69.60% Train). Generation: 14s, Training: 34s. Estimated remaining time: 39h 58m 46s. Estimated total time: 41h 48m 57s. Time estimates for 10 more iterations: 8m 21s, 100 more iterations: 1h 23m 37s, 500 more iterations: 6h 58m 9s. -[2025-11-09 19:49:07,845][__main__][INFO] - Starting iteration 127. -[2025-11-09 19:49:08,311][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 12 and human policies 1. -[2025-11-09 19:49:08,312][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:49:20,684][__main__][INFO] - Number of regex retries in iteration 127: 0 -[2025-11-09 19:49:20,685][__main__][INFO] - agents played in iteration 127 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:49:21,541][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:49:21,564][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:49:21,586][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:49:21,608][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:49:21,609][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:49:21,609][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:49:22,272][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:49:22,731][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:49:23,233][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:49:23,735][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:49:24,234][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:49:24,734][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:49:25,242][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:49:25,741][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:49:26,253][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:49:26,755][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:49:27,259][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:49:27,764][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:49:28,265][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:49:28,767][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:49:29,267][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:49:29,772][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:49:30,275][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:49:30,775][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:49:31,276][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:49:31,775][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:49:32,274][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:49:32,774][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:49:33,273][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:49:33,771][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:49:34,274][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:49:34,773][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:49:35,270][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:49:35,769][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:49:36,266][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:49:36,771][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:49:37,269][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:49:37,767][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:49:38,267][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:49:38,765][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:49:39,268][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:49:39,768][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:49:40,266][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:49:40,775][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:49:41,273][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:49:41,773][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:49:42,272][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:49:42,771][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:49:43,275][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:49:43,775][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:49:44,275][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:49:44,776][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:49:45,284][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:49:45,790][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:49:46,296][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:49:46,805][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:49:47,305][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:49:47,806][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:49:48,305][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:49:48,808][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:49:49,306][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:49:49,804][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:49:50,303][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:49:50,804][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:49:51,304][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:49:51,802][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:49:52,301][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:49:52,800][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:49:53,298][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:49:53,797][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:49:54,297][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10864 tokens. -[2025-11-09 19:49:54,944][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.27%, Current % of VRAM taken: 55.76%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:32 -[2025-11-09 19:49:55,712][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:49:55,714][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:49:55,715][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:49:57,942][__main__][INFO] - Iteration 128 took 49s (24.93% Gen, 70.58% Train). Generation: 12s, Training: 35s. Estimated remaining time: 39h 30m 31s. Estimated total time: 41h 21m 32s. Time estimates for 10 more iterations: 8m 16s, 100 more iterations: 1h 22m 43s, 500 more iterations: 6h 53m 35s. -[2025-11-09 19:49:57,944][__main__][INFO] - Starting iteration 128. -[2025-11-09 19:49:58,413][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 12 and human policies 1. -[2025-11-09 19:49:58,413][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:50:01,269][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:50:01,369][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:50:03,363][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:50:12,660][__main__][INFO] - Number of regex retries in iteration 128: 3 -[2025-11-09 19:50:12,661][__main__][INFO] - agents played in iteration 128 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:50:13,424][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:50:13,451][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:50:13,488][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:50:13,514][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:50:13,515][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:50:13,515][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:50:14,157][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:50:14,616][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:50:15,123][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:50:15,624][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:50:16,126][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:50:16,626][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:50:17,128][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:50:17,630][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:50:18,131][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:50:18,630][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:50:19,131][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:50:19,634][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:50:20,137][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:50:20,638][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:50:21,137][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:50:21,639][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:50:22,138][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:50:22,637][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:50:23,142][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:50:23,640][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:50:24,137][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:50:24,637][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:50:25,137][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:50:25,653][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:50:26,152][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:50:26,651][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:50:27,154][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:50:27,653][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:50:28,160][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:50:28,659][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:50:29,158][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:50:29,671][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:50:30,175][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:50:30,677][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:50:31,175][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:50:31,673][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:50:32,176][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:50:32,675][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:50:33,175][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:50:33,674][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:50:34,174][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:50:34,680][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:50:35,180][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:50:35,681][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:50:36,181][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:50:36,679][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:50:37,179][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:50:37,681][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:50:38,181][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:50:38,681][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:50:39,179][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:50:39,677][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:50:40,181][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:50:40,677][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:50:41,177][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:50:41,674][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:50:42,174][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:50:42,673][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:50:43,173][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:50:43,668][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:50:44,167][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:50:44,662][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:50:45,158][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:50:45,657][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:50:46,154][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10846 tokens. -[2025-11-09 19:50:46,805][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 55.67%, Block Peak % of device VRAM: 62.35%, ΔTime: 00:00:32 -[2025-11-09 19:50:47,548][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:50:47,550][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:50:47,552][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:50:48,711][__main__][INFO] - Iteration 129 took 50s (28.33% Gen, 69.37% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 3m 6s. Estimated total time: 41h 54m 58s. Time estimates for 10 more iterations: 8m 22s, 100 more iterations: 1h 23m 49s, 500 more iterations: 6h 59m 9s. -[2025-11-09 19:50:48,713][__main__][INFO] - Starting iteration 129. -[2025-11-09 19:50:49,189][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 12 and human policies 1. -[2025-11-09 19:50:49,189][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:50:58,344][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:51:02,948][__main__][INFO] - Number of regex retries in iteration 129: 1 -[2025-11-09 19:51:02,949][__main__][INFO] - agents played in iteration 129 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:51:03,729][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:51:03,756][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:51:03,783][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:51:03,805][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:51:03,806][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:51:03,807][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:51:04,420][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:51:04,877][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:51:05,381][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:51:05,889][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:51:06,394][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:51:06,896][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:51:07,395][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:51:07,894][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:51:08,397][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:51:08,897][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:51:09,399][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:51:09,898][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:51:10,401][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:51:10,899][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:51:11,398][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:51:11,896][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:51:12,392][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:51:12,891][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:51:13,388][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:51:13,887][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:51:14,385][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:51:14,884][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:51:15,381][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:51:15,879][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:51:16,379][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:51:16,889][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:51:17,388][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:51:17,886][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:51:18,395][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:51:18,893][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:51:19,393][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:51:19,892][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:51:20,392][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:51:20,896][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:51:21,398][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:51:21,899][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:51:22,401][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:51:22,898][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:51:23,401][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:51:23,898][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:51:24,398][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:51:24,902][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:51:25,401][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:51:25,899][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:51:26,399][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:51:26,897][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:51:27,402][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:51:27,905][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:51:28,404][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:51:28,901][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:51:29,399][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:51:29,895][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:51:30,392][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:51:30,890][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:51:31,384][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:51:31,882][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:51:32,377][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:51:32,882][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:51:33,381][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:51:33,879][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:51:34,377][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:51:34,877][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:51:35,382][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:51:35,879][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:51:36,376][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10873 tokens. -[2025-11-09 19:51:37,041][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 55.65%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 -[2025-11-09 19:51:37,790][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:51:37,792][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:51:37,794][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:51:38,781][__main__][INFO] - Iteration 130 took 49s (27.74% Gen, 70.26% Train). Generation: 13s, Training: 34s. Estimated remaining time: 39h 26m 55s. Estimated total time: 41h 19m 36s. Time estimates for 10 more iterations: 8m 15s, 100 more iterations: 1h 22m 39s, 500 more iterations: 6h 53m 16s. -[2025-11-09 19:51:38,783][__main__][INFO] - Starting iteration 130. -[2025-11-09 19:51:39,256][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 12 and human policies 1. -[2025-11-09 19:51:39,256][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:51:53,011][__main__][INFO] - Number of regex retries in iteration 130: 0 -[2025-11-09 19:51:53,012][__main__][INFO] - agents played in iteration 130 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:51:53,829][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:51:53,852][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:51:53,874][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:51:53,896][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:51:53,897][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:51:53,898][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:51:54,587][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:51:55,052][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:51:55,560][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:51:56,065][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:51:56,566][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:51:57,069][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:51:57,571][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:51:58,073][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:51:58,575][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:51:59,089][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:51:59,590][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:52:00,090][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:52:00,591][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:52:01,090][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:52:01,588][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:52:02,087][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:52:02,585][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:52:03,087][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:52:03,586][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:52:04,084][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:52:04,584][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:52:05,083][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:52:05,583][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:52:06,081][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:52:06,581][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:52:07,080][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:52:07,579][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:52:08,079][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:52:08,579][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:52:09,078][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:52:09,598][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:52:10,099][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:52:10,602][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:52:11,102][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:52:11,604][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:52:12,109][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:52:12,612][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:52:13,116][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:52:13,617][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:52:14,118][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:52:14,619][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:52:15,117][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:52:15,618][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:52:16,118][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:52:16,617][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:52:17,117][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:52:17,620][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:52:18,118][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:52:18,617][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:52:19,115][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:52:19,613][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:52:20,114][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:52:20,612][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:52:21,108][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:52:21,605][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:52:22,104][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:52:22,606][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:52:23,104][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:52:23,604][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:52:24,103][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:52:24,601][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:52:25,099][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:52:25,597][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:52:26,095][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:52:26,598][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10855 tokens. -[2025-11-09 19:52:27,251][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.25%, Current % of VRAM taken: 55.75%, Block Peak % of device VRAM: 62.41%, ΔTime: 00:00:32 -[2025-11-09 19:52:28,004][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:52:28,006][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:52:28,008][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:52:29,997][__main__][INFO] - Iteration 131 took 50s (27.11% Gen, 68.97% Train). Generation: 13s, Training: 34s. Estimated remaining time: 40h 23m 32s. Estimated total time: 42h 17m 5s. Time estimates for 10 more iterations: 8m 27s, 100 more iterations: 1h 24m 34s, 500 more iterations: 7h 2m 50s. -[2025-11-09 19:52:29,999][__main__][INFO] - Starting iteration 131. -[2025-11-09 19:52:30,493][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 13 and human policies 1. -[2025-11-09 19:52:30,494][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:52:34,769][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:52:44,357][__main__][INFO] - Number of regex retries in iteration 131: 1 -[2025-11-09 19:52:44,358][__main__][INFO] - agents played in iteration 131 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:52:45,152][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:52:45,177][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:52:45,201][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:52:45,224][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:52:45,224][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:52:45,225][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:52:45,904][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:52:46,369][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:52:46,890][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:52:47,395][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:52:47,902][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:52:48,404][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:52:48,906][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:52:49,407][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:52:49,908][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:52:50,413][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:52:50,915][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:52:51,413][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:52:51,911][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:52:52,415][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:52:52,918][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:52:53,421][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:52:53,923][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:52:54,423][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:52:54,928][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:52:55,426][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:52:55,924][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:52:56,424][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:52:56,924][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:52:57,422][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:52:57,920][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:52:58,420][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:52:58,918][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:52:59,415][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:52:59,916][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:53:00,415][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:53:00,915][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:53:01,417][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:53:01,919][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:53:02,418][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:53:02,917][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:53:03,422][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:53:03,925][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:53:04,428][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:53:04,926][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:53:05,427][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:53:05,924][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:53:06,423][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:53:06,921][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:53:07,422][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:53:07,921][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:53:08,419][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:53:08,921][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:53:09,418][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:53:09,922][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:53:10,419][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:53:10,918][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:53:11,420][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:53:11,918][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:53:12,417][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:53:12,916][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:53:13,415][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:53:13,914][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:53:14,414][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:53:14,914][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:53:15,417][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:53:15,917][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:53:16,415][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:53:16,914][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:53:17,414][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:53:17,913][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10854 tokens. -[2025-11-09 19:53:18,554][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 -[2025-11-09 19:53:19,315][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:53:19,317][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:53:19,318][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:53:20,296][__main__][INFO] - Iteration 132 took 49s (27.84% Gen, 70.20% Train). Generation: 13s, Training: 34s. Estimated remaining time: 39h 35m 46s. Estimated total time: 41h 30m 9s. Time estimates for 10 more iterations: 8m 18s, 100 more iterations: 1h 23m 0s, 500 more iterations: 6h 55m 1s. -[2025-11-09 19:53:20,298][__main__][INFO] - Starting iteration 132. -[2025-11-09 19:53:20,766][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 13 and human policies 1. -[2025-11-09 19:53:20,767][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:53:23,946][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:53:34,624][__main__][INFO] - Number of regex retries in iteration 132: 1 -[2025-11-09 19:53:34,624][__main__][INFO] - agents played in iteration 132 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:53:35,411][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:53:35,436][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:53:35,461][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:53:35,484][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:53:35,484][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:53:35,486][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:53:36,160][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:53:36,622][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:53:37,129][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:53:37,633][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:53:38,136][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:53:38,636][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:53:39,138][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:53:39,638][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:53:40,137][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:53:40,639][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:53:41,142][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:53:41,644][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:53:42,142][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:53:42,640][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:53:43,146][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:53:43,649][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:53:44,150][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:53:44,651][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:53:45,147][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:53:45,646][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:53:46,144][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:53:46,641][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:53:47,139][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:53:47,636][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:53:48,137][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:53:48,640][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:53:49,139][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:53:49,643][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:53:50,141][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:53:50,640][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:53:51,139][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:53:51,640][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:53:52,142][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:53:52,638][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:53:53,136][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:53:53,635][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:53:54,132][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:53:54,629][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:53:55,128][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:53:55,628][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:53:56,134][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:53:56,637][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:53:57,137][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:53:57,637][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:53:58,141][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:53:58,642][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:53:59,141][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:53:59,642][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:54:00,146][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:54:00,647][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:54:01,145][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:54:01,646][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:54:02,142][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:54:02,638][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:54:03,138][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:54:03,632][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:54:04,129][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:54:04,623][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:54:05,120][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:54:05,617][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:54:06,110][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:54:06,607][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:54:07,100][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:54:07,598][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:54:08,096][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10855 tokens. -[2025-11-09 19:54:08,742][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 55.64%, Block Peak % of device VRAM: 62.42%, ΔTime: 00:00:32 -[2025-11-09 19:54:09,494][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:54:09,495][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:54:09,497][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:54:10,471][__main__][INFO] - Iteration 133 took 49s (27.88% Gen, 70.16% Train). Generation: 13s, Training: 34s. Estimated remaining time: 39h 30m 1s. Estimated total time: 41h 25m 14s. Time estimates for 10 more iterations: 8m 17s, 100 more iterations: 1h 22m 50s, 500 more iterations: 6h 54m 12s. -[2025-11-09 19:54:10,473][__main__][INFO] - Starting iteration 133. -[2025-11-09 19:54:10,934][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 13 and human policies 1. -[2025-11-09 19:54:10,935][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:54:13,818][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:54:24,315][__main__][INFO] - Number of regex retries in iteration 133: 1 -[2025-11-09 19:54:24,316][__main__][INFO] - agents played in iteration 133 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:54:25,121][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:54:25,148][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:54:25,174][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:54:25,197][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:54:25,198][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:54:25,199][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:54:25,830][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:54:26,288][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:54:26,802][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:54:27,304][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:54:27,818][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:54:28,318][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:54:28,818][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:54:29,318][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:54:29,818][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:54:30,325][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:54:30,824][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:54:31,326][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:54:31,837][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:54:32,340][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:54:32,841][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:54:33,337][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:54:33,837][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:54:34,339][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:54:34,840][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:54:35,343][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:54:35,846][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:54:36,347][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:54:36,847][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:54:37,345][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:54:37,843][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:54:38,346][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:54:38,844][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:54:39,345][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:54:39,842][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:54:40,338][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:54:40,837][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:54:41,336][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:54:41,837][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:54:42,341][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:54:42,840][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:54:43,341][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:54:43,839][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:54:44,336][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:54:44,839][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:54:45,337][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:54:45,836][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:54:46,337][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:54:46,837][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:54:47,344][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:54:47,845][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:54:48,345][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:54:48,846][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:54:49,344][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:54:49,845][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:54:50,343][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:54:50,842][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:54:51,345][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:54:51,844][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:54:52,341][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:54:52,838][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:54:53,335][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:54:53,833][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:54:54,330][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:54:54,826][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:54:55,339][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:54:55,835][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:54:56,331][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:54:56,832][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:54:57,330][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:54:57,827][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10846 tokens. -[2025-11-09 19:54:58,469][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 55.72%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:32 -[2025-11-09 19:54:59,228][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:54:59,229][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:54:59,231][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:55:00,155][__main__][INFO] - Iteration 134 took 49s (27.19% Gen, 70.93% Train). Generation: 13s, Training: 34s. Estimated remaining time: 39h 5m 2s. Estimated total time: 41h 1m 5s. Time estimates for 10 more iterations: 8m 12s, 100 more iterations: 1h 22m 2s, 500 more iterations: 6h 50m 10s. -[2025-11-09 19:55:00,158][__main__][INFO] - Starting iteration 134. -[2025-11-09 19:55:00,623][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 13 and human policies 1. -[2025-11-09 19:55:00,624][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:55:13,970][__main__][INFO] - Number of regex retries in iteration 134: 0 -[2025-11-09 19:55:13,970][__main__][INFO] - agents played in iteration 134 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:55:14,756][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:55:14,779][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:55:14,802][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:55:14,824][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:55:14,825][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:55:14,826][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:55:15,478][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:55:15,935][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:55:16,439][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:55:16,935][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:55:17,432][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:55:17,934][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:55:18,432][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:55:18,932][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:55:19,433][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:55:19,933][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:55:20,433][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:55:20,931][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:55:21,431][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:55:21,933][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:55:22,434][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:55:22,935][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:55:23,435][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:55:23,939][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:55:24,438][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:55:24,946][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:55:25,445][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:55:25,945][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:55:26,446][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:55:26,946][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:55:27,445][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:55:27,944][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:55:28,441][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:55:28,963][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:55:29,466][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:55:29,964][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:55:30,467][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:55:30,967][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:55:31,468][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:55:31,966][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:55:32,462][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:55:32,958][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:55:33,456][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:55:33,952][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:55:34,447][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:55:34,947][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:55:35,449][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:55:35,949][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:55:36,449][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:55:36,949][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:55:37,448][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:55:37,950][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:55:38,452][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:55:38,953][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:55:39,456][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:55:39,955][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:55:40,456][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:55:40,957][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:55:41,455][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:55:41,954][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:55:42,449][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:55:42,947][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:55:43,445][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:55:43,942][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:55:44,439][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:55:44,937][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:55:45,435][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:55:45,944][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:55:46,441][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:55:46,939][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:55:47,438][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10860 tokens. -[2025-11-09 19:55:48,073][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 -[2025-11-09 19:55:48,834][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:55:48,836][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:55:48,838][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:55:49,770][__main__][INFO] - Iteration 135 took 49s (27.16% Gen, 70.94% Train). Generation: 13s, Training: 34s. Estimated remaining time: 39h 0m 28s. Estimated total time: 40h 57m 21s. Time estimates for 10 more iterations: 8m 11s, 100 more iterations: 1h 21m 54s, 500 more iterations: 6h 49m 33s. -[2025-11-09 19:55:49,772][__main__][INFO] - Starting iteration 135. -[2025-11-09 19:55:50,244][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 13 and human policies 1. -[2025-11-09 19:55:50,245][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:56:04,741][__main__][INFO] - Number of regex retries in iteration 135: 0 -[2025-11-09 19:56:04,742][__main__][INFO] - agents played in iteration 135 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:56:05,517][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:56:05,540][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:56:05,562][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:56:05,584][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:56:05,585][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:56:05,585][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:56:06,215][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:56:06,673][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:56:07,187][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:56:07,687][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:56:08,185][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:56:08,683][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:56:09,179][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:56:09,681][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:56:10,179][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:56:10,676][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:56:11,175][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:56:11,669][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:56:12,169][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:56:12,662][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:56:13,164][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:56:13,664][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:56:14,162][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:56:14,662][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:56:15,165][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:56:15,663][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:56:16,167][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:56:16,669][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:56:17,170][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:56:17,672][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:56:18,173][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:56:18,672][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:56:19,176][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:56:19,677][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:56:20,175][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:56:20,673][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:56:21,168][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:56:21,665][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:56:22,160][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:56:22,659][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:56:23,158][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:56:23,656][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:56:24,155][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:56:24,656][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:56:25,156][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:56:25,655][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:56:26,156][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:56:26,679][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:56:27,186][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:56:27,691][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:56:28,195][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:56:28,697][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:56:29,201][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:56:29,709][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:56:30,210][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:56:30,709][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:56:31,216][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:56:31,717][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:56:32,233][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:56:32,735][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:56:33,236][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:56:33,739][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:56:34,237][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:56:34,736][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:56:35,235][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:56:35,734][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:56:36,248][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:56:36,747][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:56:37,247][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:56:37,748][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:56:38,248][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10876 tokens. -[2025-11-09 19:56:38,940][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.31%, Current % of VRAM taken: 55.80%, Block Peak % of device VRAM: 62.41%, ΔTime: 00:00:32 -[2025-11-09 19:56:39,708][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:56:39,709][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:56:39,711][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:56:40,725][__main__][INFO] - Iteration 136 took 50s (28.72% Gen, 69.27% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 6m 20s. Estimated total time: 42h 4m 3s. Time estimates for 10 more iterations: 8m 24s, 100 more iterations: 1h 24m 8s, 500 more iterations: 7h 0m 40s. -[2025-11-09 19:56:40,727][__main__][INFO] - Starting iteration 136. -[2025-11-09 19:56:41,188][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 13 and human policies 1. -[2025-11-09 19:56:41,189][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:56:49,152][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:56:55,026][__main__][INFO] - Number of regex retries in iteration 136: 1 -[2025-11-09 19:56:55,026][__main__][INFO] - agents played in iteration 136 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:56:55,817][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:56:55,843][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:56:55,873][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:56:55,902][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:56:55,903][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:56:55,904][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:56:56,551][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:56:57,010][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:56:57,515][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:56:58,014][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:56:58,512][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:56:59,009][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:56:59,512][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:57:00,014][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:57:00,513][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:57:01,014][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:57:01,517][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:57:02,015][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:57:02,512][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:57:03,010][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:57:03,504][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:57:04,001][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:57:04,498][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:57:04,996][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:57:05,497][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:57:05,996][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:57:06,499][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:57:06,998][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:57:07,499][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:57:08,003][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:57:08,502][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:57:09,005][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:57:09,505][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:57:10,004][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:57:10,505][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:57:11,004][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:57:11,501][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:57:12,002][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:57:12,499][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:57:12,996][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:57:13,493][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:57:13,993][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:57:14,495][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:57:14,995][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:57:15,496][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:57:15,998][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:57:16,497][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:57:16,997][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:57:17,498][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:57:17,999][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:57:18,502][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:57:19,000][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:57:19,498][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:57:19,997][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:57:20,491][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:57:20,993][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:57:21,494][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:57:21,995][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:57:22,512][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:57:23,011][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:57:23,511][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:57:24,016][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:57:24,517][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:57:25,021][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:57:25,520][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:57:26,018][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:57:26,518][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:57:27,016][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:57:27,515][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:57:28,013][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:57:28,513][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10856 tokens. -[2025-11-09 19:57:29,161][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.28%, Current % of VRAM taken: 55.78%, Block Peak % of device VRAM: 62.35%, ΔTime: 00:00:32 -[2025-11-09 19:57:29,912][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:57:29,914][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:57:29,915][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:57:30,874][__main__][INFO] - Iteration 137 took 49s (27.85% Gen, 70.22% Train). Generation: 13s, Training: 34s. Estimated remaining time: 39h 25m 45s. Estimated total time: 41h 24m 19s. Time estimates for 10 more iterations: 8m 16s, 100 more iterations: 1h 22m 48s, 500 more iterations: 6h 54m 3s. -[2025-11-09 19:57:30,876][__main__][INFO] - Starting iteration 137. -[2025-11-09 19:57:31,343][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 13 and human policies 1. -[2025-11-09 19:57:31,345][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:57:44,989][__main__][INFO] - Number of regex retries in iteration 137: 0 -[2025-11-09 19:57:44,990][__main__][INFO] - agents played in iteration 137 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:57:45,776][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:57:45,799][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:57:45,822][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:57:45,844][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:57:45,845][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:57:45,846][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:57:46,488][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:57:46,943][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:57:47,461][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:57:47,959][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:57:48,457][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:57:48,956][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:57:49,454][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:57:49,968][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:57:50,466][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:57:50,963][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:57:51,460][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:57:51,959][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:57:52,457][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:57:52,955][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:57:53,453][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:57:53,949][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:57:54,445][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:57:54,944][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:57:55,442][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:57:55,939][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:57:56,451][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:57:56,948][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:57:57,454][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:57:57,958][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:57:58,460][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:57:58,960][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:57:59,460][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:57:59,956][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:58:00,472][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:58:00,973][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:58:01,475][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:58:01,974][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:58:02,475][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:58:02,976][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:58:03,474][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:58:03,972][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:58:04,473][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:58:04,975][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:58:05,475][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:58:05,976][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:58:06,477][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:58:06,974][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:58:07,475][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:58:07,975][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:58:08,479][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:58:08,979][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:58:09,484][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:58:09,983][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:58:10,481][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:58:10,979][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:58:11,480][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:58:11,985][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:58:12,483][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:58:12,980][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:58:13,480][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:58:13,978][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:58:14,475][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:58:14,976][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:58:15,476][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:58:15,976][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:58:16,478][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:58:16,976][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:58:17,474][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:58:17,972][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:58:18,477][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10852 tokens. -[2025-11-09 19:58:19,130][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 55.70%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 -[2025-11-09 19:58:19,898][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:58:19,899][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:58:19,901][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:58:20,872][__main__][INFO] - Iteration 138 took 49s (27.55% Gen, 70.49% Train). Generation: 13s, Training: 34s. Estimated remaining time: 39h 17m 2s. Estimated total time: 41h 16m 26s. Time estimates for 10 more iterations: 8m 15s, 100 more iterations: 1h 22m 32s, 500 more iterations: 6h 52m 44s. -[2025-11-09 19:58:20,874][__main__][INFO] - Starting iteration 138. -[2025-11-09 19:58:21,345][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 13 and human policies 1. -[2025-11-09 19:58:21,346][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:58:24,559][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:58:25,662][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:58:35,208][__main__][INFO] - Number of regex retries in iteration 138: 2 -[2025-11-09 19:58:35,209][__main__][INFO] - agents played in iteration 138 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:58:35,999][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:58:36,023][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:58:36,045][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:58:36,067][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:58:36,069][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:58:36,069][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:58:36,689][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:58:37,144][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:58:37,650][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:58:38,150][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:58:38,650][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:58:39,150][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:58:39,650][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:58:40,148][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:58:40,647][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:58:41,145][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:58:41,643][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:58:42,142][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:58:42,653][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:58:43,155][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:58:43,662][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:58:44,161][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:58:44,660][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:58:45,159][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:58:45,658][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:58:46,156][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:58:46,657][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:58:47,157][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:58:47,664][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:58:48,165][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:58:48,663][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:58:49,164][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:58:49,666][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:58:50,177][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:58:50,680][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:58:51,177][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:58:51,687][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:58:52,190][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:58:52,690][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:58:53,196][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:58:53,699][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:58:54,203][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:58:54,706][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:58:55,207][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:58:55,709][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:58:56,212][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:58:56,714][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:58:57,217][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:58:57,717][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:58:58,219][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:58:58,720][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:58:59,220][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:58:59,721][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:59:00,220][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:59:00,719][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:59:01,223][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:59:01,721][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:59:02,224][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:59:02,724][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:59:03,222][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:59:03,723][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:59:04,223][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:59:04,723][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:59:05,222][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:59:05,723][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:59:06,226][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:59:06,725][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:59:07,224][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:59:07,725][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:59:08,227][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:59:08,728][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10858 tokens. -[2025-11-09 19:59:09,416][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.34%, ΔTime: 00:00:32 -[2025-11-09 19:59:10,163][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 19:59:10,165][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 19:59:10,166][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 19:59:11,158][__main__][INFO] - Iteration 139 took 49s (27.83% Gen, 70.18% Train). Generation: 13s, Training: 34s. Estimated remaining time: 39h 30m 26s. Estimated total time: 41h 30m 41s. Time estimates for 10 more iterations: 8m 18s, 100 more iterations: 1h 23m 1s, 500 more iterations: 6h 55m 6s. -[2025-11-09 19:59:11,160][__main__][INFO] - Starting iteration 139. -[2025-11-09 19:59:11,642][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 13 and human policies 1. -[2025-11-09 19:59:11,644][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 19:59:15,176][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 30 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:59:21,107][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 19:59:25,337][__main__][INFO] - Number of regex retries in iteration 139: 2 -[2025-11-09 19:59:25,337][__main__][INFO] - agents played in iteration 139 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 19:59:26,119][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:59:26,145][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:59:26,171][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:59:26,194][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 19:59:26,195][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 19:59:26,196][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 19:59:26,808][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 19:59:27,268][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 19:59:27,769][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 19:59:28,266][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 19:59:28,777][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 19:59:29,277][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 19:59:29,777][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 19:59:30,282][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 19:59:30,779][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 19:59:31,289][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 19:59:31,786][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 19:59:32,284][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 19:59:32,794][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 19:59:33,290][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 19:59:33,789][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 19:59:34,305][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 19:59:34,801][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 19:59:35,303][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 19:59:35,805][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 19:59:36,303][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 19:59:36,808][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 19:59:37,306][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 19:59:37,807][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 19:59:38,309][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 19:59:38,807][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 19:59:39,308][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 19:59:39,805][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 19:59:40,301][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 19:59:40,806][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 19:59:41,308][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 19:59:41,808][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 19:59:42,307][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 19:59:42,806][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 19:59:43,306][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 19:59:43,804][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 19:59:44,305][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 19:59:44,807][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 19:59:45,311][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 19:59:45,812][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 19:59:46,312][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 19:59:46,811][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 19:59:47,312][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 19:59:47,808][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 19:59:48,305][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 19:59:48,811][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 19:59:49,307][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 19:59:49,803][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 19:59:50,300][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 19:59:50,795][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 19:59:51,305][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 19:59:51,802][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 19:59:52,298][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 19:59:52,806][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 19:59:53,302][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 19:59:53,798][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 19:59:54,297][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 19:59:54,795][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 19:59:55,301][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 19:59:55,799][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 19:59:56,296][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 19:59:56,793][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 19:59:57,291][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 19:59:57,789][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 19:59:58,285][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 19:59:58,782][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10849 tokens. -[2025-11-09 19:59:59,509][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 55.68%, Block Peak % of device VRAM: 62.35%, ΔTime: 00:00:32 -[2025-11-09 20:00:00,262][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:00:00,263][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:00:00,265][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:00:01,307][__main__][INFO] - Iteration 140 took 49s (27.57% Gen, 70.33% Train). Generation: 13s, Training: 34s. Estimated remaining time: 39h 22m 12s. Estimated total time: 41h 23m 16s. Time estimates for 10 more iterations: 8m 16s, 100 more iterations: 1h 22m 46s, 500 more iterations: 6h 53m 52s. -[2025-11-09 20:00:01,309][__main__][INFO] - Starting iteration 140. -[2025-11-09 20:00:01,778][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 13 and human policies 1. -[2025-11-09 20:00:01,778][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:00:15,164][__main__][INFO] - Number of regex retries in iteration 140: 0 -[2025-11-09 20:00:15,165][__main__][INFO] - agents played in iteration 140 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:00:15,959][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:00:15,982][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:00:16,009][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:00:16,045][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:00:16,045][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:00:16,046][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:00:16,698][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:00:17,152][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:00:17,658][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:00:18,157][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:00:18,654][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:00:19,156][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:00:19,653][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:00:20,153][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:00:20,650][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:00:21,149][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:00:21,647][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:00:22,144][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:00:22,642][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:00:23,140][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:00:23,641][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:00:24,141][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:00:24,639][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:00:25,136][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:00:25,637][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:00:26,134][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:00:26,634][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:00:27,132][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:00:27,631][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:00:28,133][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:00:28,633][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:00:29,135][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:00:29,634][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:00:30,128][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:00:30,625][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:00:31,122][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:00:31,620][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:00:32,118][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:00:32,616][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:00:33,116][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:00:33,618][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:00:34,120][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:00:34,625][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:00:35,129][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:00:35,630][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:00:36,131][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:00:36,632][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:00:37,140][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:00:37,640][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:00:38,141][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:00:38,643][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:00:39,143][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:00:39,642][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:00:40,139][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:00:40,639][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:00:41,136][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:00:41,634][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:00:42,133][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:00:42,630][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:00:43,128][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:00:43,627][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:00:44,124][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:00:44,627][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:00:45,135][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:00:45,634][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:00:46,133][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:00:46,639][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:00:47,138][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:00:47,637][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:00:48,137][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:00:48,636][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10871 tokens. -[2025-11-09 20:00:49,303][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 55.71%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 -[2025-11-09 20:00:50,075][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:00:50,077][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:00:50,079][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:00:52,451][__main__][INFO] - Iteration 141 took 50s (26.42% Gen, 68.90% Train). Generation: 13s, Training: 34s. Estimated remaining time: 40h 11m 47s. Estimated total time: 42h 13m 42s. Time estimates for 10 more iterations: 8m 26s, 100 more iterations: 1h 24m 27s, 500 more iterations: 7h 2m 17s. -[2025-11-09 20:00:52,454][__main__][INFO] - Starting iteration 141. -[2025-11-09 20:00:52,919][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 14 and human policies 1. -[2025-11-09 20:00:52,920][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:01:07,724][__main__][INFO] - Number of regex retries in iteration 141: 0 -[2025-11-09 20:01:07,725][__main__][INFO] - agents played in iteration 141 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:01:08,512][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:01:08,547][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:01:08,572][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:01:08,594][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:01:08,595][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:01:08,596][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:01:09,241][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:01:09,697][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:01:10,199][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:01:10,699][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:01:11,197][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:01:11,696][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:01:12,195][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:01:12,699][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:01:13,197][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:01:13,698][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:01:14,198][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:01:14,698][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:01:15,196][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:01:15,695][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:01:16,190][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:01:16,701][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:01:17,201][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:01:17,700][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:01:18,207][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:01:18,704][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:01:19,201][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:01:19,702][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:01:20,200][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:01:20,705][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:01:21,205][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:01:21,702][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:01:22,203][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:01:22,699][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:01:23,196][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:01:23,696][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:01:24,197][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:01:24,698][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:01:25,199][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:01:25,697][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:01:26,207][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:01:26,711][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:01:27,211][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:01:27,709][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:01:28,209][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:01:28,726][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:01:29,223][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:01:29,724][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:01:30,228][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:01:30,781][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:01:31,279][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:01:31,778][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:01:32,276][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:01:32,775][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:01:33,273][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:01:33,776][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:01:34,275][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:01:34,772][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:01:35,271][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:01:35,769][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:01:36,268][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:01:36,767][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:01:37,265][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:01:37,762][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:01:38,259][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:01:38,760][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:01:39,260][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:01:39,757][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:01:40,256][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:01:40,758][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:01:41,258][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10859 tokens. -[2025-11-09 20:01:41,912][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 55.69%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:32 -[2025-11-09 20:01:42,655][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:01:42,656][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:01:42,658][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:01:43,593][__main__][INFO] - Iteration 142 took 50s (29.22% Gen, 68.94% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 10m 55s. Estimated total time: 42h 13m 41s. Time estimates for 10 more iterations: 8m 26s, 100 more iterations: 1h 24m 27s, 500 more iterations: 7h 2m 16s. -[2025-11-09 20:01:43,595][__main__][INFO] - Starting iteration 142. -[2025-11-09 20:01:44,065][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 14 and human policies 1. -[2025-11-09 20:01:44,065][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:01:58,026][__main__][INFO] - Number of regex retries in iteration 142: 0 -[2025-11-09 20:01:58,027][__main__][INFO] - agents played in iteration 142 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:01:58,819][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:01:58,856][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:01:58,882][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:01:58,903][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:01:58,904][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:01:58,905][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:01:59,560][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:02:00,014][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:02:00,519][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:02:01,022][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:02:01,519][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:02:02,019][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:02:02,516][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:02:03,032][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:02:03,529][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:02:04,030][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:02:04,532][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:02:05,030][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:02:05,538][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:02:06,035][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:02:06,537][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:02:07,040][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:02:07,538][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:02:08,040][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:02:08,538][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:02:09,040][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:02:09,551][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:02:10,049][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:02:10,550][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:02:11,049][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:02:11,549][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:02:12,045][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:02:12,539][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:02:13,035][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:02:13,533][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:02:14,033][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:02:14,536][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:02:15,037][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:02:15,537][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:02:16,052][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:02:16,550][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:02:17,053][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:02:17,549][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:02:18,049][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:02:18,549][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:02:19,047][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:02:19,549][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:02:20,049][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:02:20,553][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:02:21,060][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:02:21,557][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:02:22,058][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:02:22,562][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:02:23,063][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:02:23,564][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:02:24,065][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:02:24,566][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:02:25,066][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:02:25,565][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:02:26,068][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:02:26,570][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:02:27,069][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:02:27,568][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:02:28,071][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:02:28,569][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:02:29,069][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:02:29,565][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:02:30,063][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:02:30,566][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:02:31,065][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:02:31,566][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10863 tokens. -[2025-11-09 20:02:32,226][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 55.66%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:32 -[2025-11-09 20:02:32,975][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:02:32,979][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:02:32,981][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:02:34,026][__main__][INFO] - Iteration 143 took 49s (27.95% Gen, 69.96% Train). Generation: 13s, Training: 34s. Estimated remaining time: 39h 34m 27s. Estimated total time: 41h 38m 4s. Time estimates for 10 more iterations: 8m 19s, 100 more iterations: 1h 23m 16s, 500 more iterations: 6h 56m 20s. -[2025-11-09 20:02:34,028][__main__][INFO] - Starting iteration 143. -[2025-11-09 20:02:34,514][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 14 and human policies 1. -[2025-11-09 20:02:34,514][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:02:39,203][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:02:40,684][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:02:48,082][__main__][INFO] - Number of regex retries in iteration 143: 2 -[2025-11-09 20:02:48,082][__main__][INFO] - agents played in iteration 143 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:02:48,884][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:02:48,906][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:02:48,928][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:02:48,950][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:02:48,951][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:02:48,952][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:02:49,590][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:02:50,048][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:02:50,550][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:02:51,051][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:02:51,558][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:02:52,057][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:02:52,560][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:02:53,058][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:02:53,558][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:02:54,060][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:02:54,557][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:02:55,056][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:02:55,557][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:02:56,056][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:02:56,554][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:02:57,054][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:02:57,554][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:02:58,059][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:02:58,560][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:02:59,065][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:02:59,567][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:03:00,065][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:03:00,564][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:03:01,064][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:03:01,565][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:03:02,071][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:03:02,573][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:03:03,072][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:03:03,574][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:03:04,079][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:03:04,584][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:03:05,088][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:03:05,588][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:03:06,089][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:03:06,592][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:03:07,093][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:03:07,596][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:03:08,096][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:03:08,598][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:03:09,098][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:03:09,597][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:03:10,116][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:03:10,619][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:03:11,120][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:03:11,621][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:03:12,123][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:03:12,624][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:03:13,123][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:03:13,620][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:03:14,120][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:03:14,619][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:03:15,121][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:03:15,621][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:03:16,121][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:03:16,620][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:03:17,120][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:03:17,633][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:03:18,130][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:03:18,631][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:03:19,144][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:03:19,643][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:03:20,151][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:03:20,650][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:03:21,151][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:03:21,656][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10828 tokens. -[2025-11-09 20:03:22,326][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 55.74%, Block Peak % of device VRAM: 62.35%, ΔTime: 00:00:32 -[2025-11-09 20:03:23,074][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:03:23,075][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:03:23,077][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:03:24,279][__main__][INFO] - Iteration 144 took 49s (27.26% Gen, 70.32% Train). Generation: 13s, Training: 34s. Estimated remaining time: 39h 23m 52s. Estimated total time: 41h 28m 19s. Time estimates for 10 more iterations: 8m 17s, 100 more iterations: 1h 22m 56s, 500 more iterations: 6h 54m 43s. -[2025-11-09 20:03:24,282][__main__][INFO] - Starting iteration 144. -[2025-11-09 20:03:24,745][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 14 and human policies 1. -[2025-11-09 20:03:24,746][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:03:38,215][__main__][INFO] - Number of regex retries in iteration 144: 0 -[2025-11-09 20:03:38,216][__main__][INFO] - agents played in iteration 144 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:03:38,997][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:03:39,020][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:03:39,043][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:03:39,065][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:03:39,066][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:03:39,066][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:03:39,718][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:03:40,174][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:03:40,677][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:03:41,181][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:03:41,679][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:03:42,180][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:03:42,677][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:03:43,176][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:03:43,678][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:03:44,175][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:03:44,675][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:03:45,174][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:03:45,672][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:03:46,170][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:03:46,670][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:03:47,168][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:03:47,664][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:03:48,165][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:03:48,664][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:03:49,177][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:03:49,677][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:03:50,179][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:03:50,679][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:03:51,179][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:03:51,687][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:03:52,191][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:03:52,692][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:03:53,192][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:03:53,689][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:03:54,189][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:03:54,688][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:03:55,186][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:03:55,684][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:03:56,184][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:03:56,681][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:03:57,186][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:03:57,686][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:03:58,187][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:03:58,687][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:03:59,187][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:03:59,690][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:04:00,183][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:04:00,681][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:04:01,179][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:04:01,677][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:04:02,174][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:04:02,679][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:04:03,182][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:04:03,679][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:04:04,177][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:04:04,681][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:04:05,180][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:04:05,680][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:04:06,180][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:04:06,681][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:04:07,179][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:04:07,681][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:04:08,182][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:04:08,683][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:04:09,190][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:04:09,692][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:04:10,192][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:04:10,694][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:04:11,191][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:04:11,691][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10853 tokens. -[2025-11-09 20:04:12,398][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 55.72%, Block Peak % of device VRAM: 62.50%, ΔTime: 00:00:32 -[2025-11-09 20:04:13,160][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:04:13,162][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:04:13,163][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:04:14,175][__main__][INFO] - Iteration 145 took 49s (27.25% Gen, 70.70% Train). Generation: 13s, Training: 34s. Estimated remaining time: 39h 6m 16s. Estimated total time: 41h 11m 33s. Time estimates for 10 more iterations: 8m 14s, 100 more iterations: 1h 22m 23s, 500 more iterations: 6h 51m 55s. -[2025-11-09 20:04:14,177][__main__][INFO] - Starting iteration 145. -[2025-11-09 20:04:14,664][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 14 and human policies 1. -[2025-11-09 20:04:14,665][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:04:17,804][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:04:28,118][__main__][INFO] - Number of regex retries in iteration 145: 1 -[2025-11-09 20:04:28,118][__main__][INFO] - agents played in iteration 145 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:04:28,923][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:04:28,945][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:04:28,968][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:04:28,990][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:04:28,991][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:04:28,991][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:04:29,647][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:04:30,103][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:04:30,614][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:04:31,113][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:04:31,613][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:04:32,115][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:04:32,614][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:04:33,114][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:04:33,619][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:04:34,118][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:04:34,618][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:04:35,116][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:04:35,616][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:04:36,121][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:04:36,621][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:04:37,121][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:04:37,618][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:04:38,120][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:04:38,631][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:04:39,129][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:04:39,629][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:04:40,140][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:04:40,643][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:04:41,164][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:04:41,664][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:04:42,165][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:04:42,679][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:04:43,178][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:04:43,676][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:04:44,177][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:04:44,674][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:04:45,175][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:04:45,676][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:04:46,176][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:04:46,684][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:04:47,185][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:04:47,684][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:04:48,182][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:04:48,681][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:04:49,182][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:04:49,681][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:04:50,181][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:04:50,683][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:04:51,183][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:04:51,681][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:04:52,178][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:04:52,676][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:04:53,178][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:04:53,675][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:04:54,173][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:04:54,668][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:04:55,164][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:04:55,660][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:04:56,157][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:04:56,654][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:04:57,154][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:04:57,654][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:04:58,154][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:04:58,655][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:04:59,156][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:04:59,665][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:05:00,166][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:05:00,667][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:05:01,168][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:05:01,670][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10871 tokens. -[2025-11-09 20:05:02,398][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 55.69%, Block Peak % of device VRAM: 62.21%, ΔTime: 00:00:32 -[2025-11-09 20:05:03,144][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:05:03,146][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:05:03,147][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:05:08,438][__main__][INFO] - Iteration 146 took 53s (25.02% Gen, 65.14% Train). Generation: 13s, Training: 35s. Estimated remaining time: 42h 42m 32s. Estimated total time: 44h 48m 43s. Time estimates for 10 more iterations: 8m 57s, 100 more iterations: 1h 29m 37s, 500 more iterations: 7h 28m 7s. -[2025-11-09 20:05:08,440][__main__][INFO] - Starting iteration 146. -[2025-11-09 20:05:08,929][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 14 and human policies 1. -[2025-11-09 20:05:08,930][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:05:18,532][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:05:23,433][__main__][INFO] - Number of regex retries in iteration 146: 1 -[2025-11-09 20:05:23,433][__main__][INFO] - agents played in iteration 146 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:05:24,201][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:05:24,229][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:05:24,255][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:05:24,278][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:05:24,278][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:05:24,279][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:05:24,909][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:05:25,363][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:05:25,883][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:05:26,385][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:05:26,887][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:05:27,387][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:05:27,886][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:05:28,390][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:05:28,889][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:05:29,396][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:05:30,006][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:05:30,506][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:05:31,006][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:05:31,504][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:05:32,008][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:05:32,509][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:05:33,012][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:05:33,509][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:05:34,004][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:05:34,500][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:05:34,999][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:05:35,498][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:05:35,998][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:05:36,498][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:05:36,999][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:05:37,500][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:05:38,000][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:05:38,497][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:05:38,999][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:05:39,499][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:05:40,002][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:05:40,501][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:05:41,000][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:05:41,502][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:05:42,001][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:05:42,499][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:05:43,001][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:05:43,500][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:05:44,002][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:05:44,504][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:05:45,009][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:05:45,508][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:05:46,006][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:05:46,504][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:05:47,003][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:05:47,501][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:05:48,000][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:05:48,499][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:05:49,001][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:05:49,507][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:05:50,006][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:05:50,507][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:05:51,005][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:05:51,502][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:05:52,006][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:05:52,507][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:05:53,009][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:05:53,512][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:05:54,014][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:05:54,515][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:05:55,014][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:05:55,514][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:05:56,013][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:05:56,513][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:05:57,014][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10858 tokens. -[2025-11-09 20:05:57,679][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 55.72%, Block Peak % of device VRAM: 62.59%, ΔTime: 00:00:32 -[2025-11-09 20:05:58,447][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:05:58,449][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:05:58,451][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:05:59,518][__main__][INFO] - Iteration 147 took 50s (28.67% Gen, 69.22% Train). Generation: 14s, Training: 35s. Estimated remaining time: 40h 2m 24s. Estimated total time: 42h 9m 26s. Time estimates for 10 more iterations: 8m 25s, 100 more iterations: 1h 24m 18s, 500 more iterations: 7h 1m 34s. -[2025-11-09 20:05:59,520][__main__][INFO] - Starting iteration 147. -[2025-11-09 20:05:59,988][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 14 and human policies 1. -[2025-11-09 20:05:59,988][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:06:14,154][__main__][INFO] - Number of regex retries in iteration 147: 0 -[2025-11-09 20:06:14,155][__main__][INFO] - agents played in iteration 147 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:06:14,930][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:06:14,953][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:06:14,976][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:06:14,998][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:06:14,998][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:06:14,999][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:06:15,645][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:06:16,101][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:06:16,607][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:06:17,108][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:06:17,606][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:06:18,106][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:06:18,602][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:06:19,100][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:06:19,606][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:06:20,108][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:06:20,609][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:06:21,113][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:06:21,613][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:06:22,113][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:06:22,615][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:06:23,114][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:06:23,614][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:06:24,110][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:06:24,605][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:06:25,101][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:06:25,600][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:06:26,106][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:06:26,605][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:06:27,103][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:06:27,604][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:06:28,101][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:06:28,602][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:06:29,102][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:06:29,600][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:06:30,109][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:06:30,609][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:06:31,105][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:06:31,603][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:06:32,098][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:06:32,604][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:06:33,101][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:06:33,598][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:06:34,111][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:06:34,608][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:06:35,105][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:06:35,603][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:06:36,099][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:06:36,606][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:06:37,103][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:06:37,601][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:06:38,111][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:06:38,608][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:06:39,106][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:06:39,603][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:06:40,105][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:06:40,607][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:06:41,106][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:06:41,607][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:06:42,118][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:06:42,619][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:06:43,122][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:06:43,622][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:06:44,119][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:06:44,625][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:06:45,125][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:06:45,623][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:06:46,120][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:06:46,617][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:06:47,120][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:06:47,616][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10870 tokens. -[2025-11-09 20:06:48,276][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 55.67%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 -[2025-11-09 20:06:49,035][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:06:49,036][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:06:49,038][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:06:50,010][__main__][INFO] - Iteration 148 took 50s (28.32% Gen, 69.73% Train). Generation: 14s, Training: 34s. Estimated remaining time: 39h 33m 14s. Estimated total time: 41h 41m 8s. Time estimates for 10 more iterations: 8m 20s, 100 more iterations: 1h 23m 22s, 500 more iterations: 6h 56m 51s. -[2025-11-09 20:06:50,012][__main__][INFO] - Starting iteration 148. -[2025-11-09 20:06:50,476][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 14 and human policies 1. -[2025-11-09 20:06:50,476][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:07:04,913][__main__][INFO] - Number of regex retries in iteration 148: 0 -[2025-11-09 20:07:04,913][__main__][INFO] - agents played in iteration 148 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:07:05,692][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:07:05,715][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:07:05,737][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:07:05,759][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:07:05,760][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:07:05,761][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:07:06,377][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:07:06,833][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:07:07,335][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:07:07,831][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:07:08,335][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:07:08,835][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:07:09,334][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:07:09,834][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:07:10,335][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:07:10,837][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:07:11,339][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:07:11,840][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:07:12,341][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:07:12,835][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:07:13,335][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:07:13,832][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:07:14,330][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:07:14,831][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:07:15,329][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:07:15,831][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:07:16,335][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:07:16,836][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:07:17,339][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:07:17,839][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:07:18,338][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:07:18,840][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:07:19,340][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:07:19,840][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:07:20,339][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:07:20,837][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:07:21,336][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:07:21,834][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:07:22,331][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:07:22,830][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:07:23,327][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:07:23,827][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:07:24,326][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:07:24,822][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:07:25,320][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:07:25,816][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:07:26,314][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:07:26,843][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:07:27,342][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:07:27,840][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:07:28,339][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:07:28,840][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:07:29,343][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:07:29,867][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:07:30,369][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:07:30,869][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:07:31,367][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:07:31,868][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:07:32,378][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:07:32,875][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:07:33,376][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:07:33,873][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:07:34,373][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:07:34,883][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:07:35,382][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:07:35,887][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:07:36,387][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:07:36,883][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:07:37,382][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:07:37,881][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:07:38,379][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10844 tokens. -[2025-11-09 20:07:39,037][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 55.68%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:32 -[2025-11-09 20:07:39,779][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:07:39,780][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:07:39,782][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:07:41,602][__main__][INFO] - Iteration 149 took 51s (28.24% Gen, 68.20% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 27m 34s. Estimated total time: 42h 36m 19s. Time estimates for 10 more iterations: 8m 31s, 100 more iterations: 1h 25m 12s, 500 more iterations: 7h 6m 3s. -[2025-11-09 20:07:41,604][__main__][INFO] - Starting iteration 149. -[2025-11-09 20:07:42,062][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 14 and human policies 1. -[2025-11-09 20:07:42,063][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:07:56,530][__main__][INFO] - Number of regex retries in iteration 149: 0 -[2025-11-09 20:07:56,531][__main__][INFO] - agents played in iteration 149 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:07:57,323][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:07:57,351][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:07:57,378][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:07:57,402][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:07:57,402][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:07:57,403][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:07:58,031][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:07:58,501][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:07:59,007][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:07:59,524][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:08:00,023][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:08:00,521][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:08:01,021][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:08:01,523][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:08:02,031][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:08:02,530][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:08:03,024][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:08:03,536][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:08:04,035][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:08:04,532][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:08:05,032][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:08:05,533][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:08:06,038][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:08:06,540][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:08:07,037][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:08:07,541][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:08:08,041][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:08:08,544][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:08:09,044][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:08:09,543][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:08:10,047][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:08:10,547][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:08:11,047][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:08:11,543][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:08:12,041][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:08:12,544][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:08:13,041][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:08:13,539][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:08:14,049][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:08:14,546][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:08:15,045][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:08:15,546][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:08:16,044][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:08:16,545][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:08:17,041][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:08:17,538][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:08:18,047][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:08:18,543][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:08:19,040][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:08:19,537][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:08:20,034][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:08:20,532][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:08:21,028][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:08:21,524][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:08:22,023][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:08:22,524][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:08:23,024][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:08:23,524][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:08:24,022][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:08:24,520][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:08:25,022][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:08:25,522][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:08:26,026][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:08:26,523][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:08:27,023][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:08:27,521][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:08:28,017][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:08:28,521][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:08:29,018][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:08:29,520][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:08:30,019][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10847 tokens. -[2025-11-09 20:08:30,672][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 -[2025-11-09 20:08:31,446][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:08:31,447][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:08:31,449][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:08:32,462][__main__][INFO] - Iteration 150 took 50s (28.71% Gen, 69.28% Train). Generation: 14s, Training: 34s. Estimated remaining time: 39h 50m 23s. Estimated total time: 41h 59m 59s. Time estimates for 10 more iterations: 8m 23s, 100 more iterations: 1h 23m 59s, 500 more iterations: 6h 59m 59s. -[2025-11-09 20:08:32,464][__main__][INFO] - Starting iteration 150. -[2025-11-09 20:08:32,937][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 14 and human policies 1. -[2025-11-09 20:08:32,938][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:08:38,168][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:08:47,093][__main__][INFO] - Number of regex retries in iteration 150: 1 -[2025-11-09 20:08:47,094][__main__][INFO] - agents played in iteration 150 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:08:47,872][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:08:47,896][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:08:47,921][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:08:47,944][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:08:47,944][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:08:47,945][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:08:48,567][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:08:49,020][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:08:49,541][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:08:50,041][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:08:50,540][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:08:51,037][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:08:51,534][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:08:52,043][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:08:52,542][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:08:53,043][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:08:53,541][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:08:54,037][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:08:54,532][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:08:55,032][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:08:55,529][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:08:56,048][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:08:56,547][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:08:57,050][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:08:57,550][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:08:58,051][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:08:58,551][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:08:59,052][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:08:59,551][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:09:00,065][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:09:00,565][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:09:01,064][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:09:01,564][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:09:02,063][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:09:02,568][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:09:03,069][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:09:03,567][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:09:04,072][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:09:04,570][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:09:05,069][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:09:05,562][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:09:06,057][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:09:06,551][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:09:07,045][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:09:07,538][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:09:08,034][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:09:08,530][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:09:09,031][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:09:09,541][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:09:10,042][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:09:10,553][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:09:11,054][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:09:11,552][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:09:12,053][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:09:12,551][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:09:13,069][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:09:13,571][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:09:14,073][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:09:14,586][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:09:15,085][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:09:15,588][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:09:16,089][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:09:16,589][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:09:17,090][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:09:17,593][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:09:18,092][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:09:18,591][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:09:19,090][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:09:19,587][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:09:20,085][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:09:20,581][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10874 tokens. -[2025-11-09 20:09:21,231][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 55.61%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:32 -[2025-11-09 20:09:21,979][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:09:21,980][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:09:21,982][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:09:23,988][__main__][INFO] - Iteration 151 took 51s (27.73% Gen, 68.34% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 22m 8s. Estimated total time: 42h 32m 35s. Time estimates for 10 more iterations: 8m 30s, 100 more iterations: 1h 25m 5s, 500 more iterations: 7h 5m 25s. -[2025-11-09 20:09:23,990][__main__][INFO] - Starting iteration 151. -[2025-11-09 20:09:24,484][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 15 and human policies 1. -[2025-11-09 20:09:24,484][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:09:39,082][__main__][INFO] - Number of regex retries in iteration 151: 0 -[2025-11-09 20:09:39,082][__main__][INFO] - agents played in iteration 151 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:09:39,867][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:09:39,894][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:09:39,921][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:09:39,944][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:09:39,944][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:09:39,945][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:09:40,580][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:09:41,034][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:09:41,538][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:09:42,039][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:09:42,538][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:09:43,038][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:09:43,549][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:09:44,046][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:09:44,544][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:09:45,041][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:09:45,535][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:09:46,034][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:09:46,532][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:09:47,031][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:09:47,530][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:09:48,029][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:09:48,531][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:09:49,040][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:09:49,539][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:09:50,053][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:09:50,553][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:09:51,051][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:09:51,556][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:09:52,054][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:09:52,564][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:09:53,063][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:09:53,564][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:09:54,064][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:09:54,563][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:09:55,064][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:09:55,563][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:09:56,061][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:09:56,581][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:09:57,077][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:09:57,578][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:09:58,075][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:09:58,575][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:09:59,073][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:09:59,572][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:10:00,069][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:10:00,567][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:10:01,066][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:10:01,564][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:10:02,063][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:10:02,561][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:10:03,083][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:10:03,584][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:10:04,086][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:10:04,586][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:10:05,083][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:10:05,582][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:10:06,083][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:10:06,584][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:10:07,092][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:10:07,593][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:10:08,094][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:10:08,588][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:10:09,088][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:10:09,585][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:10:10,085][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:10:10,584][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:10:11,086][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:10:11,586][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:10:12,086][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:10:12,584][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10871 tokens. -[2025-11-09 20:10:13,241][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 -[2025-11-09 20:10:14,001][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:10:14,003][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:10:14,005][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:10:16,396][__main__][INFO] - Iteration 152 took 51s (28.12% Gen, 67.27% Train). Generation: 14s, Training: 34s. Estimated remaining time: 41h 4m 18s. Estimated total time: 43h 15m 37s. Time estimates for 10 more iterations: 8m 39s, 100 more iterations: 1h 26m 31s, 500 more iterations: 7h 12m 36s. -[2025-11-09 20:10:16,398][__main__][INFO] - Starting iteration 152. -[2025-11-09 20:10:16,885][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 15 and human policies 1. -[2025-11-09 20:10:16,885][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:10:22,730][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:10:32,648][__main__][INFO] - Number of regex retries in iteration 152: 1 -[2025-11-09 20:10:32,649][__main__][INFO] - agents played in iteration 152 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:10:33,429][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:10:33,454][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:10:33,479][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:10:33,501][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:10:33,502][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:10:33,503][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:10:34,161][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:10:34,628][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:10:35,132][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:10:35,633][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:10:36,133][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:10:36,636][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:10:37,136][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:10:37,635][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:10:38,137][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:10:38,661][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:10:39,160][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:10:39,671][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:10:40,171][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:10:40,672][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:10:41,179][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:10:41,683][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:10:42,188][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:10:42,705][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:10:43,206][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:10:43,706][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:10:44,209][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:10:44,710][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:10:45,210][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:10:45,709][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:10:46,206][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:10:46,707][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:10:47,206][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:10:47,707][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:10:48,206][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:10:48,706][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:10:49,205][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:10:49,703][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:10:50,203][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:10:50,702][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:10:51,201][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:10:51,702][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:10:52,201][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:10:52,703][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:10:53,206][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:10:53,705][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:10:54,204][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:10:54,699][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:10:55,196][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:10:55,698][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:10:56,194][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:10:56,694][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:10:57,187][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:10:57,685][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:10:58,187][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:10:58,688][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:10:59,190][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:10:59,691][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:11:00,190][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:11:00,690][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:11:01,191][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:11:01,695][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:11:02,199][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:11:02,699][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:11:03,199][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:11:03,699][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:11:04,199][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:11:04,705][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:11:05,209][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:11:05,708][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:11:06,210][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10859 tokens. -[2025-11-09 20:11:06,872][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 55.66%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 -[2025-11-09 20:11:07,649][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:11:07,651][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:11:07,653][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:11:09,189][__main__][INFO] - Iteration 153 took 52s (30.14% Gen, 66.92% Train). Generation: 15s, Training: 35s. Estimated remaining time: 41h 23m 2s. Estimated total time: 43h 35m 15s. Time estimates for 10 more iterations: 8m 43s, 100 more iterations: 1h 27m 10s, 500 more iterations: 7h 15m 52s. -[2025-11-09 20:11:09,192][__main__][INFO] - Starting iteration 153. -[2025-11-09 20:11:09,650][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 15 and human policies 1. -[2025-11-09 20:11:09,651][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:11:12,981][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:11:23,934][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls +Proposal: 1 hat, 1 book, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:38:16,315][__main__][INFO] - Number of regex retries in iteration 106: 1 +[2026-03-25 22:38:16,316][__main__][INFO] - agents played in iteration 106 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:38:17,091][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:38:17,114][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:38:17,137][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:38:17,161][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:38:17,162][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:38:17,162][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:38:17,767][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:38:18,219][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:38:18,723][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:38:19,224][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:38:19,727][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:38:20,223][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:38:20,725][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:38:21,246][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:38:21,746][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:38:22,242][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:38:22,741][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:38:23,243][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:38:23,743][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:38:24,241][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:38:24,741][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:38:25,242][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:38:25,743][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:38:26,242][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:38:26,742][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:38:27,243][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:38:27,741][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:38:28,239][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:38:28,742][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:38:29,241][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:38:29,740][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:38:30,239][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:38:30,741][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:38:31,241][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:38:31,742][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:38:32,241][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:38:32,740][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:38:33,241][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:38:33,742][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:38:34,243][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:38:34,746][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:38:35,248][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:38:35,745][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:38:36,246][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:38:36,748][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:38:37,249][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:38:37,748][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:38:38,248][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:38:38,749][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:38:39,248][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:38:39,746][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:38:40,245][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:38:40,745][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:38:41,246][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:38:41,745][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:38:42,246][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:38:42,748][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:38:43,246][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:38:43,746][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:38:44,248][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:38:44,749][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:38:45,251][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:38:45,751][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:38:46,247][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:38:46,749][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:38:47,246][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:38:47,743][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:38:48,242][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:38:48,742][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:38:49,240][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:38:49,739][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10504 tokens. +[2026-03-25 22:38:50,382][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 60.57%, Block Peak % of device VRAM: 62.14%, ΔTime: 00:00:32 +[2026-03-25 22:38:51,097][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:38:51,099][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:38:51,101][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:38:51,796][__main__][INFO] - Iteration 107 took 51s (31.54% Gen, 67.11% Train). Generation: 16s, Training: 34s. Estimated remaining time: 41h 38m 20s. Estimated total time: 43h 11m 32s. Time estimates for 10 more iterations: 8m 38s, 100 more iterations: 1h 26m 23s, 500 more iterations: 7h 11m 55s. +[2026-03-25 22:38:51,798][__main__][INFO] - Starting iteration 107. +[2026-03-25 22:38:52,198][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 10 and human policies 1. +[2026-03-25 22:38:52,198][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:38:58,362][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:39:07,844][__main__][INFO] - Number of regex retries in iteration 107: 1 +[2026-03-25 22:39:07,845][__main__][INFO] - agents played in iteration 107 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:39:08,619][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:39:08,642][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:39:08,665][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:39:08,688][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:39:08,689][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:39:08,689][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:39:09,284][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:39:09,736][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:39:10,243][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:39:10,745][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:39:11,245][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:39:11,744][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:39:12,243][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:39:12,743][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:39:13,242][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:39:13,740][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:39:14,240][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:39:14,740][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:39:15,245][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:39:15,745][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:39:16,243][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:39:16,744][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:39:17,243][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:39:17,744][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:39:18,243][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:39:18,744][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:39:19,245][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:39:19,744][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:39:20,245][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:39:20,749][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:39:21,250][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:39:21,745][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:39:22,244][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:39:22,743][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:39:23,243][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:39:23,740][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:39:24,239][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:39:24,737][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:39:25,236][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:39:25,734][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:39:26,234][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:39:26,732][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:39:27,229][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:39:27,727][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:39:28,226][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:39:28,726][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:39:29,225][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:39:29,720][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:39:30,218][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:39:30,714][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:39:31,271][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:39:31,775][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:39:32,273][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:39:32,773][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:39:33,275][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:39:33,773][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:39:34,274][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:39:34,775][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:39:35,273][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:39:35,773][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:39:36,274][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:39:36,773][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:39:37,272][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:39:37,771][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:39:38,299][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:39:38,800][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:39:39,299][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:39:39,800][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:39:40,302][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:39:40,804][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:39:41,307][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10680 tokens. +[2026-03-25 22:39:41,953][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 60.67%, Block Peak % of device VRAM: 62.35%, ΔTime: 00:00:32 +[2026-03-25 22:39:42,682][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:39:42,684][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:39:42,686][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:39:43,386][__main__][INFO] - Iteration 108 took 51s (30.57% Gen, 68.06% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 5m 24s. Estimated total time: 42h 39m 27s. Time estimates for 10 more iterations: 8m 31s, 100 more iterations: 1h 25m 18s, 500 more iterations: 7h 6m 34s. +[2026-03-25 22:39:43,390][__main__][INFO] - Starting iteration 108. +[2026-03-25 22:39:43,790][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 10 and human policies 1. +[2026-03-25 22:39:43,791][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:39:46,575][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:39:47,618][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:39:50,617][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:39:56,027][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:39:57,752][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 1, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:39:58,009][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 1, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:39:58,676][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 1, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 2/3 +[2026-03-25 22:39:59,681][__main__][INFO] - Number of regex retries in iteration 108: 7 +[2026-03-25 22:39:59,682][__main__][INFO] - agents played in iteration 108 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:40:00,479][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:40:00,502][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:40:00,525][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:40:00,548][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:40:00,549][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:40:00,549][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:40:01,172][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:40:01,627][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:40:02,130][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:40:02,631][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:40:03,129][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:40:03,632][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:40:04,131][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:40:04,627][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:40:05,126][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:40:05,626][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:40:06,128][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:40:06,627][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:40:07,127][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:40:07,629][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:40:08,126][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:40:08,626][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:40:09,128][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:40:09,630][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:40:10,130][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:40:10,628][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:40:11,128][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:40:11,627][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:40:12,127][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:40:12,626][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:40:13,127][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:40:13,648][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:40:14,148][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:40:14,649][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:40:15,150][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:40:15,654][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:40:16,153][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:40:16,653][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:40:17,154][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:40:17,657][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:40:18,154][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:40:18,654][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:40:19,154][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:40:19,651][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:40:20,147][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:40:20,647][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:40:21,148][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:40:21,650][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:40:22,148][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:40:22,647][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:40:23,149][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:40:23,648][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:40:24,147][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:40:24,649][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:40:25,147][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:40:25,648][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:40:26,147][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:40:26,647][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:40:27,147][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:40:27,644][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:40:28,144][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:40:28,645][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:40:29,145][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:40:29,646][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:40:30,145][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:40:30,646][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:40:31,144][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:40:31,645][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:40:32,144][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:40:32,645][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:40:33,146][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10598 tokens. +[2026-03-25 22:40:33,771][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 60.64%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 +[2026-03-25 22:40:34,501][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:40:34,503][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:40:34,505][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:40:35,309][__main__][INFO] - Iteration 109 took 51s (30.85% Gen, 67.59% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 21m 3s. Estimated total time: 42h 55m 58s. Time estimates for 10 more iterations: 8m 35s, 100 more iterations: 1h 25m 51s, 500 more iterations: 7h 9m 19s. +[2026-03-25 22:40:35,311][__main__][INFO] - Starting iteration 109. +[2026-03-25 22:40:35,710][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 10 and human policies 1. +[2026-03-25 22:40:35,710][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:40:38,276][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:40:39,576][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:40:51,365][__main__][INFO] - Number of regex retries in iteration 109: 2 +[2026-03-25 22:40:51,366][__main__][INFO] - agents played in iteration 109 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:40:52,149][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:40:52,172][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:40:52,195][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:40:52,218][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:40:52,219][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:40:52,219][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:40:52,822][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:40:53,274][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:40:53,776][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:40:54,274][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:40:54,774][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:40:55,271][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:40:55,772][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:40:56,271][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:40:56,767][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:40:57,268][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:40:57,772][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:40:58,273][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:40:58,773][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:40:59,272][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:40:59,772][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:41:00,270][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:41:00,772][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:41:01,269][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:41:01,767][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:41:02,268][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:41:02,766][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:41:03,267][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:41:03,766][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:41:04,263][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:41:04,763][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:41:05,265][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:41:05,763][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:41:06,261][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:41:06,761][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:41:07,261][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:41:07,759][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:41:08,255][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:41:08,753][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:41:09,256][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:41:09,756][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:41:10,253][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:41:10,752][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:41:11,252][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:41:11,750][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:41:12,250][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:41:12,751][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:41:13,249][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:41:13,749][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:41:14,245][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:41:14,746][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:41:15,249][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:41:15,745][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:41:16,244][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:41:16,745][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:41:17,243][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:41:17,739][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:41:18,238][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:41:18,735][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:41:19,232][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:41:19,728][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:41:20,227][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:41:20,722][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:41:21,222][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:41:21,717][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:41:22,217][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:41:22,716][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:41:23,214][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:41:23,709][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:41:24,208][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:41:24,708][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10550 tokens. +[2026-03-25 22:41:25,344][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.58%, Block Peak % of device VRAM: 62.38%, ΔTime: 00:00:32 +[2026-03-25 22:41:26,066][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:41:26,068][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:41:26,070][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:41:26,706][__main__][INFO] - Iteration 110 took 50s (30.70% Gen, 68.05% Train). Generation: 15s, Training: 34s. Estimated remaining time: 40h 54m 2s. Estimated total time: 42h 29m 49s. Time estimates for 10 more iterations: 8m 29s, 100 more iterations: 1h 24m 59s, 500 more iterations: 7h 4m 58s. +[2026-03-25 22:41:26,708][__main__][INFO] - Starting iteration 110. +[2026-03-25 22:41:27,111][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 10 and human policies 1. +[2026-03-25 22:41:27,112][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:41:39,511][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:41:42,261][__main__][INFO] - Number of regex retries in iteration 110: 1 +[2026-03-25 22:41:42,262][__main__][INFO] - agents played in iteration 110 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:41:43,066][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:41:43,090][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:41:43,112][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:41:43,135][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:41:43,136][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:41:43,136][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:41:43,744][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:41:44,200][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:41:44,703][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:41:45,202][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:41:45,702][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:41:46,204][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:41:46,703][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:41:47,204][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:41:47,704][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:41:48,204][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:41:48,705][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:41:49,202][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:41:49,701][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:41:50,201][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:41:50,698][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:41:51,196][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:41:51,695][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:41:52,193][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:41:52,694][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:41:53,193][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:41:53,692][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:41:54,191][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:41:54,689][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:41:55,189][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:41:55,687][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:41:56,187][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:41:56,687][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:41:57,184][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:41:57,680][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:41:58,179][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:41:58,677][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:41:59,174][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:41:59,669][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:42:00,171][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:42:00,670][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:42:01,170][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:42:01,670][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:42:02,171][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:42:02,669][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:42:03,169][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:42:03,668][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:42:04,167][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:42:04,665][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:42:05,166][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:42:05,665][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:42:06,164][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:42:06,665][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:42:07,164][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:42:07,662][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:42:08,161][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:42:08,660][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:42:09,157][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:42:09,654][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:42:10,152][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:42:10,650][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:42:11,151][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:42:11,650][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:42:12,150][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:42:12,649][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:42:13,148][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:42:13,649][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:42:14,151][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:42:14,651][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:42:15,150][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:42:15,649][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10602 tokens. +[2026-03-25 22:42:16,288][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 60.56%, Block Peak % of device VRAM: 62.09%, ΔTime: 00:00:32 +[2026-03-25 22:42:17,017][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:42:17,019][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:42:17,021][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:42:18,283][__main__][INFO] - Iteration 111 took 51s (29.61% Gen, 67.93% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 1m 58s. Estimated total time: 42h 38m 36s. Time estimates for 10 more iterations: 8m 31s, 100 more iterations: 1h 25m 17s, 500 more iterations: 7h 6m 26s. +[2026-03-25 22:42:18,285][__main__][INFO] - Starting iteration 111. +[2026-03-25 22:42:18,687][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 11 and human policies 1. +[2026-03-25 22:42:18,687][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:42:21,344][mllm.models.large_language_model_local][WARNING] - Response Proposal: 5 hats, 5 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:42:34,424][__main__][INFO] - Number of regex retries in iteration 111: 1 +[2026-03-25 22:42:34,425][__main__][INFO] - agents played in iteration 111 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:42:35,214][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:42:35,237][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:42:35,260][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:42:35,282][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:42:35,283][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:42:35,284][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:42:35,887][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:42:36,345][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:42:36,851][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:42:37,351][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:42:37,851][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:42:38,352][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:42:38,848][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:42:39,349][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:42:39,850][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:42:40,350][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:42:40,848][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:42:41,349][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:42:41,850][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:42:42,352][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:42:42,851][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:42:43,351][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:42:43,850][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:42:44,347][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:42:44,846][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:42:45,348][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:42:45,849][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:42:46,348][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:42:46,847][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:42:47,349][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:42:47,850][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:42:48,346][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:42:48,843][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:42:49,345][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:42:49,843][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:42:50,345][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:42:50,845][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:42:51,342][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:42:51,843][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:42:52,342][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:42:52,843][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:42:53,341][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:42:53,837][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:42:54,338][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:42:54,835][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:42:55,334][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:42:55,831][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:42:56,333][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:42:56,835][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:42:57,333][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:42:57,834][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:42:58,334][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:42:58,833][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:42:59,329][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:42:59,828][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:43:00,330][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:43:00,831][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:43:01,330][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:43:01,834][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:43:02,335][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:43:02,834][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:43:03,337][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:43:03,839][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:43:04,336][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:43:04,833][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:43:05,330][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:43:05,829][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:43:06,327][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:43:06,825][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:43:07,322][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:43:07,822][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10647 tokens. +[2026-03-25 22:43:08,476][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.58%, Block Peak % of device VRAM: 62.36%, ΔTime: 00:00:32 +[2026-03-25 22:43:09,202][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:43:09,204][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:43:09,206][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:43:09,904][__main__][INFO] - Iteration 112 took 51s (30.73% Gen, 67.91% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 3m 24s. Estimated total time: 42h 40m 54s. Time estimates for 10 more iterations: 8m 32s, 100 more iterations: 1h 25m 21s, 500 more iterations: 7h 6m 49s. +[2026-03-25 22:43:09,906][__main__][INFO] - Starting iteration 112. +[2026-03-25 22:43:10,308][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 11 and human policies 1. +[2026-03-25 22:43:10,309][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:43:12,950][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:43:12,952][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:43:25,578][__main__][INFO] - Number of regex retries in iteration 112: 2 +[2026-03-25 22:43:25,578][__main__][INFO] - agents played in iteration 112 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:43:26,362][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:43:26,385][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:43:26,409][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:43:26,432][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:43:26,433][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:43:26,434][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:43:27,045][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:43:27,496][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:43:28,004][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:43:28,501][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:43:28,998][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:43:29,498][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:43:29,995][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:43:30,500][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:43:31,002][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:43:31,500][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:43:32,003][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:43:32,503][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:43:33,002][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:43:33,501][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:43:34,003][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:43:34,504][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:43:35,004][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:43:35,500][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:43:36,002][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:43:36,503][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:43:37,003][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:43:37,500][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:43:38,002][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:43:38,502][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:43:39,001][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:43:39,500][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:43:39,999][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:43:40,501][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:43:41,001][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:43:41,500][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:43:41,998][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:43:42,496][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:43:42,992][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:43:43,492][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:43:43,997][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:43:44,501][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:43:45,002][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:43:45,504][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:43:46,006][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:43:46,505][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:43:47,003][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:43:47,504][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:43:48,007][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:43:48,507][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:43:49,006][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:43:49,505][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:43:50,004][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:43:50,505][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:43:51,007][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:43:51,506][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:43:52,005][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:43:52,506][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:43:53,029][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:43:53,531][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:43:54,032][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:43:54,535][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:43:55,036][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:43:55,536][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:43:56,035][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:43:56,538][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:43:57,037][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:43:57,535][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:43:58,031][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:43:58,530][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:43:59,031][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10601 tokens. +[2026-03-25 22:43:59,665][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 60.63%, Block Peak % of device VRAM: 62.22%, ΔTime: 00:00:32 +[2026-03-25 22:44:00,395][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:44:00,397][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:44:00,399][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:44:01,104][__main__][INFO] - Iteration 113 took 50s (30.06% Gen, 68.55% Train). Generation: 15s, Training: 34s. Estimated remaining time: 40h 41m 28s. Estimated total time: 42h 19m 49s. Time estimates for 10 more iterations: 8m 27s, 100 more iterations: 1h 24m 39s, 500 more iterations: 7h 3m 18s. +[2026-03-25 22:44:01,107][__main__][INFO] - Starting iteration 113. +[2026-03-25 22:44:01,506][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 11 and human policies 1. +[2026-03-25 22:44:01,506][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:44:12,712][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:44:16,844][__main__][INFO] - Number of regex retries in iteration 113: 1 +[2026-03-25 22:44:16,845][__main__][INFO] - agents played in iteration 113 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:44:17,622][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:44:17,645][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:44:17,668][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:44:17,690][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:44:17,691][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:44:17,691][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:44:18,307][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:44:18,759][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:44:19,263][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:44:19,761][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:44:20,261][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:44:20,764][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:44:21,262][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:44:21,761][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:44:22,261][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:44:22,767][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:44:23,264][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:44:23,762][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:44:24,264][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:44:24,767][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:44:25,264][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:44:25,761][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:44:26,261][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:44:26,763][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:44:27,261][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:44:27,759][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:44:28,259][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:44:28,759][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:44:29,258][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:44:29,755][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:44:30,254][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:44:30,756][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:44:31,273][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:44:31,773][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:44:32,275][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:44:32,778][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:44:33,279][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:44:33,781][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:44:34,284][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:44:34,781][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:44:35,277][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:44:35,775][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:44:36,275][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:44:36,775][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:44:37,272][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:44:37,771][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:44:38,272][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:44:38,776][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:44:39,276][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:44:39,775][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:44:40,278][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:44:40,780][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:44:41,279][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:44:41,782][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:44:42,281][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:44:42,778][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:44:43,274][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:44:43,773][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:44:44,274][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:44:44,776][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:44:45,273][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:44:45,774][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:44:46,273][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:44:46,774][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:44:47,272][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:44:47,774][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:44:48,273][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:44:48,775][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:44:49,275][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:44:49,778][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:44:50,279][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10627 tokens. +[2026-03-25 22:44:50,901][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 60.61%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 +[2026-03-25 22:44:51,622][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:44:51,624][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:44:51,626][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:44:52,281][__main__][INFO] - Iteration 114 took 50s (30.21% Gen, 68.50% Train). Generation: 15s, Training: 34s. Estimated remaining time: 40h 39m 34s. Estimated total time: 42h 18m 46s. Time estimates for 10 more iterations: 8m 27s, 100 more iterations: 1h 24m 37s, 500 more iterations: 7h 3m 7s. +[2026-03-25 22:44:52,283][__main__][INFO] - Starting iteration 114. +[2026-03-25 22:44:52,683][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 11 and human policies 1. +[2026-03-25 22:44:52,684][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:45:08,860][__main__][INFO] - Number of regex retries in iteration 114: 0 +[2026-03-25 22:45:08,861][__main__][INFO] - agents played in iteration 114 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:45:09,645][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:45:09,669][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:45:09,692][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:45:09,714][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:45:09,715][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:45:09,715][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:45:10,328][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:45:10,781][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:45:11,283][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:45:11,784][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:45:12,286][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:45:12,786][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:45:13,284][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:45:13,782][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:45:14,285][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:45:14,791][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:45:15,290][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:45:15,792][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:45:16,293][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:45:16,793][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:45:17,291][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:45:17,792][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:45:18,291][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:45:18,789][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:45:19,284][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:45:19,784][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:45:20,283][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:45:20,780][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:45:21,274][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:45:21,773][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:45:22,272][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:45:22,773][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:45:23,274][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:45:23,776][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:45:24,278][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:45:24,777][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:45:25,275][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:45:25,777][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:45:26,281][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:45:26,782][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:45:27,282][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:45:27,780][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:45:28,283][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:45:28,783][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:45:29,280][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:45:29,784][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:45:30,284][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:45:30,780][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:45:31,276][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:45:31,777][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:45:32,278][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:45:32,776][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:45:33,276][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:45:33,779][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:45:34,279][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:45:34,779][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:45:35,276][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:45:35,776][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:45:36,278][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:45:36,771][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:45:37,270][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:45:37,772][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:45:38,274][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:45:38,774][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:45:39,299][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:45:39,801][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:45:40,301][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:45:40,799][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:45:41,300][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:45:41,802][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:45:42,303][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10571 tokens. +[2026-03-25 22:45:42,954][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 60.62%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:32 +[2026-03-25 22:45:43,686][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:45:43,688][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:45:43,690][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:45:44,389][__main__][INFO] - Iteration 115 took 51s (31.29% Gen, 67.36% Train). Generation: 16s, Training: 34s. Estimated remaining time: 41h 25m 14s. Estimated total time: 43h 5m 18s. Time estimates for 10 more iterations: 8m 37s, 100 more iterations: 1h 26m 10s, 500 more iterations: 7h 10m 53s. +[2026-03-25 22:45:44,391][__main__][INFO] - Starting iteration 115. +[2026-03-25 22:45:44,789][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 11 and human policies 1. +[2026-03-25 22:45:44,790][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:46:00,981][__main__][INFO] - Number of regex retries in iteration 115: 0 +[2026-03-25 22:46:00,982][__main__][INFO] - agents played in iteration 115 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:46:01,771][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:46:01,794][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:46:01,818][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:46:01,841][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:46:01,842][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:46:01,842][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:46:02,457][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:46:02,913][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:46:03,415][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:46:03,915][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:46:04,421][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:46:04,921][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:46:05,419][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:46:05,920][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:46:06,420][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:46:06,925][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:46:07,425][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:46:07,926][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:46:08,427][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:46:08,930][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:46:09,430][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:46:09,932][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:46:10,447][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:46:10,953][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:46:11,452][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:46:11,954][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:46:12,453][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:46:12,951][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:46:13,447][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:46:13,950][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:46:14,452][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:46:14,950][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:46:15,450][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:46:15,951][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:46:16,451][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:46:16,948][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:46:17,449][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:46:17,947][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:46:18,446][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:46:18,946][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:46:19,445][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:46:19,944][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:46:20,445][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:46:20,942][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:46:21,442][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:46:21,943][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:46:22,445][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:46:22,943][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:46:23,442][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:46:23,942][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:46:24,441][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:46:24,940][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:46:25,444][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:46:25,944][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:46:26,446][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:46:26,945][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:46:27,447][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:46:27,947][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:46:28,449][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:46:28,948][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:46:29,448][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:46:29,947][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:46:30,446][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:46:30,946][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:46:31,448][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:46:31,950][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:46:32,451][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:46:32,951][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:46:33,451][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:46:33,951][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:46:34,449][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10687 tokens. +[2026-03-25 22:46:35,092][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 60.62%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:32 +[2026-03-25 22:46:35,814][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:46:35,816][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:46:35,818][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:46:36,514][__main__][INFO] - Iteration 116 took 51s (31.30% Gen, 67.35% Train). Generation: 16s, Training: 34s. Estimated remaining time: 41h 25m 22s. Estimated total time: 43h 6m 18s. Time estimates for 10 more iterations: 8m 37s, 100 more iterations: 1h 26m 12s, 500 more iterations: 7h 11m 3s. +[2026-03-25 22:46:36,517][__main__][INFO] - Starting iteration 116. +[2026-03-25 22:46:36,915][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 11 and human policies 1. +[2026-03-25 22:46:36,915][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:46:39,852][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:46:52,269][__main__][INFO] - Number of regex retries in iteration 116: 1 +[2026-03-25 22:46:52,270][__main__][INFO] - agents played in iteration 116 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:46:53,053][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:46:53,077][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:46:53,100][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:46:53,122][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:46:53,123][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:46:53,124][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:46:53,808][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:46:54,270][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:46:54,772][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:46:55,275][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:46:55,776][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:46:56,276][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:46:56,775][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:46:57,276][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:46:57,773][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:46:58,273][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:46:58,771][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:46:59,275][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:46:59,774][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:47:00,272][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:47:00,770][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:47:01,270][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:47:01,771][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:47:02,271][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:47:02,771][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:47:03,299][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:47:03,804][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:47:04,306][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:47:04,805][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:47:05,306][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:47:05,807][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:47:06,307][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:47:06,806][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:47:07,304][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:47:07,805][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:47:08,304][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:47:08,803][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:47:09,305][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:47:09,805][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:47:10,303][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:47:10,803][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:47:11,304][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:47:11,805][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:47:12,308][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:47:12,808][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:47:13,309][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:47:13,809][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:47:14,307][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:47:14,806][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:47:15,306][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:47:15,806][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:47:16,303][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:47:16,803][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:47:17,304][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:47:17,804][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:47:18,303][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:47:18,803][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:47:19,304][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:47:19,805][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:47:20,303][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:47:20,803][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:47:21,301][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:47:21,802][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:47:22,298][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:47:22,796][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:47:23,294][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:47:23,793][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:47:24,289][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:47:24,785][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:47:25,283][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:47:25,782][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10693 tokens. +[2026-03-25 22:47:26,429][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 60.56%, Block Peak % of device VRAM: 62.20%, ΔTime: 00:00:32 +[2026-03-25 22:47:27,136][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:47:27,138][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:47:27,140][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:47:27,846][__main__][INFO] - Iteration 117 took 50s (30.15% Gen, 68.46% Train). Generation: 15s, Training: 34s. Estimated remaining time: 40h 44m 47s. Estimated total time: 42h 26m 35s. Time estimates for 10 more iterations: 8m 29s, 100 more iterations: 1h 24m 53s, 500 more iterations: 7h 4m 25s. +[2026-03-25 22:47:27,848][__main__][INFO] - Starting iteration 117. +[2026-03-25 22:47:28,247][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 11 and human policies 1. +[2026-03-25 22:47:28,247][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:47:43,336][__main__][INFO] - Number of regex retries in iteration 117: 0 +[2026-03-25 22:47:43,337][__main__][INFO] - agents played in iteration 117 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:47:44,117][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:47:44,140][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:47:44,163][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:47:44,186][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:47:44,186][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:47:44,187][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:47:44,816][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:47:45,270][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:47:45,772][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:47:46,268][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:47:46,770][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:47:47,268][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:47:47,767][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:47:48,264][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:47:48,765][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:47:49,267][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:47:49,769][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:47:50,271][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:47:50,773][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:47:51,271][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:47:51,772][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:47:52,270][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:47:52,769][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:47:53,268][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:47:53,785][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:47:54,287][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:47:54,783][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:47:55,284][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:47:55,787][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:47:56,286][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:47:56,787][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:47:57,285][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:47:57,784][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:47:58,285][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:47:58,783][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:47:59,283][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:47:59,782][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:48:00,281][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:48:00,780][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:48:01,278][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:48:01,777][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:48:02,278][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:48:02,778][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:48:03,278][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:48:03,780][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:48:04,281][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:48:04,783][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:48:05,283][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:48:05,783][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:48:06,279][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:48:06,782][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:48:07,281][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:48:07,782][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:48:08,282][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:48:08,782][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:48:09,278][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:48:09,777][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:48:10,278][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:48:10,781][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:48:11,276][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:48:11,775][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:48:12,272][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:48:12,770][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:48:13,271][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:48:13,768][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:48:14,269][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:48:14,768][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:48:15,263][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:48:15,765][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:48:16,266][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:48:16,767][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10658 tokens. +[2026-03-25 22:48:17,406][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.17%, ΔTime: 00:00:32 +[2026-03-25 22:48:18,119][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:48:18,122][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:48:18,123][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:48:18,827][__main__][INFO] - Iteration 118 took 50s (29.83% Gen, 68.77% Train). Generation: 15s, Training: 34s. Estimated remaining time: 40h 26m 24s. Estimated total time: 42h 9m 3s. Time estimates for 10 more iterations: 8m 25s, 100 more iterations: 1h 24m 18s, 500 more iterations: 7h 1m 30s. +[2026-03-25 22:48:18,830][__main__][INFO] - Starting iteration 118. +[2026-03-25 22:48:19,228][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 11 and human policies 1. +[2026-03-25 22:48:19,229][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:48:34,159][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:48:35,001][__main__][INFO] - Number of regex retries in iteration 118: 1 +[2026-03-25 22:48:35,002][__main__][INFO] - agents played in iteration 118 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:48:35,805][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:48:35,828][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:48:35,851][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:48:35,874][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:48:35,875][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:48:35,875][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:48:36,491][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:48:36,948][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:48:37,455][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:48:37,955][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:48:38,456][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:48:38,956][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:48:39,455][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:48:39,954][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:48:40,458][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:48:40,957][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:48:41,459][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:48:41,961][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:48:42,463][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:48:42,962][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:48:43,462][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:48:43,963][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:48:44,467][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:48:44,967][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:48:45,466][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:48:45,964][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:48:46,469][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:48:46,969][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:48:47,469][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:48:47,967][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:48:48,468][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:48:48,964][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:48:49,464][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:48:49,963][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:48:50,464][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:48:50,962][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:48:51,459][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:48:51,956][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:48:52,457][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:48:52,954][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:48:53,454][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:48:53,953][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:48:54,456][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:48:54,953][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:48:55,456][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:48:55,956][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:48:56,457][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:48:56,955][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:48:57,478][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:48:57,982][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:48:58,483][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:48:58,982][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:48:59,485][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:48:59,986][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:49:00,490][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:49:00,988][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:49:01,487][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:49:01,986][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:49:02,483][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:49:02,982][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:49:03,483][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:49:03,981][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:49:04,477][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:49:04,980][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:49:05,477][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:49:05,979][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:49:06,476][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:49:06,974][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:49:07,476][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:49:07,977][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:49:08,473][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10619 tokens. +[2026-03-25 22:49:09,123][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.58%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:32 +[2026-03-25 22:49:09,842][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:49:09,844][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:49:09,846][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:49:10,550][__main__][INFO] - Iteration 119 took 51s (30.73% Gen, 67.89% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 2m 35s. Estimated total time: 42h 46m 5s. Time estimates for 10 more iterations: 8m 33s, 100 more iterations: 1h 25m 32s, 500 more iterations: 7h 7m 40s. +[2026-03-25 22:49:10,552][__main__][INFO] - Starting iteration 119. +[2026-03-25 22:49:10,952][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 11 and human policies 1. +[2026-03-25 22:49:10,953][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:49:13,768][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:49:26,184][__main__][INFO] - Number of regex retries in iteration 119: 1 +[2026-03-25 22:49:26,184][__main__][INFO] - agents played in iteration 119 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:49:26,969][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:49:26,992][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:49:27,015][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:49:27,037][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:49:27,038][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:49:27,038][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:49:27,658][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:49:28,112][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:49:28,622][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:49:29,126][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:49:29,622][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:49:30,123][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:49:30,624][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:49:31,125][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:49:31,624][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:49:32,124][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:49:32,625][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:49:33,124][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:49:33,630][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:49:34,128][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:49:34,626][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:49:35,126][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:49:35,628][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:49:36,129][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:49:36,631][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:49:37,129][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:49:37,635][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:49:38,136][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:49:38,636][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:49:39,137][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:49:39,637][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:49:40,136][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:49:40,638][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:49:41,140][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:49:41,643][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:49:42,144][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:49:42,642][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:49:43,145][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:49:43,646][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:49:44,144][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:49:44,646][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:49:45,146][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:49:45,646][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:49:46,145][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:49:46,647][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:49:47,147][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:49:47,649][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:49:48,150][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:49:48,648][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:49:49,148][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:49:49,647][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:49:50,147][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:49:50,648][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:49:51,148][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:49:51,644][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:49:52,144][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:49:52,645][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:49:53,142][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:49:53,643][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:49:54,144][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:49:54,644][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:49:55,145][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:49:55,644][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:49:56,145][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:49:56,645][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:49:57,145][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:49:57,645][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:49:58,143][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:49:58,645][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:49:59,145][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:49:59,647][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10709 tokens. +[2026-03-25 22:50:00,299][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 60.57%, Block Peak % of device VRAM: 62.11%, ΔTime: 00:00:32 +[2026-03-25 22:50:01,031][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:50:01,033][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:50:01,035][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:50:01,746][__main__][INFO] - Iteration 120 took 50s (29.99% Gen, 68.61% Train). Generation: 15s, Training: 34s. Estimated remaining time: 40h 35m 21s. Estimated total time: 42h 19m 42s. Time estimates for 10 more iterations: 8m 27s, 100 more iterations: 1h 24m 39s, 500 more iterations: 7h 3m 17s. +[2026-03-25 22:50:01,748][__main__][INFO] - Starting iteration 120. +[2026-03-25 22:50:02,151][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 11 and human policies 1. +[2026-03-25 22:50:02,151][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:50:16,289][__main__][INFO] - Number of regex retries in iteration 120: 0 +[2026-03-25 22:50:16,290][__main__][INFO] - agents played in iteration 120 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:50:17,067][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:50:17,090][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:50:17,113][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:50:17,135][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:50:17,136][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:50:17,136][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:50:17,744][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:50:18,201][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:50:18,706][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:50:19,206][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:50:19,704][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:50:20,207][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:50:20,708][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:50:21,211][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:50:21,707][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:50:22,206][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:50:22,707][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:50:23,207][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:50:23,706][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:50:24,208][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:50:24,707][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:50:25,211][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:50:25,710][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:50:26,207][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:50:26,703][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:50:27,202][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:50:27,697][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:50:28,197][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:50:28,694][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:50:29,190][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:50:29,688][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:50:30,192][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:50:30,692][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:50:31,203][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:50:31,706][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:50:32,209][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:50:32,710][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:50:33,210][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:50:33,710][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:50:34,210][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:50:34,710][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:50:35,210][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:50:35,710][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:50:36,212][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:50:36,711][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:50:37,210][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:50:37,710][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:50:38,211][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:50:38,709][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:50:39,208][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:50:39,708][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:50:40,208][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:50:40,707][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:50:41,207][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:50:41,707][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:50:42,209][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:50:42,707][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:50:43,205][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:50:43,703][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:50:44,201][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:50:44,702][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:50:45,202][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:50:45,697][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:50:46,200][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:50:46,699][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:50:47,198][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:50:47,692][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:50:48,192][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:50:48,692][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:50:49,192][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:50:49,691][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10649 tokens. +[2026-03-25 22:50:50,345][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.25%, Current % of VRAM taken: 60.67%, Block Peak % of device VRAM: 62.46%, ΔTime: 00:00:32 +[2026-03-25 22:50:51,077][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:50:51,079][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:50:51,081][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:50:52,361][__main__][INFO] - Iteration 121 took 50s (28.16% Gen, 69.29% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 5m 19s. Estimated total time: 41h 50m 31s. Time estimates for 10 more iterations: 8m 22s, 100 more iterations: 1h 23m 41s, 500 more iterations: 6h 58m 25s. +[2026-03-25 22:50:52,363][__main__][INFO] - Starting iteration 121. +[2026-03-25 22:50:52,765][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 12 and human policies 1. +[2026-03-25 22:50:52,766][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:51:08,234][__main__][INFO] - Number of regex retries in iteration 121: 0 +[2026-03-25 22:51:08,235][__main__][INFO] - agents played in iteration 121 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:51:09,031][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:51:09,055][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:51:09,078][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:51:09,102][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:51:09,102][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:51:09,103][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:51:09,718][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:51:10,173][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:51:10,676][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:51:11,175][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:51:11,674][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:51:12,176][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:51:12,674][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:51:13,175][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:51:13,674][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:51:14,176][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:51:14,674][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:51:15,177][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:51:15,679][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:51:16,183][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:51:16,682][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:51:17,182][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:51:17,684][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:51:18,186][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:51:18,687][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:51:19,191][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:51:19,693][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:51:20,193][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:51:20,692][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:51:21,195][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:51:21,698][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:51:22,199][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:51:22,698][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:51:23,200][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:51:23,703][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:51:24,203][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:51:24,702][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:51:25,200][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:51:25,702][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:51:26,201][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:51:26,701][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:51:27,204][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:51:27,705][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:51:28,206][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:51:28,702][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:51:29,204][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:51:29,705][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:51:30,204][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:51:30,703][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:51:31,205][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:51:31,708][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:51:32,209][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:51:32,704][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:51:33,206][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:51:33,708][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:51:34,207][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:51:34,706][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:51:35,209][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:51:35,710][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:51:36,208][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:51:36,707][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:51:37,210][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:51:37,711][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:51:38,212][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:51:38,714][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:51:39,217][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:51:39,724][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:51:40,224][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:51:40,723][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:51:41,227][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:51:41,728][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10714 tokens. +[2026-03-25 22:51:42,373][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 60.66%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 +[2026-03-25 22:51:43,098][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:51:43,100][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:51:43,102][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:51:43,842][__main__][INFO] - Iteration 122 took 51s (30.29% Gen, 68.26% Train). Generation: 15s, Training: 34s. Estimated remaining time: 40h 47m 50s. Estimated total time: 42h 33m 54s. Time estimates for 10 more iterations: 8m 30s, 100 more iterations: 1h 25m 7s, 500 more iterations: 7h 5m 39s. +[2026-03-25 22:51:43,844][__main__][INFO] - Starting iteration 122. +[2026-03-25 22:51:44,246][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 12 and human policies 1. +[2026-03-25 22:51:44,247][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:51:47,257][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:51:59,836][__main__][INFO] - Number of regex retries in iteration 122: 1 +[2026-03-25 22:51:59,837][__main__][INFO] - agents played in iteration 122 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:52:00,623][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:52:00,647][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:52:00,671][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:52:00,694][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:52:00,694][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:52:00,695][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:52:01,300][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:52:01,757][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:52:02,264][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:52:02,766][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:52:03,268][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:52:03,769][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:52:04,267][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:52:04,768][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:52:05,270][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:52:05,770][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:52:06,268][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:52:06,768][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:52:07,270][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:52:07,768][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:52:08,267][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:52:08,769][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:52:09,273][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:52:09,774][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:52:10,273][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:52:10,774][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:52:11,277][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:52:11,778][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:52:12,279][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:52:12,778][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:52:13,280][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:52:13,778][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:52:14,282][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:52:14,786][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:52:15,290][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:52:15,791][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:52:16,289][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:52:16,791][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:52:17,294][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:52:17,813][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:52:18,313][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:52:18,817][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:52:19,320][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:52:19,824][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:52:20,325][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:52:20,829][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:52:21,332][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:52:21,834][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:52:22,336][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:52:22,838][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:52:23,342][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:52:23,839][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:52:24,338][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:52:24,840][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:52:25,343][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:52:25,842][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:52:26,344][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:52:26,847][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:52:27,349][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:52:27,849][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:52:28,349][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:52:28,851][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:52:29,351][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:52:29,850][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:52:30,352][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:52:30,854][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:52:31,356][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:52:31,858][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:52:32,361][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:52:32,865][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:52:33,368][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10702 tokens. +[2026-03-25 22:52:34,004][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.26%, Current % of VRAM taken: 60.68%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:32 +[2026-03-25 22:52:34,730][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:52:34,732][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:52:34,733][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:52:35,510][__main__][INFO] - Iteration 123 took 51s (30.41% Gen, 68.07% Train). Generation: 15s, Training: 34s. Estimated remaining time: 40h 56m 17s. Estimated total time: 42h 43m 12s. Time estimates for 10 more iterations: 8m 32s, 100 more iterations: 1h 25m 26s, 500 more iterations: 7h 7m 12s. +[2026-03-25 22:52:35,512][__main__][INFO] - Starting iteration 123. +[2026-03-25 22:52:35,912][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 12 and human policies 1. +[2026-03-25 22:52:35,913][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:52:50,324][__main__][INFO] - Number of regex retries in iteration 123: 0 +[2026-03-25 22:52:50,326][__main__][INFO] - agents played in iteration 123 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:52:51,099][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:52:51,123][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:52:51,146][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:52:51,169][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:52:51,169][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:52:51,170][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:52:51,785][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:52:52,238][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:52:52,749][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:52:53,250][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:52:53,750][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:52:54,254][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:52:54,757][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:52:55,257][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:52:55,758][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:52:56,261][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:52:56,762][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:52:57,261][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:52:57,759][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:52:58,263][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:52:58,765][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:52:59,266][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:52:59,766][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:53:00,269][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:53:00,773][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:53:01,274][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:53:01,775][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:53:02,278][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:53:02,780][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:53:03,283][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:53:03,784][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:53:04,286][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:53:04,788][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:53:05,287][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:53:05,784][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:53:06,288][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:53:06,788][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:53:07,287][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:53:07,788][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:53:08,294][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:53:08,796][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:53:09,297][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:53:09,821][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:53:10,326][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:53:10,828][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:53:11,329][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:53:11,832][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:53:12,334][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:53:12,835][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:53:13,334][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:53:13,836][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:53:14,338][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:53:14,838][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:53:15,339][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:53:15,842][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:53:16,344][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:53:16,843][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:53:17,342][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:53:17,845][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:53:18,347][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:53:18,848][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:53:19,347][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:53:19,850][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:53:20,352][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:53:20,851][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:53:21,350][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:53:21,852][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:53:22,354][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:53:22,855][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:53:23,356][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:53:23,858][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10816 tokens. +[2026-03-25 22:53:24,484][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.27%, Current % of VRAM taken: 60.69%, Block Peak % of device VRAM: 62.48%, ΔTime: 00:00:32 +[2026-03-25 22:53:25,216][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:53:25,218][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:53:25,220][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:53:25,970][__main__][INFO] - Iteration 124 took 50s (28.79% Gen, 69.71% Train). Generation: 14s, Training: 34s. Estimated remaining time: 39h 55m 11s. Estimated total time: 41h 42m 57s. Time estimates for 10 more iterations: 8m 20s, 100 more iterations: 1h 23m 25s, 500 more iterations: 6h 57m 9s. +[2026-03-25 22:53:25,973][__main__][INFO] - Starting iteration 124. +[2026-03-25 22:53:26,371][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 12 and human policies 1. +[2026-03-25 22:53:26,372][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:53:29,403][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:53:36,533][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 1 z ball did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:53:42,295][__main__][INFO] - Number of regex retries in iteration 124: 2 +[2026-03-25 22:53:42,296][__main__][INFO] - agents played in iteration 124 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:53:43,082][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.59%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:53:43,106][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.59%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:53:43,130][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.59%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:53:43,153][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.59%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:53:43,154][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:53:43,154][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:53:43,769][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:53:44,228][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:53:44,731][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:53:45,233][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:53:45,735][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:53:46,237][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:53:46,737][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:53:47,240][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:53:47,743][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:53:48,244][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:53:48,742][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:53:49,245][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:53:49,747][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:53:50,246][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:53:50,746][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:53:51,249][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:53:51,752][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:53:52,253][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:53:52,750][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:53:53,252][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:53:53,753][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:53:54,251][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:53:54,750][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:53:55,252][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:53:55,755][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:53:56,252][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:53:56,751][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:53:57,251][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:53:57,750][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:53:58,249][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:53:58,748][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:53:59,248][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:53:59,748][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:54:00,247][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:54:00,748][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:54:01,250][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:54:01,752][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:54:02,252][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:54:02,751][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:54:03,251][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:54:03,753][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:54:04,254][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:54:04,753][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:54:05,253][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:54:05,755][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:54:06,255][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:54:06,756][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:54:07,257][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:54:07,754][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:54:08,256][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:54:08,758][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:54:09,262][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:54:09,766][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:54:10,262][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:54:10,758][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:54:11,261][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:54:11,765][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:54:12,264][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:54:12,764][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:54:13,265][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:54:13,765][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:54:14,263][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:54:14,762][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:54:15,262][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:54:15,763][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10757 tokens. +[2026-03-25 22:54:16,404][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 60.61%, Block Peak % of device VRAM: 62.34%, ΔTime: 00:00:32 +[2026-03-25 22:54:17,127][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:54:17,129][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:54:17,130][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:54:17,837][__main__][INFO] - Iteration 125 took 51s (30.94% Gen, 67.68% Train). Generation: 15s, Training: 34s. Estimated remaining time: 41h 4m 40s. Estimated total time: 42h 53m 17s. Time estimates for 10 more iterations: 8m 34s, 100 more iterations: 1h 25m 46s, 500 more iterations: 7h 8m 52s. +[2026-03-25 22:54:17,839][__main__][INFO] - Starting iteration 125. +[2026-03-25 22:54:18,239][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 12 and human policies 1. +[2026-03-25 22:54:18,239][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:54:32,912][__main__][INFO] - Number of regex retries in iteration 125: 0 +[2026-03-25 22:54:32,913][__main__][INFO] - agents played in iteration 125 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:54:33,695][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:54:33,719][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:54:33,742][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:54:33,764][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:54:33,765][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:54:33,766][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:54:34,356][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:54:34,811][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:54:35,317][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:54:35,819][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:54:36,319][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:54:36,822][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:54:37,324][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:54:37,826][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:54:38,326][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:54:38,826][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:54:39,328][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:54:39,829][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:54:40,331][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:54:40,831][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:54:41,331][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:54:41,831][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:54:42,330][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:54:42,829][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:54:43,330][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:54:43,830][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:54:44,329][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:54:44,829][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:54:45,330][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:54:45,832][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:54:46,331][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:54:46,830][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:54:47,329][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:54:47,831][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:54:48,330][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:54:48,830][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:54:49,327][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:54:49,827][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:54:50,325][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:54:50,825][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:54:51,326][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:54:51,828][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:54:52,328][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:54:52,828][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:54:53,329][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:54:53,831][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:54:54,332][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:54:54,829][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:54:55,326][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:54:55,824][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:54:56,323][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:54:56,824][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:54:57,322][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:54:57,822][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:54:58,321][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:54:58,828][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:54:59,346][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:54:59,849][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:55:00,353][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:55:00,856][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:55:01,358][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:55:01,860][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:55:02,363][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:55:02,863][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:55:03,364][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:55:03,865][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:55:04,364][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:55:04,864][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:55:05,365][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:55:05,863][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:55:06,364][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10754 tokens. +[2026-03-25 22:55:06,992][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 60.66%, Block Peak % of device VRAM: 62.41%, ΔTime: 00:00:32 +[2026-03-25 22:55:07,717][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:55:07,719][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:55:07,721][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:55:08,433][__main__][INFO] - Iteration 126 took 50s (29.23% Gen, 69.35% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 0m 17s. Estimated total time: 41h 49m 45s. Time estimates for 10 more iterations: 8m 21s, 100 more iterations: 1h 23m 39s, 500 more iterations: 6h 58m 17s. +[2026-03-25 22:55:08,435][__main__][INFO] - Starting iteration 126. +[2026-03-25 22:55:08,834][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 12 and human policies 1. +[2026-03-25 22:55:08,835][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:55:11,846][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:55:24,122][__main__][INFO] - Number of regex retries in iteration 126: 1 +[2026-03-25 22:55:24,122][__main__][INFO] - agents played in iteration 126 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:55:24,904][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:55:24,927][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:55:24,950][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:55:24,972][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:55:24,973][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:55:24,973][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:55:25,578][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:55:26,034][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:55:26,543][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:55:27,044][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:55:27,544][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:55:28,045][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:55:28,548][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:55:29,051][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:55:29,554][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:55:30,054][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:55:30,554][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:55:31,050][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:55:31,547][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:55:32,048][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:55:32,548][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:55:33,049][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:55:33,547][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:55:34,047][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:55:34,546][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:55:35,051][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:55:35,552][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:55:36,052][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:55:36,553][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:55:37,055][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:55:37,554][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:55:38,052][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:55:38,553][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:55:39,050][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:55:39,548][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:55:40,049][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:55:40,550][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:55:41,052][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:55:41,550][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:55:42,051][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:55:42,552][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:55:43,055][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:55:43,555][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:55:44,055][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:55:44,555][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:55:45,054][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:55:45,553][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:55:46,053][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:55:46,553][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:55:47,052][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:55:47,550][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:55:48,051][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:55:48,549][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:55:49,049][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:55:49,547][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:55:50,048][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:55:50,549][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:55:51,050][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:55:51,550][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:55:52,050][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:55:52,550][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:55:53,049][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:55:53,548][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:55:54,051][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:55:54,551][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:55:55,050][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:55:55,549][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:55:56,051][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:55:56,551][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:55:57,050][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:55:57,549][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10746 tokens. +[2026-03-25 22:55:58,195][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 60.62%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:32 +[2026-03-25 22:55:58,928][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:55:58,930][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:55:58,931][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:55:59,773][__main__][INFO] - Iteration 127 took 50s (30.01% Gen, 68.33% Train). Generation: 15s, Training: 34s. Estimated remaining time: 40h 36m 39s. Estimated total time: 42h 26m 58s. Time estimates for 10 more iterations: 8m 29s, 100 more iterations: 1h 24m 53s, 500 more iterations: 7h 4m 29s. +[2026-03-25 22:55:59,775][__main__][INFO] - Starting iteration 127. +[2026-03-25 22:56:00,174][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 12 and human policies 1. +[2026-03-25 22:56:00,175][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:56:03,187][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:56:14,637][__main__][INFO] - Number of regex retries in iteration 127: 1 +[2026-03-25 22:56:14,638][__main__][INFO] - agents played in iteration 127 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:56:15,414][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:56:15,437][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:56:15,460][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:56:15,482][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:56:15,483][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:56:15,484][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:56:16,090][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:56:16,547][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:56:17,049][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:56:17,550][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:56:18,052][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:56:18,555][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:56:19,055][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:56:19,554][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:56:20,053][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:56:20,553][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:56:21,057][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:56:21,557][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:56:22,057][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:56:22,559][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:56:23,059][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:56:23,558][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:56:24,057][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:56:24,556][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:56:25,053][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:56:25,554][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:56:26,052][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:56:26,551][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:56:27,050][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:56:27,549][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:56:28,049][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:56:28,550][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:56:29,052][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:56:29,551][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:56:30,050][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:56:30,551][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:56:31,051][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:56:31,549][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:56:32,047][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:56:32,549][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:56:33,049][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:56:33,549][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:56:34,050][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:56:34,551][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:56:35,053][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:56:35,551][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:56:36,052][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:56:36,553][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:56:37,053][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:56:37,551][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:56:38,051][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:56:38,552][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:56:39,054][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:56:39,550][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:56:40,051][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:56:40,552][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:56:41,052][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:56:41,551][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:56:42,053][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:56:42,553][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:56:43,052][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:56:43,552][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:56:44,053][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:56:44,554][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:56:45,054][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:56:45,576][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:56:46,077][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:56:46,575][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:56:47,074][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:56:47,574][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:56:48,075][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10790 tokens. +[2026-03-25 22:56:48,709][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 60.66%, Block Peak % of device VRAM: 62.23%, ΔTime: 00:00:32 +[2026-03-25 22:56:49,403][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:56:49,405][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:56:49,407][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:56:50,099][__main__][INFO] - Iteration 128 took 49s (28.97% Gen, 69.64% Train). Generation: 14s, Training: 34s. Estimated remaining time: 39h 45m 5s. Estimated total time: 41h 36m 15s. Time estimates for 10 more iterations: 8m 19s, 100 more iterations: 1h 23m 12s, 500 more iterations: 6h 56m 2s. +[2026-03-25 22:56:50,101][__main__][INFO] - Starting iteration 128. +[2026-03-25 22:56:50,499][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 12 and human policies 1. +[2026-03-25 22:56:50,499][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:57:05,818][__main__][INFO] - Number of regex retries in iteration 128: 0 +[2026-03-25 22:57:05,819][__main__][INFO] - agents played in iteration 128 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:57:06,606][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:57:06,630][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:57:06,652][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:57:06,675][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:57:06,676][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:57:06,676][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:57:07,299][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:57:07,755][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:57:08,264][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:57:08,766][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:57:09,268][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:57:09,768][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:57:10,270][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:57:10,772][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:57:11,274][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:57:11,770][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:57:12,271][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:57:12,772][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:57:13,278][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:57:13,778][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:57:14,280][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:57:14,780][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:57:15,281][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:57:15,779][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:57:16,282][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:57:16,782][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:57:17,278][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:57:17,777][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:57:18,277][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:57:18,777][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:57:19,274][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:57:19,773][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:57:20,275][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:57:20,776][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:57:21,276][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:57:21,777][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:57:22,278][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:57:22,777][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:57:23,275][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:57:23,773][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:57:24,274][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:57:24,774][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:57:25,275][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:57:25,774][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:57:26,274][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:57:26,773][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:57:27,271][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:57:27,791][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:57:28,298][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:57:28,801][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:57:29,302][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:57:29,805][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:57:30,307][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:57:30,809][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:57:31,311][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:57:31,810][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:57:32,312][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:57:32,812][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:57:33,311][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:57:33,813][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:57:34,313][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:57:34,815][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:57:35,312][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:57:35,812][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:57:36,308][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:57:36,804][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:57:37,302][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:57:37,801][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:57:38,299][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:57:38,798][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:57:39,297][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10759 tokens. +[2026-03-25 22:57:39,943][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 60.63%, Block Peak % of device VRAM: 62.33%, ΔTime: 00:00:32 +[2026-03-25 22:57:40,655][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:57:40,658][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:57:40,659][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:57:41,405][__main__][INFO] - Iteration 129 took 50s (30.09% Gen, 68.44% Train). Generation: 15s, Training: 34s. Estimated remaining time: 40h 33m 20s. Estimated total time: 42h 25m 21s. Time estimates for 10 more iterations: 8m 29s, 100 more iterations: 1h 24m 50s, 500 more iterations: 7h 4m 13s. +[2026-03-25 22:57:41,408][__main__][INFO] - Starting iteration 129. +[2026-03-25 22:57:41,807][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 12 and human policies 1. +[2026-03-25 22:57:41,807][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:57:51,596][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:57:56,950][__main__][INFO] - Number of regex retries in iteration 129: 1 +[2026-03-25 22:57:56,951][__main__][INFO] - agents played in iteration 129 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:57:57,725][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:57:57,749][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:57:57,772][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:57:57,795][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:57:57,796][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:57:57,796][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:57:58,391][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:57:58,848][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:57:59,353][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:57:59,855][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:58:00,361][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:58:00,861][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:58:01,364][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:58:01,867][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:58:02,369][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:58:02,868][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:58:03,370][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:58:03,870][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:58:04,373][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:58:04,872][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:58:05,371][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:58:05,871][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:58:06,372][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:58:06,873][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:58:07,371][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:58:07,869][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:58:08,372][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:58:08,872][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:58:09,372][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:58:09,872][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:58:10,374][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:58:10,876][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:58:11,375][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:58:11,876][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:58:12,378][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:58:12,880][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:58:13,381][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:58:13,882][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:58:14,386][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:58:14,887][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:58:15,386][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:58:15,888][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:58:16,389][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:58:16,889][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:58:17,388][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:58:17,891][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:58:18,394][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:58:18,893][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:58:19,392][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:58:19,894][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:58:20,397][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:58:20,896][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:58:21,395][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:58:21,897][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:58:22,399][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:58:22,895][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:58:23,393][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:58:23,894][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:58:24,393][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:58:24,891][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:58:25,387][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:58:25,889][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:58:26,387][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:58:26,885][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:58:27,385][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:58:27,886][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:58:28,388][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:58:28,886][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:58:29,386][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:58:29,886][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:58:30,385][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10814 tokens. +[2026-03-25 22:58:31,019][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 60.57%, Block Peak % of device VRAM: 62.24%, ΔTime: 00:00:32 +[2026-03-25 22:58:31,737][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:58:31,740][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:58:31,741][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:58:32,445][__main__][INFO] - Iteration 130 took 50s (29.91% Gen, 68.70% Train). Generation: 15s, Training: 34s. Estimated remaining time: 40h 19m 6s. Estimated total time: 42h 11m 58s. Time estimates for 10 more iterations: 8m 26s, 100 more iterations: 1h 24m 23s, 500 more iterations: 7h 1m 59s. +[2026-03-25 22:58:32,448][__main__][INFO] - Starting iteration 130. +[2026-03-25 22:58:33,111][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 12 and human policies 1. +[2026-03-25 22:58:33,112][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:58:47,777][__main__][INFO] - Number of regex retries in iteration 130: 0 +[2026-03-25 22:58:47,779][__main__][INFO] - agents played in iteration 130 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:58:48,560][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:58:48,583][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:58:48,606][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:58:48,629][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:58:48,630][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:58:48,630][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:58:49,238][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:58:49,695][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:58:50,201][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:58:50,702][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:58:51,204][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:58:51,709][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:58:52,214][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:58:52,715][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:58:53,218][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:58:53,723][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:58:54,224][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:58:54,722][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:58:55,224][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:58:55,728][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:58:56,232][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:58:56,729][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:58:57,232][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:58:57,736][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:58:58,239][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:58:58,738][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:58:59,239][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:58:59,745][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:59:00,247][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:59:00,748][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:59:01,251][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:59:01,752][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:59:02,252][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:59:02,752][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:59:03,254][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:59:03,756][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:59:04,256][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:59:04,755][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:59:05,257][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:59:05,765][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:59:06,266][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:59:06,766][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:59:07,270][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:59:07,773][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:59:08,276][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:59:08,777][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:59:09,283][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 22:59:09,786][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 22:59:10,287][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 22:59:10,786][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 22:59:11,288][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 22:59:11,789][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 22:59:12,291][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 22:59:12,790][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 22:59:13,292][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 22:59:13,794][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 22:59:14,293][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 22:59:14,792][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 22:59:15,294][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 22:59:15,796][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 22:59:16,295][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 22:59:16,794][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 22:59:17,295][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 22:59:17,797][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 22:59:18,295][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 22:59:18,795][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 22:59:19,299][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 22:59:19,801][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 22:59:20,301][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 22:59:20,800][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 22:59:21,304][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10814 tokens. +[2026-03-25 22:59:21,934][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.25%, Current % of VRAM taken: 60.67%, Block Peak % of device VRAM: 62.42%, ΔTime: 00:00:32 +[2026-03-25 22:59:22,667][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 22:59:22,669][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 22:59:22,671][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 22:59:24,009][__main__][INFO] - Iteration 131 took 50s (28.82% Gen, 68.55% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 31m 13s. Estimated total time: 42h 24m 57s. Time estimates for 10 more iterations: 8m 28s, 100 more iterations: 1h 24m 49s, 500 more iterations: 7h 4m 9s. +[2026-03-25 22:59:24,012][__main__][INFO] - Starting iteration 131. +[2026-03-25 22:59:24,415][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 13 and human policies 1. +[2026-03-25 22:59:24,415][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 22:59:34,131][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 22:59:38,471][__main__][INFO] - Number of regex retries in iteration 131: 1 +[2026-03-25 22:59:38,472][__main__][INFO] - agents played in iteration 131 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 22:59:39,277][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:59:39,300][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:59:39,323][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:59:39,346][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 22:59:39,347][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 22:59:39,347][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 22:59:39,949][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 22:59:40,408][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 22:59:40,915][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 22:59:41,417][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 22:59:41,918][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 22:59:42,419][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 22:59:42,923][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 22:59:43,428][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 22:59:43,929][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 22:59:44,433][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 22:59:44,936][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 22:59:45,438][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 22:59:45,934][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 22:59:46,436][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 22:59:46,940][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 22:59:47,445][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 22:59:47,947][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 22:59:48,448][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 22:59:48,951][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 22:59:49,454][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 22:59:49,955][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 22:59:50,456][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 22:59:50,958][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 22:59:51,459][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 22:59:51,955][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 22:59:52,456][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 22:59:52,959][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 22:59:53,460][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 22:59:53,960][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 22:59:54,461][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 22:59:54,960][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 22:59:55,464][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 22:59:55,963][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 22:59:56,465][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 22:59:56,968][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 22:59:57,471][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 22:59:57,972][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 22:59:58,474][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 22:59:58,976][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 22:59:59,475][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 22:59:59,976][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:00:00,478][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:00:00,979][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:00:01,476][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:00:01,974][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:00:02,478][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:00:02,979][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:00:03,477][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:00:03,976][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:00:04,474][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:00:04,976][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:00:05,475][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:00:05,975][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:00:06,477][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:00:06,978][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:00:07,476][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:00:07,976][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:00:08,478][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:00:08,981][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:00:09,481][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:00:09,981][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:00:10,484][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:00:10,987][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:00:11,488][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:00:11,987][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10839 tokens. +[2026-03-25 23:00:12,637][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 60.64%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 +[2026-03-25 23:00:13,360][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:00:13,363][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:00:13,364][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:00:14,171][__main__][INFO] - Iteration 132 took 49s (28.25% Gen, 70.13% Train). Generation: 14s, Training: 34s. Estimated remaining time: 39h 33m 17s. Estimated total time: 41h 27m 50s. Time estimates for 10 more iterations: 8m 17s, 100 more iterations: 1h 22m 55s, 500 more iterations: 6h 54m 38s. +[2026-03-25 23:00:14,173][__main__][INFO] - Starting iteration 132. +[2026-03-25 23:00:14,577][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 13 and human policies 1. +[2026-03-25 23:00:14,577][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:00:29,932][__main__][INFO] - Number of regex retries in iteration 132: 0 +[2026-03-25 23:00:29,933][__main__][INFO] - agents played in iteration 132 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:00:30,725][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:00:30,748][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:00:30,771][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:00:30,794][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:00:30,794][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:00:30,795][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:00:31,402][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:00:31,859][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:00:32,365][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:00:32,868][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:00:33,368][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:00:33,869][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:00:34,373][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:00:34,874][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:00:35,374][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:00:35,872][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:00:36,376][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:00:36,878][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:00:37,377][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:00:37,876][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:00:38,378][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:00:38,878][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:00:39,378][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:00:39,880][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:00:40,379][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:00:40,879][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:00:41,380][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:00:41,883][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:00:42,385][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:00:42,886][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:00:43,386][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:00:43,889][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:00:44,395][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:00:44,895][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:00:45,395][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:00:45,898][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:00:46,400][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:00:46,899][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:00:47,402][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:00:47,904][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:00:48,405][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:00:48,903][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:00:49,402][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:00:49,903][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:00:50,400][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:00:50,902][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:00:51,402][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:00:51,905][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:00:52,407][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:00:52,905][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:00:53,405][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:00:53,907][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:00:54,409][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:00:54,910][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:00:55,410][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:00:55,913][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:00:56,413][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:00:56,911][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:00:57,410][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:00:57,913][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:00:58,415][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:00:58,911][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:00:59,413][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:00:59,910][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:01:00,412][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:01:00,909][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:01:01,407][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:01:01,911][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:01:02,411][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:01:02,909][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:01:03,410][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10782 tokens. +[2026-03-25 23:01:04,056][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.58%, Block Peak % of device VRAM: 62.40%, ΔTime: 00:00:32 +[2026-03-25 23:01:04,827][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:01:04,829][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:01:04,831][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:01:05,675][__main__][INFO] - Iteration 133 took 51s (30.05% Gen, 68.29% Train). Generation: 15s, Training: 34s. Estimated remaining time: 40h 39m 32s. Estimated total time: 42h 34m 57s. Time estimates for 10 more iterations: 8m 30s, 100 more iterations: 1h 25m 9s, 500 more iterations: 7h 5m 49s. +[2026-03-25 23:01:05,678][__main__][INFO] - Starting iteration 133. +[2026-03-25 23:01:06,081][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 13 and human policies 1. +[2026-03-25 23:01:06,082][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:01:21,115][__main__][INFO] - Number of regex retries in iteration 133: 0 +[2026-03-25 23:01:21,116][__main__][INFO] - agents played in iteration 133 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:01:21,902][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:01:21,925][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:01:21,948][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:01:21,972][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:01:21,972][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:01:21,973][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:01:22,587][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:01:23,043][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:01:23,548][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:01:24,050][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:01:24,554][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:01:25,055][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:01:25,557][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:01:26,057][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:01:26,560][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:01:27,058][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:01:27,558][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:01:28,060][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:01:28,563][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:01:29,062][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:01:29,559][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:01:30,062][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:01:30,562][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:01:31,060][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:01:31,562][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:01:32,067][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:01:32,567][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:01:33,066][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:01:33,569][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:01:34,072][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:01:34,573][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:01:35,071][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:01:35,572][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:01:36,076][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:01:36,576][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:01:37,074][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:01:37,573][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:01:38,075][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:01:38,577][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:01:39,077][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:01:39,580][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:01:40,082][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:01:40,583][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:01:41,081][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:01:41,586][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:01:42,088][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:01:42,591][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:01:43,089][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:01:43,590][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:01:44,091][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:01:44,593][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:01:45,091][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:01:45,591][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:01:46,089][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:01:46,589][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:01:47,088][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:01:47,589][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:01:48,088][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:01:48,587][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:01:49,085][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:01:49,588][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:01:50,087][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:01:50,586][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:01:51,085][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:01:51,585][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:01:52,086][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:01:52,585][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:01:53,085][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:01:53,586][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:01:54,086][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:01:54,586][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10765 tokens. +[2026-03-25 23:01:55,230][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 60.66%, Block Peak % of device VRAM: 62.27%, ΔTime: 00:00:32 +[2026-03-25 23:01:55,923][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:01:55,925][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:01:55,927][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:01:56,720][__main__][INFO] - Iteration 134 took 50s (29.69% Gen, 68.74% Train). Generation: 15s, Training: 34s. Estimated remaining time: 40h 15m 42s. Estimated total time: 42h 11m 58s. Time estimates for 10 more iterations: 8m 26s, 100 more iterations: 1h 24m 23s, 500 more iterations: 7h 1m 59s. +[2026-03-25 23:01:56,723][__main__][INFO] - Starting iteration 134. +[2026-03-25 23:01:57,123][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 13 and human policies 1. +[2026-03-25 23:01:57,124][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:02:10,714][__main__][INFO] - Number of regex retries in iteration 134: 0 +[2026-03-25 23:02:10,715][__main__][INFO] - agents played in iteration 134 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:02:11,495][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:02:11,518][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:02:11,541][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:02:11,564][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:02:11,565][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:02:11,565][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:02:12,188][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:02:12,644][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:02:13,148][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:02:13,648][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:02:14,148][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:02:14,651][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:02:15,151][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:02:15,652][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:02:16,152][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:02:16,652][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:02:17,155][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:02:17,656][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:02:18,159][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:02:18,659][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:02:19,159][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:02:19,660][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:02:20,165][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:02:20,669][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:02:21,172][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:02:21,672][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:02:22,173][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:02:22,687][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:02:23,193][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:02:23,694][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:02:24,197][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:02:24,698][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:02:25,201][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:02:25,702][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:02:26,204][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:02:26,700][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:02:27,203][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:02:27,703][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:02:28,205][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:02:28,704][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:02:29,204][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:02:29,705][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:02:30,207][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:02:30,705][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:02:31,203][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:02:31,707][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:02:32,207][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:02:32,707][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:02:33,209][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:02:33,710][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:02:34,210][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:02:34,711][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:02:35,214][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:02:35,714][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:02:36,213][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:02:36,712][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:02:37,212][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:02:37,711][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:02:38,210][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:02:38,709][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:02:39,208][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:02:39,710][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:02:40,210][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:02:40,710][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:02:41,210][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:02:41,710][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:02:42,210][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:02:42,711][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:02:43,212][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:02:43,714][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:02:44,213][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10829 tokens. +[2026-03-25 23:02:44,851][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 60.65%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:32 +[2026-03-25 23:02:45,565][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:02:45,568][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:02:45,569][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:02:46,263][__main__][INFO] - Iteration 135 took 49s (27.66% Gen, 70.93% Train). Generation: 13s, Training: 34s. Estimated remaining time: 38h 59m 53s. Estimated total time: 40h 56m 59s. Time estimates for 10 more iterations: 8m 11s, 100 more iterations: 1h 21m 53s, 500 more iterations: 6h 49m 29s. +[2026-03-25 23:02:46,265][__main__][INFO] - Starting iteration 135. +[2026-03-25 23:02:46,664][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 13 and human policies 1. +[2026-03-25 23:02:46,664][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:02:50,941][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 23:03:01,882][__main__][INFO] - Number of regex retries in iteration 135: 1 +[2026-03-25 23:03:01,882][__main__][INFO] - agents played in iteration 135 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:03:02,657][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:03:02,680][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:03:02,703][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:03:02,726][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:03:02,727][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:03:02,727][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:03:03,335][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:03:03,786][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:03:04,294][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:03:04,794][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:03:05,295][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:03:05,791][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:03:06,291][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:03:06,791][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:03:07,294][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:03:07,794][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:03:08,295][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:03:08,791][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:03:09,291][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:03:09,787][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:03:10,287][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:03:10,785][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:03:11,286][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:03:11,787][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:03:12,288][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:03:12,788][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:03:13,289][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:03:13,792][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:03:14,290][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:03:14,790][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:03:15,291][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:03:15,789][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:03:16,287][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:03:16,787][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:03:17,288][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:03:17,789][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:03:18,287][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:03:18,786][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:03:19,286][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:03:19,788][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:03:20,289][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:03:20,790][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:03:21,290][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:03:21,792][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:03:22,291][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:03:22,790][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:03:23,291][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:03:23,798][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:03:24,299][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:03:24,799][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:03:25,317][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:03:25,819][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:03:26,318][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:03:26,819][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:03:27,323][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:03:27,825][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:03:28,324][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:03:28,827][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:03:29,328][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:03:29,829][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:03:30,328][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:03:30,830][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:03:31,331][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:03:31,833][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:03:32,332][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:03:32,834][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:03:33,336][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:03:33,838][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:03:34,338][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:03:34,840][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:03:35,341][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10847 tokens. +[2026-03-25 23:03:35,984][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.29%, Current % of VRAM taken: 60.71%, Block Peak % of device VRAM: 62.41%, ΔTime: 00:00:32 +[2026-03-25 23:03:36,706][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:03:36,708][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:03:36,710][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:03:37,404][__main__][INFO] - Iteration 136 took 50s (29.99% Gen, 68.64% Train). Generation: 15s, Training: 34s. Estimated remaining time: 40h 19m 7s. Estimated total time: 42h 17m 4s. Time estimates for 10 more iterations: 8m 27s, 100 more iterations: 1h 24m 34s, 500 more iterations: 7h 2m 50s. +[2026-03-25 23:03:37,407][__main__][INFO] - Starting iteration 136. +[2026-03-25 23:03:37,806][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 13 and human policies 1. +[2026-03-25 23:03:37,806][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:03:51,927][__main__][INFO] - Number of regex retries in iteration 136: 0 +[2026-03-25 23:03:51,928][__main__][INFO] - agents played in iteration 136 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:03:52,713][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:03:52,737][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:03:52,760][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:03:52,782][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:03:52,783][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:03:52,783][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:03:53,389][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:03:53,845][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:03:54,349][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:03:54,851][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:03:55,354][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:03:55,855][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:03:56,354][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:03:56,855][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:03:57,354][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:03:57,852][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:03:58,350][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:03:58,852][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:03:59,352][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:03:59,851][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:04:00,349][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:04:00,851][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:04:01,350][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:04:01,850][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:04:02,349][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:04:02,850][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:04:03,350][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:04:03,849][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:04:04,348][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:04:04,849][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:04:05,350][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:04:05,853][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:04:06,355][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:04:06,853][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:04:07,356][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:04:07,855][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:04:08,356][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:04:08,855][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:04:09,356][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:04:09,855][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:04:10,357][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:04:10,858][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:04:11,360][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:04:11,859][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:04:12,360][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:04:12,863][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:04:13,365][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:04:13,865][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:04:14,364][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:04:14,864][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:04:15,363][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:04:15,862][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:04:16,362][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:04:16,860][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:04:17,360][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:04:17,861][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:04:18,363][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:04:18,864][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:04:19,379][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:04:19,879][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:04:20,379][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:04:20,880][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:04:21,382][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:04:21,883][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:04:22,388][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:04:22,888][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:04:23,389][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:04:23,888][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:04:24,390][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:04:24,890][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:04:25,391][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10849 tokens. +[2026-03-25 23:04:26,028][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.25%, Current % of VRAM taken: 60.67%, Block Peak % of device VRAM: 62.34%, ΔTime: 00:00:32 +[2026-03-25 23:04:26,760][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:04:26,763][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:04:26,764][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:04:27,504][__main__][INFO] - Iteration 137 took 49s (28.41% Gen, 70.10% Train). Generation: 14s, Training: 34s. Estimated remaining time: 39h 26m 8s. Estimated total time: 41h 24m 55s. Time estimates for 10 more iterations: 8m 16s, 100 more iterations: 1h 22m 49s, 500 more iterations: 6h 54m 9s. +[2026-03-25 23:04:27,506][__main__][INFO] - Starting iteration 137. +[2026-03-25 23:04:27,906][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 13 and human policies 1. +[2026-03-25 23:04:27,906][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:04:38,787][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 23:04:42,185][__main__][INFO] - Number of regex retries in iteration 137: 1 +[2026-03-25 23:04:42,186][__main__][INFO] - agents played in iteration 137 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:04:42,965][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:04:42,988][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:04:43,011][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:04:43,033][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:04:43,034][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:04:43,034][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:04:43,655][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:04:44,111][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:04:44,616][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:04:45,115][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:04:45,616][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:04:46,117][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:04:46,617][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:04:47,116][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:04:47,616][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:04:48,116][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:04:48,614][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:04:49,113][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:04:49,613][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:04:50,113][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:04:50,615][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:04:51,115][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:04:51,614][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:04:52,113][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:04:52,616][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:04:53,117][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:04:53,619][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:04:54,119][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:04:54,620][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:04:55,118][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:04:55,619][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:04:56,119][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:04:56,619][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:04:57,118][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:04:57,619][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:04:58,116][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:04:58,616][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:04:59,114][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:04:59,615][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:05:00,116][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:05:00,620][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:05:01,118][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:05:01,620][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:05:02,121][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:05:02,623][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:05:03,125][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:05:03,627][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:05:04,128][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:05:04,630][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:05:05,129][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:05:05,630][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:05:06,130][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:05:06,630][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:05:07,130][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:05:07,631][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:05:08,133][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:05:08,632][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:05:09,132][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:05:09,632][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:05:10,130][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:05:10,629][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:05:11,129][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:05:11,630][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:05:12,130][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:05:12,631][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:05:13,130][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:05:13,632][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:05:14,132][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:05:14,633][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:05:15,131][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:05:15,632][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10829 tokens. +[2026-03-25 23:05:16,281][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 60.64%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 +[2026-03-25 23:05:17,014][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:05:17,016][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:05:17,018][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:05:17,777][__main__][INFO] - Iteration 138 took 49s (28.63% Gen, 69.84% Train). Generation: 14s, Training: 34s. Estimated remaining time: 39h 33m 59s. Estimated total time: 41h 33m 36s. Time estimates for 10 more iterations: 8m 18s, 100 more iterations: 1h 23m 7s, 500 more iterations: 6h 55m 36s. +[2026-03-25 23:05:17,779][__main__][INFO] - Starting iteration 138. +[2026-03-25 23:05:18,179][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 13 and human policies 1. +[2026-03-25 23:05:18,180][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:05:33,294][__main__][INFO] - Number of regex retries in iteration 138: 0 +[2026-03-25 23:05:33,294][__main__][INFO] - agents played in iteration 138 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:05:34,071][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:05:34,094][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:05:34,117][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:05:34,140][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:05:34,140][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:05:34,141][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:05:34,734][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:05:35,189][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:05:35,693][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:05:36,194][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:05:36,693][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:05:37,192][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:05:37,691][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:05:38,190][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:05:38,690][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:05:39,189][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:05:39,686][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:05:40,185][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:05:40,684][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:05:41,182][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:05:41,677][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:05:42,175][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:05:42,674][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:05:43,171][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:05:43,670][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:05:44,169][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:05:44,667][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:05:45,165][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:05:45,663][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:05:46,163][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:05:46,661][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:05:47,159][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:05:47,657][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:05:48,155][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:05:48,654][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:05:49,152][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:05:49,651][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:05:50,151][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:05:50,650][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:05:51,149][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:05:51,648][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:05:52,146][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:05:52,645][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:05:53,145][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:05:53,644][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:05:54,144][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:05:54,644][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:05:55,144][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:05:55,643][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:05:56,140][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:05:56,638][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:05:57,135][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:05:57,632][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:05:58,130][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:05:58,628][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:05:59,127][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:05:59,627][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:06:00,126][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:06:00,627][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:06:01,126][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:06:01,626][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:06:02,128][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:06:02,628][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:06:03,126][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:06:03,623][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:06:04,121][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:06:04,617][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:06:05,115][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:06:05,613][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:06:06,111][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:06:06,609][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10831 tokens. +[2026-03-25 23:06:07,229][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 60.66%, Block Peak % of device VRAM: 62.34%, ΔTime: 00:00:32 +[2026-03-25 23:06:07,959][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:06:07,961][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:06:07,963][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:06:08,682][__main__][INFO] - Iteration 139 took 50s (29.93% Gen, 68.64% Train). Generation: 15s, Training: 34s. Estimated remaining time: 40h 4m 44s. Estimated total time: 42h 5m 12s. Time estimates for 10 more iterations: 8m 25s, 100 more iterations: 1h 24m 10s, 500 more iterations: 7h 0m 52s. +[2026-03-25 23:06:08,685][__main__][INFO] - Starting iteration 139. +[2026-03-25 23:06:09,083][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 13 and human policies 1. +[2026-03-25 23:06:09,084][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:06:17,238][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 23:06:25,853][__main__][INFO] - Number of regex retries in iteration 139: 1 +[2026-03-25 23:06:25,854][__main__][INFO] - agents played in iteration 139 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:06:26,628][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:06:26,651][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:06:26,674][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:06:26,696][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:06:26,697][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:06:26,698][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:06:27,300][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:06:27,757][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:06:28,262][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:06:28,761][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:06:29,261][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:06:29,760][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:06:30,259][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:06:30,758][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:06:31,258][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:06:31,756][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:06:32,257][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:06:32,756][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:06:33,255][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:06:33,753][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:06:34,250][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:06:34,749][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:06:35,248][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:06:35,749][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:06:36,250][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:06:36,751][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:06:37,252][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:06:37,754][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:06:38,253][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:06:38,754][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:06:39,254][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:06:39,754][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:06:40,253][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:06:40,753][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:06:41,254][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:06:41,754][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:06:42,254][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:06:42,754][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:06:43,254][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:06:43,753][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:06:44,252][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:06:44,754][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:06:45,254][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:06:45,754][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:06:46,253][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:06:46,753][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:06:47,252][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:06:47,749][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:06:48,245][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:06:48,742][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:06:49,240][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:06:49,735][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:06:50,232][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:06:50,726][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:06:51,220][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:06:51,715][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:06:52,211][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:06:52,708][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:06:53,205][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:06:53,702][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:06:54,198][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:06:54,694][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:06:55,191][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:06:55,690][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:06:56,187][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:06:56,686][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:06:57,184][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:06:57,681][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:06:58,179][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:06:58,676][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:06:59,174][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10831 tokens. +[2026-03-25 23:06:59,806][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 60.61%, Block Peak % of device VRAM: 62.34%, ΔTime: 00:00:32 +[2026-03-25 23:07:00,531][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:07:00,533][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:07:00,535][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:07:01,175][__main__][INFO] - Iteration 140 took 52s (32.19% Gen, 66.58% Train). Generation: 16s, Training: 34s. Estimated remaining time: 41h 23m 17s. Estimated total time: 43h 24m 38s. Time estimates for 10 more iterations: 8m 40s, 100 more iterations: 1h 26m 49s, 500 more iterations: 7h 14m 6s. +[2026-03-25 23:07:01,183][__main__][INFO] - Starting iteration 140. +[2026-03-25 23:07:01,584][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 13 and human policies 1. +[2026-03-25 23:07:01,585][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:07:09,182][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 23:07:15,981][__main__][INFO] - Number of regex retries in iteration 140: 1 +[2026-03-25 23:07:15,982][__main__][INFO] - agents played in iteration 140 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:07:16,778][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:07:16,802][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:07:16,824][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:07:16,847][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:07:16,848][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:07:16,849][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:07:17,443][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:07:17,898][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:07:18,401][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:07:18,901][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:07:19,399][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:07:19,898][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:07:20,396][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:07:20,896][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:07:21,394][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:07:21,894][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:07:22,392][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:07:22,891][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:07:23,390][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:07:23,890][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:07:24,388][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:07:24,888][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:07:25,387][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:07:25,886][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:07:26,385][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:07:26,884][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:07:27,383][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:07:27,884][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:07:28,383][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:07:28,882][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:07:29,379][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:07:29,877][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:07:30,375][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:07:30,871][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:07:31,370][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:07:31,872][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:07:32,371][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:07:32,870][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:07:33,370][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:07:33,869][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:07:34,368][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:07:34,867][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:07:35,365][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:07:35,864][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:07:36,364][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:07:36,863][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:07:37,361][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:07:37,860][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:07:38,358][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:07:38,857][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:07:39,356][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:07:39,855][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:07:40,354][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:07:40,853][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:07:41,352][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:07:41,851][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:07:42,349][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:07:42,847][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:07:43,347][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:07:43,843][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:07:44,342][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:07:44,840][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:07:45,337][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:07:45,836][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:07:46,332][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:07:46,832][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:07:47,330][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:07:47,829][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:07:48,329][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:07:48,827][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:07:49,326][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10859 tokens. +[2026-03-25 23:07:49,953][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 60.61%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 +[2026-03-25 23:07:50,680][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:07:50,682][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:07:50,684][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:07:51,938][__main__][INFO] - Iteration 141 took 50s (28.59% Gen, 68.92% Train). Generation: 14s, Training: 34s. Estimated remaining time: 39h 55m 33s. Estimated total time: 41h 57m 44s. Time estimates for 10 more iterations: 8m 23s, 100 more iterations: 1h 23m 55s, 500 more iterations: 6h 59m 37s. +[2026-03-25 23:07:51,941][__main__][INFO] - Starting iteration 141. +[2026-03-25 23:07:52,341][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 14 and human policies 1. +[2026-03-25 23:07:52,341][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:07:55,743][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 23:08:07,483][__main__][INFO] - Number of regex retries in iteration 141: 1 +[2026-03-25 23:08:07,484][__main__][INFO] - agents played in iteration 141 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:08:08,267][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:08:08,290][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:08:08,313][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:08:08,336][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:08:08,337][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:08:08,337][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:08:08,938][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:08:09,393][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:08:09,895][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:08:10,393][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:08:10,892][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:08:11,391][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:08:11,889][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:08:12,389][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:08:12,887][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:08:13,386][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:08:13,885][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:08:14,383][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:08:14,880][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:08:15,379][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:08:15,876][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:08:16,374][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:08:16,872][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:08:17,371][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:08:17,871][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:08:18,368][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:08:18,867][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:08:19,364][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:08:19,861][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:08:20,359][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:08:20,857][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:08:21,354][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:08:21,853][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:08:22,351][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:08:22,850][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:08:23,349][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:08:23,847][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:08:24,346][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:08:24,845][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:08:25,341][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:08:25,836][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:08:26,332][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:08:26,830][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:08:27,325][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:08:27,824][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:08:28,320][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:08:28,817][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:08:29,316][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:08:29,815][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:08:30,313][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:08:30,811][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:08:31,310][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:08:31,807][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:08:32,307][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:08:32,805][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:08:33,307][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:08:33,805][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:08:34,300][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:08:34,800][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:08:35,299][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:08:35,799][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:08:36,297][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:08:36,796][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:08:37,292][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:08:37,790][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:08:38,284][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:08:38,787][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:08:39,285][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:08:39,779][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:08:40,280][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:08:40,780][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10842 tokens. +[2026-03-25 23:08:41,424][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 60.63%, Block Peak % of device VRAM: 62.28%, ΔTime: 00:00:32 +[2026-03-25 23:08:42,152][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:08:42,154][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:08:42,156][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:08:42,887][__main__][INFO] - Iteration 142 took 50s (29.96% Gen, 68.59% Train). Generation: 15s, Training: 34s. Estimated remaining time: 40h 4m 18s. Estimated total time: 42h 7m 20s. Time estimates for 10 more iterations: 8m 25s, 100 more iterations: 1h 24m 14s, 500 more iterations: 7h 1m 13s. +[2026-03-25 23:08:42,889][__main__][INFO] - Starting iteration 142. +[2026-03-25 23:08:43,293][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 14 and human policies 1. +[2026-03-25 23:08:43,294][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:08:49,255][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 23:08:57,995][__main__][INFO] - Number of regex retries in iteration 142: 1 +[2026-03-25 23:08:57,996][__main__][INFO] - agents played in iteration 142 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:08:58,775][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:08:58,798][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:08:58,821][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:08:58,843][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:08:58,844][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:08:58,844][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:08:59,450][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:08:59,905][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:09:00,408][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:09:00,907][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:09:01,406][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:09:01,907][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:09:02,407][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:09:02,906][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:09:03,404][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:09:03,901][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:09:04,399][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:09:04,896][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:09:05,391][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:09:05,886][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:09:06,385][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:09:06,881][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:09:07,379][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:09:07,877][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:09:08,376][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:09:08,875][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:09:09,374][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:09:09,872][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:09:10,371][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:09:10,869][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:09:11,368][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:09:11,863][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:09:12,360][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:09:12,856][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:09:13,354][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:09:13,851][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:09:14,351][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:09:14,850][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:09:15,348][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:09:15,846][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:09:16,346][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:09:16,846][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:09:17,346][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:09:17,845][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:09:18,343][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:09:18,840][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:09:19,339][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:09:19,836][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:09:20,336][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:09:20,834][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:09:21,333][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:09:21,832][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:09:22,331][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:09:22,843][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:09:23,342][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:09:23,843][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:09:24,341][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:09:24,840][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:09:25,338][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:09:25,836][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:09:26,334][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:09:26,832][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:09:27,330][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:09:27,828][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:09:28,326][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:09:28,824][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:09:29,323][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:09:29,822][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:09:30,318][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:09:30,816][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:09:31,315][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10833 tokens. +[2026-03-25 23:09:31,979][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 60.60%, Block Peak % of device VRAM: 62.28%, ΔTime: 00:00:32 +[2026-03-25 23:09:32,713][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:09:32,715][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:09:32,716][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:09:33,369][__main__][INFO] - Iteration 143 took 50s (29.36% Gen, 69.33% Train). Generation: 14s, Training: 34s. Estimated remaining time: 39h 39m 57s. Estimated total time: 41h 43m 50s. Time estimates for 10 more iterations: 8m 20s, 100 more iterations: 1h 23m 27s, 500 more iterations: 6h 57m 18s. +[2026-03-25 23:09:33,374][__main__][INFO] - Starting iteration 143. +[2026-03-25 23:09:33,777][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 14 and human policies 1. +[2026-03-25 23:09:33,778][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:09:37,016][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 23:09:39,077][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 23:09:47,854][__main__][INFO] - Number of regex retries in iteration 143: 2 +[2026-03-25 23:09:47,855][__main__][INFO] - agents played in iteration 143 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:09:48,632][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:09:48,655][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:09:48,678][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:09:48,701][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:09:48,701][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:09:48,702][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:09:49,302][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:09:49,758][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:09:50,262][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:09:50,761][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:09:51,259][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:09:51,757][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:09:52,257][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:09:52,754][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:09:53,252][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:09:53,752][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:09:54,251][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:09:54,749][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:09:55,247][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:09:55,745][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:09:56,241][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:09:56,739][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:09:57,236][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:09:57,736][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:09:58,234][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:09:58,732][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:09:59,231][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:09:59,730][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:10:00,228][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:10:00,726][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:10:01,226][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:10:01,726][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:10:02,231][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:10:02,731][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:10:03,231][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:10:03,730][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:10:04,245][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:10:04,786][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:10:05,286][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:10:05,786][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:10:06,287][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:10:06,786][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:10:07,286][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:10:07,786][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:10:08,286][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:10:08,786][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:10:09,285][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:10:09,787][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:10:10,286][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:10:10,784][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:10:11,285][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:10:11,782][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:10:12,282][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:10:12,780][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:10:13,278][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:10:13,776][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:10:14,274][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:10:14,771][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:10:15,270][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:10:15,769][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:10:16,267][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:10:16,765][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:10:17,264][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:10:17,763][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:10:18,261][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:10:18,759][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:10:19,257][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:10:19,756][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:10:20,254][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:10:20,752][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:10:21,251][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10837 tokens. +[2026-03-25 23:10:21,889][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 60.67%, Block Peak % of device VRAM: 62.34%, ΔTime: 00:00:32 +[2026-03-25 23:10:22,613][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:10:22,615][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:10:22,617][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:10:23,318][__main__][INFO] - Iteration 144 took 49s (28.41% Gen, 70.17% Train). Generation: 14s, Training: 34s. Estimated remaining time: 39h 12m 20s. Estimated total time: 41h 17m 3s. Time estimates for 10 more iterations: 8m 15s, 100 more iterations: 1h 22m 34s, 500 more iterations: 6h 52m 50s. +[2026-03-25 23:10:23,322][__main__][INFO] - Starting iteration 144. +[2026-03-25 23:10:23,719][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 14 and human policies 1. +[2026-03-25 23:10:23,720][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:10:37,475][__main__][INFO] - Number of regex retries in iteration 144: 0 +[2026-03-25 23:10:37,476][__main__][INFO] - agents played in iteration 144 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:10:38,248][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:10:38,272][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:10:38,295][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:10:38,317][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:10:38,318][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:10:38,318][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:10:38,920][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:10:39,374][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:10:39,876][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:10:40,376][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:10:40,877][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:10:41,378][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:10:41,877][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:10:42,377][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:10:42,877][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:10:43,376][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:10:43,877][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:10:44,375][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:10:44,874][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:10:45,373][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:10:45,872][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:10:46,371][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:10:46,867][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:10:47,367][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:10:47,867][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:10:48,369][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:10:48,870][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:10:49,371][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:10:49,872][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:10:50,372][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:10:50,874][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:10:51,372][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:10:51,871][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:10:52,370][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:10:52,868][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:10:53,367][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:10:53,863][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:10:54,362][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:10:54,863][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:10:55,363][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:10:55,863][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:10:56,360][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:10:56,859][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:10:57,358][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:10:57,856][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:10:58,355][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:10:58,854][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:10:59,351][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:10:59,849][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:11:00,345][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:11:00,841][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:11:01,337][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:11:01,835][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:11:02,334][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:11:02,831][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:11:03,328][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:11:03,826][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:11:04,324][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:11:04,820][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:11:05,318][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:11:05,812][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:11:06,310][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:11:06,808][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:11:07,306][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:11:07,805][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:11:08,304][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:11:08,803][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:11:09,303][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:11:09,801][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:11:10,299][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:11:10,797][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10851 tokens. +[2026-03-25 23:11:11,438][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 60.65%, Block Peak % of device VRAM: 62.49%, ΔTime: 00:00:32 +[2026-03-25 23:11:12,165][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:11:12,168][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:11:12,169][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:11:12,875][__main__][INFO] - Iteration 145 took 49s (27.98% Gen, 70.58% Train). Generation: 13s, Training: 34s. Estimated remaining time: 38h 52m 14s. Estimated total time: 40h 57m 47s. Time estimates for 10 more iterations: 8m 11s, 100 more iterations: 1h 21m 55s, 500 more iterations: 6h 49m 37s. +[2026-03-25 23:11:12,877][__main__][INFO] - Starting iteration 145. +[2026-03-25 23:11:13,278][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 14 and human policies 1. +[2026-03-25 23:11:13,279][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:11:16,774][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 23:11:27,159][__main__][INFO] - Number of regex retries in iteration 145: 1 +[2026-03-25 23:11:27,160][__main__][INFO] - agents played in iteration 145 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:11:27,945][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:11:27,968][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:11:27,991][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:11:28,015][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:11:28,016][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:11:28,017][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:11:28,635][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:11:29,090][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:11:29,594][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:11:30,094][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:11:30,594][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:11:31,091][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:11:31,589][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:11:32,090][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:11:32,591][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:11:33,090][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:11:33,588][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:11:34,088][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:11:34,586][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:11:35,084][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:11:35,583][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:11:36,081][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:11:36,577][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:11:37,078][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:11:37,577][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:11:38,077][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:11:38,576][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:11:39,074][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:11:39,573][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:11:40,073][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:11:40,572][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:11:41,071][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:11:41,568][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:11:42,068][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:11:42,565][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:11:43,063][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:11:43,561][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:11:44,057][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:11:44,555][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:11:45,053][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:11:45,551][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:11:46,049][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:11:46,548][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:11:47,046][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:11:47,544][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:11:48,042][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:11:48,540][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:11:49,039][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:11:49,537][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:11:50,036][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:11:50,535][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:11:51,033][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:11:51,531][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:11:52,029][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:11:52,528][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:11:53,029][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:11:53,533][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:11:54,039][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:11:54,541][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:11:55,049][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:11:55,553][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:11:56,059][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:11:56,564][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:11:57,068][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:11:57,574][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:11:58,079][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:11:58,585][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:11:59,089][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:11:59,594][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:12:00,098][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:12:00,603][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10864 tokens. +[2026-03-25 23:12:01,320][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 60.63%, Block Peak % of device VRAM: 62.20%, ΔTime: 00:00:32 +[2026-03-25 23:12:02,071][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:12:02,073][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:12:02,075][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:12:02,850][__main__][INFO] - Iteration 146 took 49s (28.00% Gen, 70.43% Train). Generation: 13s, Training: 34s. Estimated remaining time: 39h 12m 15s. Estimated total time: 41h 18m 38s. Time estimates for 10 more iterations: 8m 15s, 100 more iterations: 1h 22m 37s, 500 more iterations: 6h 53m 6s. +[2026-03-25 23:12:02,853][__main__][INFO] - Starting iteration 146. +[2026-03-25 23:12:03,255][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 14 and human policies 1. +[2026-03-25 23:12:03,255][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:12:19,597][__main__][INFO] - Number of regex retries in iteration 146: 0 +[2026-03-25 23:12:19,598][__main__][INFO] - agents played in iteration 146 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:12:20,412][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:12:20,436][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:12:20,461][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:12:20,484][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:12:20,486][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:12:20,486][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:12:21,170][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:12:21,632][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:12:22,145][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:12:22,651][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:12:23,157][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:12:23,663][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:12:24,169][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:12:24,674][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:12:25,179][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:12:25,691][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:12:26,199][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:12:26,710][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:12:27,217][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:12:27,723][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:12:28,230][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:12:28,737][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:12:29,248][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:12:29,755][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:12:30,259][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:12:30,764][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:12:31,269][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:12:31,772][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:12:32,277][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:12:32,783][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:12:33,289][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:12:33,795][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:12:34,300][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:12:34,804][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:12:35,310][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:12:35,815][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:12:36,321][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:12:36,825][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:12:37,333][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:12:37,837][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:12:38,342][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:12:38,848][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:12:39,354][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:12:39,861][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:12:40,367][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:12:40,873][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:12:41,380][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:12:41,884][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:12:42,391][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:12:42,896][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:12:43,400][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:12:43,905][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:12:44,410][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:12:44,915][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:12:45,420][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:12:45,925][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:12:46,431][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:12:46,935][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:12:47,440][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:12:47,945][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:12:48,451][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:12:48,956][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:12:49,463][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:12:49,968][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:12:50,474][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:12:50,980][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:12:51,487][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:12:51,992][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:12:52,496][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:12:53,003][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:12:53,509][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10768 tokens. +[2026-03-25 23:12:54,223][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.29%, Current % of VRAM taken: 60.72%, Block Peak % of device VRAM: 62.58%, ΔTime: 00:00:33 +[2026-03-25 23:12:54,980][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:12:54,983][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:12:54,985][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:12:55,753][__main__][INFO] - Iteration 147 took 52s (31.13% Gen, 67.40% Train). Generation: 16s, Training: 35s. Estimated remaining time: 41h 37m 41s. Estimated total time: 43h 44m 56s. Time estimates for 10 more iterations: 8m 44s, 100 more iterations: 1h 27m 29s, 500 more iterations: 7h 17m 29s. +[2026-03-25 23:12:55,755][__main__][INFO] - Starting iteration 147. +[2026-03-25 23:12:56,157][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 14 and human policies 1. +[2026-03-25 23:12:56,157][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:13:10,579][__main__][INFO] - Number of regex retries in iteration 147: 0 +[2026-03-25 23:13:10,580][__main__][INFO] - agents played in iteration 147 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:13:11,400][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.61%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:13:11,424][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.61%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:13:11,449][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.61%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:13:11,473][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.61%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:13:11,474][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:13:11,474][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:13:12,152][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:13:12,614][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:13:13,125][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:13:13,631][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:13:14,138][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:13:14,642][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:13:15,150][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:13:15,658][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:13:16,163][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:13:16,672][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:13:17,177][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:13:17,683][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:13:18,188][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:13:18,693][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:13:19,198][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:13:19,705][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:13:20,209][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:13:20,712][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:13:21,216][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:13:21,719][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:13:22,223][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:13:22,728][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:13:23,232][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:13:23,735][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:13:24,239][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:13:24,744][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:13:25,247][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:13:25,752][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:13:26,257][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:13:26,759][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:13:27,264][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:13:27,766][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:13:28,271][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:13:28,773][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:13:29,278][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:13:29,783][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:13:30,288][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:13:30,793][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:13:31,297][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:13:31,802][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:13:32,308][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:13:32,813][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:13:33,318][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:13:33,825][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:13:34,331][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:13:34,838][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:13:35,343][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:13:35,849][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:13:36,358][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:13:36,863][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:13:37,369][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:13:37,875][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:13:38,380][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:13:38,887][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:13:39,392][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:13:39,898][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:13:40,402][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:13:40,905][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:13:41,410][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:13:41,914][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:13:42,419][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:13:42,923][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:13:43,430][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:13:43,936][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:13:44,438][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10866 tokens. +[2026-03-25 23:13:45,137][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 60.56%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 +[2026-03-25 23:13:45,887][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:13:45,889][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:13:45,892][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:13:46,620][__main__][INFO] - Iteration 148 took 50s (28.58% Gen, 69.97% Train). Generation: 14s, Training: 35s. Estimated remaining time: 39h 55m 7s. Estimated total time: 42h 3m 13s. Time estimates for 10 more iterations: 8m 24s, 100 more iterations: 1h 24m 6s, 500 more iterations: 7h 0m 32s. +[2026-03-25 23:13:46,623][__main__][INFO] - Starting iteration 148. +[2026-03-25 23:13:47,026][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 14 and human policies 1. +[2026-03-25 23:13:47,026][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:13:55,089][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 23:14:03,007][__main__][INFO] - Number of regex retries in iteration 148: 1 +[2026-03-25 23:14:03,008][__main__][INFO] - agents played in iteration 148 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:14:03,833][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:14:03,857][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:14:03,882][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:14:03,906][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:14:03,907][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:14:03,907][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:14:04,583][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:14:05,044][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:14:05,554][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:14:06,059][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:14:06,564][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:14:07,068][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:14:07,572][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:14:08,077][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:14:08,583][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:14:09,089][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:14:09,595][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:14:10,101][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:14:10,607][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:14:11,109][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:14:11,613][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:14:12,116][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:14:12,621][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:14:13,129][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:14:13,635][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:14:14,140][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:14:14,649][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:14:15,156][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:14:15,662][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:14:16,166][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:14:16,672][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:14:17,177][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:14:17,682][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:14:18,188][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:14:18,693][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:14:19,199][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:14:19,705][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:14:20,209][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:14:20,715][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:14:21,218][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:14:21,721][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:14:22,226][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:14:22,727][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:14:23,231][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:14:23,733][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:14:24,237][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:14:24,741][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:14:25,266][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:14:25,774][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:14:26,281][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:14:26,786][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:14:27,290][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:14:27,796][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:14:28,302][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:14:28,807][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:14:29,312][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:14:29,814][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:14:30,317][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:14:30,819][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:14:31,322][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:14:31,827][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:14:32,331][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:14:32,834][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:14:33,340][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:14:33,846][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:14:34,350][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:14:34,855][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:14:35,358][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:14:35,864][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:14:36,368][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:14:36,874][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10843 tokens. +[2026-03-25 23:14:37,571][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 60.63%, Block Peak % of device VRAM: 62.27%, ΔTime: 00:00:32 +[2026-03-25 23:14:38,324][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:14:38,326][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:14:38,328][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:14:39,061][__main__][INFO] - Iteration 149 took 52s (30.71% Gen, 67.88% Train). Generation: 15s, Training: 35s. Estimated remaining time: 41h 12m 50s. Estimated total time: 43h 21m 49s. Time estimates for 10 more iterations: 8m 40s, 100 more iterations: 1h 26m 43s, 500 more iterations: 7h 13m 38s. +[2026-03-25 23:14:39,064][__main__][INFO] - Starting iteration 149. +[2026-03-25 23:14:39,464][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 14 and human policies 1. +[2026-03-25 23:14:39,465][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:14:47,827][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls -Given the values: -- I value hats at 10 points each. -- Bob values hats at 1 point each. -- I value books at 10 points each. -- Bob values books at 10 points each. -- I value balls at 1 point each. -- Bob values balls at 10 points each. +I propose to take all 10 hats because my value for hats is 1, while Bob values them at 10. This would allow me to maximize my share of the less valuable items to me, given the high value Bob places on them. This way, I ensure I get the full quantity of an item I value more relatively, while potentially leaving high-value items for Bob to claim. did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 23:14:55,800][__main__][INFO] - Number of regex retries in iteration 149: 1 +[2026-03-25 23:14:55,801][__main__][INFO] - agents played in iteration 149 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:14:56,631][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:14:56,655][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:14:56,679][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:14:56,703][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:14:56,704][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:14:56,705][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:14:57,383][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:14:57,847][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:14:58,357][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:14:58,863][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:14:59,369][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:14:59,877][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:15:00,383][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:15:00,889][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:15:01,395][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:15:01,901][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:15:02,403][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:15:02,908][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:15:03,414][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:15:03,918][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:15:04,426][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:15:04,935][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:15:05,437][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:15:05,944][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:15:06,448][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:15:06,955][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:15:07,461][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:15:07,968][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:15:08,474][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:15:08,980][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:15:09,486][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:15:09,993][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:15:10,498][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:15:11,008][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:15:11,514][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:15:12,021][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:15:12,528][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:15:13,035][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:15:13,541][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:15:14,047][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:15:14,553][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:15:15,058][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:15:15,561][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:15:16,067][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:15:16,570][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:15:17,076][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:15:17,582][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:15:18,087][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:15:18,592][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:15:19,097][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:15:19,601][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:15:20,107][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:15:20,612][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:15:21,114][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:15:21,619][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:15:22,126][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:15:22,630][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:15:23,136][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:15:23,642][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:15:24,148][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:15:24,655][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:15:25,161][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:15:25,666][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:15:26,168][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:15:26,673][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:15:27,179][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:15:27,683][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:15:28,188][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:15:28,693][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:15:29,198][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:15:29,705][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10803 tokens. +[2026-03-25 23:15:30,407][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 60.66%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:33 +[2026-03-25 23:15:31,168][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:15:31,172][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:15:31,174][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:15:31,909][__main__][INFO] - Iteration 150 took 52s (31.15% Gen, 67.45% Train). Generation: 16s, Training: 35s. Estimated remaining time: 41h 32m 25s. Estimated total time: 43h 42m 17s. Time estimates for 10 more iterations: 8m 44s, 100 more iterations: 1h 27m 24s, 500 more iterations: 7h 17m 2s. +[2026-03-25 23:15:31,911][__main__][INFO] - Starting iteration 150. +[2026-03-25 23:15:32,313][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 14 and human policies 1. +[2026-03-25 23:15:32,314][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:15:35,931][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 23:15:39,414][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 23:15:47,989][__main__][INFO] - Number of regex retries in iteration 150: 2 +[2026-03-25 23:15:47,989][__main__][INFO] - agents played in iteration 150 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:15:48,811][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:15:48,835][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:15:48,860][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:15:48,884][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:15:48,885][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:15:48,885][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:15:49,572][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:15:50,034][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:15:50,544][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:15:51,050][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:15:51,555][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:15:52,060][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:15:52,566][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:15:53,072][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:15:53,575][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:15:54,081][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:15:54,585][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:15:55,089][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:15:55,595][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:15:56,099][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:15:56,603][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:15:57,109][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:15:57,614][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:15:58,120][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:15:58,626][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:15:59,132][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:15:59,638][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:16:00,144][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:16:00,650][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:16:01,156][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:16:01,661][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:16:02,168][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:16:02,673][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:16:03,178][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:16:03,683][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:16:04,189][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:16:04,693][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:16:05,196][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:16:05,701][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:16:06,203][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:16:06,704][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:16:07,208][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:16:07,710][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:16:08,212][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:16:08,711][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:16:09,215][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:16:09,717][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:16:10,223][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:16:10,728][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:16:11,232][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:16:11,738][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:16:12,243][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:16:12,750][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:16:13,255][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:16:13,759][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:16:14,269][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:16:14,775][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:16:15,282][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:16:15,787][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:16:16,293][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:16:16,800][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:16:17,306][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:16:17,812][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:16:18,316][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:16:18,819][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:16:19,325][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:16:19,828][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:16:20,333][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:16:20,838][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:16:21,338][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:16:21,841][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10811 tokens. +[2026-03-25 23:16:22,547][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:32 +[2026-03-25 23:16:23,303][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:16:23,305][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:16:23,310][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:16:24,718][__main__][INFO] - Iteration 151 took 52s (29.91% Gen, 67.40% Train). Generation: 15s, Training: 35s. Estimated remaining time: 41h 29m 32s. Estimated total time: 43h 40m 16s. Time estimates for 10 more iterations: 8m 44s, 100 more iterations: 1h 27m 20s, 500 more iterations: 7h 16m 42s. +[2026-03-25 23:16:24,720][__main__][INFO] - Starting iteration 151. +[2026-03-25 23:16:25,125][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 15 and human policies 1. +[2026-03-25 23:16:25,126][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:16:40,175][__main__][INFO] - Number of regex retries in iteration 151: 0 +[2026-03-25 23:16:40,176][__main__][INFO] - agents played in iteration 151 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:16:41,007][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:16:41,031][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:16:41,055][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:16:41,079][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:16:41,080][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:16:41,081][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:16:41,762][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:16:42,223][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:16:42,734][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:16:43,236][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:16:43,742][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:16:44,248][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:16:44,753][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:16:45,258][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:16:45,763][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:16:46,269][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:16:46,774][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:16:47,279][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:16:47,784][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:16:48,289][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:16:48,793][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:16:49,298][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:16:49,803][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:16:50,308][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:16:50,811][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:16:51,314][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:16:51,818][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:16:52,321][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:16:52,826][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:16:53,331][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:16:53,835][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:16:54,344][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:16:54,869][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:16:55,378][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:16:55,885][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:16:56,391][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:16:56,896][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:16:57,402][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:16:57,909][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:16:58,410][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:16:58,915][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:16:59,418][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:16:59,922][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:17:00,428][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:17:00,933][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:17:01,437][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:17:01,939][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:17:02,445][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:17:02,951][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:17:03,457][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:17:03,961][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:17:04,470][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:17:04,976][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:17:05,482][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:17:05,988][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:17:06,492][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:17:06,997][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:17:07,502][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:17:08,007][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:17:08,510][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:17:09,015][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:17:09,519][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:17:10,023][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:17:10,528][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:17:11,033][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:17:11,539][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:17:12,045][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:17:12,550][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:17:13,055][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:17:13,560][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:17:14,065][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10841 tokens. +[2026-03-25 23:17:14,772][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 60.66%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:33 +[2026-03-25 23:17:15,525][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:17:15,528][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:17:15,529][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:17:16,205][__main__][INFO] - Iteration 152 took 51s (29.46% Gen, 69.21% Train). Generation: 15s, Training: 35s. Estimated remaining time: 40h 22m 26s. Estimated total time: 42h 34m 2s. Time estimates for 10 more iterations: 8m 30s, 100 more iterations: 1h 25m 8s, 500 more iterations: 7h 5m 40s. +[2026-03-25 23:17:16,207][__main__][INFO] - Starting iteration 152. +[2026-03-25 23:17:16,611][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 15 and human policies 1. +[2026-03-25 23:17:16,612][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:17:33,503][__main__][INFO] - Number of regex retries in iteration 152: 0 +[2026-03-25 23:17:33,504][__main__][INFO] - agents played in iteration 152 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:17:34,334][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:17:34,358][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:17:34,382][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:17:34,406][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:17:34,407][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:17:34,408][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:17:35,090][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:17:35,552][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:17:36,060][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:17:36,567][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:17:37,070][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:17:37,575][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:17:38,078][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:17:38,580][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:17:39,083][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:17:39,592][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:17:40,097][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:17:40,605][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:17:41,110][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:17:41,616][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:17:42,119][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:17:42,625][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:17:43,131][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:17:43,655][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:17:44,165][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:17:44,670][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:17:45,178][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:17:45,684][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:17:46,191][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:17:46,697][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:17:47,202][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:17:47,708][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:17:48,213][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:17:48,719][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:17:49,225][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:17:49,731][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:17:50,236][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:17:50,741][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:17:51,248][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:17:51,754][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:17:52,259][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:17:52,765][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:17:53,270][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:17:53,775][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:17:54,280][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:17:54,784][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:17:55,291][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:17:55,794][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:17:56,296][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:17:56,803][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:17:57,310][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:17:57,814][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:17:58,319][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:17:58,822][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:17:59,326][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:17:59,830][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:18:00,336][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:18:00,842][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:18:01,347][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:18:01,853][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:18:02,360][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:18:02,867][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:18:03,373][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:18:03,878][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:18:04,382][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:18:04,888][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:18:05,394][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:18:05,899][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:18:06,403][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:18:06,909][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:18:07,414][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10769 tokens. +[2026-03-25 23:18:08,119][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 60.63%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:33 +[2026-03-25 23:18:08,873][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:18:08,875][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:18:08,877][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:18:09,554][__main__][INFO] - Iteration 153 took 52s (31.91% Gen, 66.81% Train). Generation: 16s, Training: 35s. Estimated remaining time: 41h 54m 42s. Estimated total time: 44h 7m 11s. Time estimates for 10 more iterations: 8m 49s, 100 more iterations: 1h 28m 14s, 500 more iterations: 7h 21m 11s. +[2026-03-25 23:18:09,557][__main__][INFO] - Starting iteration 153. +[2026-03-25 23:18:09,957][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 15 and human policies 1. +[2026-03-25 23:18:09,958][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:18:25,419][__main__][INFO] - Number of regex retries in iteration 153: 0 +[2026-03-25 23:18:25,420][__main__][INFO] - agents played in iteration 153 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:18:26,248][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:18:26,273][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:18:26,297][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:18:26,321][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:18:26,322][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:18:26,323][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:18:27,004][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:18:27,465][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:18:27,979][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:18:28,485][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:18:28,992][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:18:29,499][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:18:30,005][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:18:30,511][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:18:31,016][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:18:31,521][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:18:32,027][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:18:32,532][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:18:33,039][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:18:33,543][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:18:34,049][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:18:34,553][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:18:35,059][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:18:35,563][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:18:36,067][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:18:36,572][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:18:37,077][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:18:37,581][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:18:38,085][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:18:38,586][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:18:39,089][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:18:39,592][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:18:40,097][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:18:40,601][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:18:41,106][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:18:41,611][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:18:42,115][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:18:42,619][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:18:43,123][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:18:43,628][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:18:44,133][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:18:44,638][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:18:45,145][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:18:45,651][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:18:46,157][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:18:46,663][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:18:47,170][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:18:47,676][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:18:48,183][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:18:48,689][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:18:49,196][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:18:49,701][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:18:50,208][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:18:50,714][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:18:51,220][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:18:51,725][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:18:52,230][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:18:52,734][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:18:53,239][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:18:53,743][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:18:54,250][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:18:54,755][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:18:55,260][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:18:55,764][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:18:56,269][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:18:56,775][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:18:57,280][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:18:57,785][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:18:58,291][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:18:58,797][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:18:59,301][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10802 tokens. +[2026-03-25 23:19:00,008][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 60.64%, Block Peak % of device VRAM: 62.34%, ΔTime: 00:00:33 +[2026-03-25 23:19:00,765][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:19:00,767][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:19:00,769][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:19:01,669][__main__][INFO] - Iteration 154 took 51s (29.90% Gen, 68.36% Train). Generation: 15s, Training: 35s. Estimated remaining time: 40h 52m 16s. Estimated total time: 43h 5m 37s. Time estimates for 10 more iterations: 8m 37s, 100 more iterations: 1h 26m 11s, 500 more iterations: 7h 10m 56s. +[2026-03-25 23:19:01,672][__main__][INFO] - Starting iteration 154. +[2026-03-25 23:19:02,075][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 15 and human policies 1. +[2026-03-25 23:19:02,076][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:19:13,825][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls -To maximize my points for this round: -- I should take all the hats since they are worth significantly more to me than to Bob. -- I should not take any books because although I value books, Bob values them much higher, and we would end up splitting them proportionally, which wouldn't maximize my points. -- I should not take any balls either for similar reasons to the books. +Given the per-item values, I aim to maximize the number of hats I receive since they provide the highest value for me. Bob values the hats very low, so he is likely to propose keeping 0 hats. By proposing to take all the hats, I ensure I get the full value from them, even if the total proposals exceed the available items and are allocated proportionally. This leaves the books and balls to be split between hats, which I propose to take none of. This strategy leverages my high value for hats and minimizes the risk of splitting the remaining items with Bob. did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 23:19:16,213][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 1 y books, 10 z balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 23:19:20,142][__main__][INFO] - Number of regex retries in iteration 154: 2 +[2026-03-25 23:19:20,143][__main__][INFO] - agents played in iteration 154 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:19:20,981][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:19:21,005][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:19:21,029][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:19:21,053][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:19:21,054][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:19:21,055][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:19:21,737][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:19:22,199][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:19:22,705][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:19:23,209][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:19:23,713][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:19:24,217][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:19:24,721][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:19:25,222][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:19:25,732][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:19:26,242][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:19:26,749][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:19:27,255][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:19:27,766][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:19:28,274][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:19:28,781][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:19:29,288][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:19:29,795][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:19:30,300][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:19:30,804][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:19:31,306][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:19:31,811][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:19:32,314][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:19:32,817][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:19:33,320][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:19:33,825][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:19:34,330][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:19:34,837][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:19:35,344][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:19:35,852][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:19:36,359][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:19:36,864][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:19:37,370][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:19:37,876][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:19:38,379][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:19:38,882][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:19:39,389][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:19:39,894][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:19:40,396][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:19:40,900][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:19:41,401][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:19:41,904][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:19:42,410][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:19:42,915][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:19:43,418][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:19:43,923][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:19:44,428][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:19:44,934][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:19:45,438][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:19:45,942][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:19:46,449][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:19:46,955][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:19:47,460][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:19:47,965][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:19:48,470][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:19:48,975][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:19:49,481][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:19:49,983][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:19:50,490][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:19:50,996][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:19:51,501][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:19:52,007][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:19:52,513][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:19:53,015][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:19:53,520][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:19:54,026][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10737 tokens. +[2026-03-25 23:19:54,749][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.38%, ΔTime: 00:00:33 +[2026-03-25 23:19:55,516][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:19:55,519][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:19:55,523][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:19:56,206][__main__][INFO] - Iteration 155 took 54s (33.38% Gen, 65.36% Train). Generation: 18s, Training: 35s. Estimated remaining time: 42h 52m 18s. Estimated total time: 45h 6m 33s. Time estimates for 10 more iterations: 9m 1s, 100 more iterations: 1h 30m 13s, 500 more iterations: 7h 31m 5s. +[2026-03-25 23:19:56,210][__main__][INFO] - Starting iteration 155. +[2026-03-25 23:19:56,615][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 15 and human policies 1. +[2026-03-25 23:19:56,616][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:20:14,572][__main__][INFO] - Number of regex retries in iteration 155: 0 +[2026-03-25 23:20:14,573][__main__][INFO] - agents played in iteration 155 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:20:15,391][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:20:15,415][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:20:15,439][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:20:15,463][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:20:15,464][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:20:15,465][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:20:16,144][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:20:16,601][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:20:17,114][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:20:17,621][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:20:18,125][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:20:18,633][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:20:19,139][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:20:19,644][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:20:20,151][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:20:20,661][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:20:21,187][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:20:21,697][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:20:22,203][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:20:22,713][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:20:23,220][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:20:23,727][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:20:24,234][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:20:24,740][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:20:25,247][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:20:25,754][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:20:26,260][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:20:26,767][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:20:27,272][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:20:27,776][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:20:28,283][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:20:28,788][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:20:29,293][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:20:29,796][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:20:30,301][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:20:30,806][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:20:31,312][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:20:31,817][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:20:32,320][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:20:32,826][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:20:33,332][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:20:33,837][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:20:34,343][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:20:34,849][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:20:35,355][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:20:35,861][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:20:36,368][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:20:36,874][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:20:37,380][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:20:37,884][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:20:38,391][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:20:38,898][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:20:39,403][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:20:39,909][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:20:40,415][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:20:40,924][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:20:41,434][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:20:41,940][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:20:42,447][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:20:42,955][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:20:43,463][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:20:43,971][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:20:44,479][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:20:44,987][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:20:45,492][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:20:45,998][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:20:46,504][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:20:47,011][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:20:47,516][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:20:48,021][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:20:48,528][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10812 tokens. +[2026-03-25 23:20:49,232][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.27%, Current % of VRAM taken: 60.69%, Block Peak % of device VRAM: 62.58%, ΔTime: 00:00:33 +[2026-03-25 23:20:49,991][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:20:49,993][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:20:49,995][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:20:50,673][__main__][INFO] - Iteration 156 took 54s (33.22% Gen, 65.53% Train). Generation: 17s, Training: 35s. Estimated remaining time: 42h 47m 43s. Estimated total time: 45h 2m 54s. Time estimates for 10 more iterations: 9m 0s, 100 more iterations: 1h 30m 5s, 500 more iterations: 7h 30m 29s. +[2026-03-25 23:20:50,675][__main__][INFO] - Starting iteration 156. +[2026-03-25 23:20:51,080][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 15 and human policies 1. +[2026-03-25 23:20:51,080][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:20:58,423][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 23:21:11,619][__main__][INFO] - Number of regex retries in iteration 156: 1 +[2026-03-25 23:21:11,620][__main__][INFO] - agents played in iteration 156 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:21:12,474][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:21:12,499][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:21:12,523][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:21:12,548][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:21:12,549][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:21:12,549][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:21:13,240][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:21:13,699][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:21:14,207][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:21:14,710][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:21:15,214][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:21:15,722][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:21:16,225][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:21:16,729][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:21:17,232][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:21:17,738][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:21:18,246][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:21:18,752][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:21:19,258][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:21:19,764][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:21:20,268][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:21:20,775][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:21:21,281][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:21:21,807][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:21:22,314][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:21:22,821][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:21:23,324][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:21:23,828][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:21:24,335][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:21:24,841][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:21:25,348][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:21:25,854][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:21:26,362][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:21:26,867][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:21:27,373][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:21:27,879][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:21:28,384][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:21:28,889][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:21:29,395][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:21:29,901][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:21:30,408][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:21:30,913][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:21:31,419][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:21:31,926][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:21:32,433][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:21:32,941][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:21:33,447][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:21:33,953][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:21:34,455][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:21:34,962][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:21:35,467][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:21:35,971][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:21:36,477][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:21:36,981][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:21:37,485][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:21:37,991][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:21:38,496][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:21:39,001][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:21:39,507][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:21:40,012][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:21:40,518][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:21:41,023][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:21:41,531][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:21:42,039][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:21:42,544][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:21:43,051][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:21:43,556][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:21:44,061][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:21:44,565][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:21:45,070][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:21:45,575][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10731 tokens. +[2026-03-25 23:21:46,291][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 60.60%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:33 +[2026-03-25 23:21:47,077][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:21:47,080][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:21:47,082][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:21:47,948][__main__][INFO] - Iteration 157 took 56s (36.12% Gen, 62.36% Train). Generation: 20s, Training: 35s. Estimated remaining time: 45h 7m 20s. Estimated total time: 47h 23m 28s. Time estimates for 10 more iterations: 9m 28s, 100 more iterations: 1h 34m 46s, 500 more iterations: 7h 53m 54s. +[2026-03-25 23:21:47,951][__main__][INFO] - Starting iteration 157. +[2026-03-25 23:21:48,351][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 15 and human policies 1. +[2026-03-25 23:21:48,352][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:22:04,249][__main__][INFO] - Number of regex retries in iteration 157: 0 +[2026-03-25 23:22:04,250][__main__][INFO] - agents played in iteration 157 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:22:05,295][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:22:05,319][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:22:05,344][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:22:05,368][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:22:05,369][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:22:05,369][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:22:06,063][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:22:06,526][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:22:07,035][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:22:07,543][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:22:08,049][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:22:08,555][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:22:09,063][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:22:09,568][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:22:10,075][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:22:10,580][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:22:11,088][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:22:11,593][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:22:12,099][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:22:12,605][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:22:13,111][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:22:13,616][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:22:14,122][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:22:14,629][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:22:15,135][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:22:15,644][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:22:16,150][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:22:16,656][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:22:17,162][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:22:17,669][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:22:18,177][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:22:18,687][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:22:19,196][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:22:19,703][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:22:20,212][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:22:20,720][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:22:21,230][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:22:21,738][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:22:22,246][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:22:22,752][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:22:23,258][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:22:23,767][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:22:24,274][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:22:24,780][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:22:25,289][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:22:25,795][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:22:26,303][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:22:26,810][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:22:27,316][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:22:27,822][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:22:28,328][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:22:28,832][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:22:29,339][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:22:29,845][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:22:30,351][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:22:30,856][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:22:31,362][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:22:31,864][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:22:32,371][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:22:32,877][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:22:33,382][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:22:33,887][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:22:34,393][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:22:34,898][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:22:35,405][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:22:35,910][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:22:36,417][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:22:36,923][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:22:37,429][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:22:37,935][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:22:38,441][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10820 tokens. +[2026-03-25 23:22:39,174][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 60.66%, Block Peak % of device VRAM: 62.41%, ΔTime: 00:00:33 +[2026-03-25 23:22:40,001][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:22:40,004][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:22:40,006][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:22:40,741][__main__][INFO] - Iteration 158 took 52s (30.34% Gen, 68.25% Train). Generation: 15s, Training: 35s. Estimated remaining time: 41h 22m 31s. Estimated total time: 43h 39m 32s. Time estimates for 10 more iterations: 8m 43s, 100 more iterations: 1h 27m 19s, 500 more iterations: 7h 16m 35s. +[2026-03-25 23:22:40,744][__main__][INFO] - Starting iteration 158. +[2026-03-25 23:22:41,153][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 15 and human policies 1. +[2026-03-25 23:22:41,153][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:22:53,469][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 23:22:58,734][__main__][INFO] - Number of regex retries in iteration 158: 1 +[2026-03-25 23:22:58,735][__main__][INFO] - agents played in iteration 158 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:22:59,636][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:22:59,661][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:22:59,686][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:22:59,711][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:22:59,712][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:22:59,712][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:23:00,419][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:23:00,882][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:23:01,394][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:23:01,901][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:23:02,410][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:23:02,916][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:23:03,422][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:23:03,931][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:23:04,440][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:23:04,946][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:23:05,454][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:23:05,962][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:23:06,469][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:23:06,976][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:23:07,482][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:23:07,991][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:23:08,499][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:23:09,002][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:23:09,509][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:23:10,014][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:23:10,521][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:23:11,025][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:23:11,533][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:23:12,036][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:23:12,542][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:23:13,050][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:23:13,556][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:23:14,064][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:23:14,572][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:23:15,078][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:23:15,586][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:23:16,094][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:23:16,602][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:23:17,107][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:23:17,613][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:23:18,117][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:23:18,621][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:23:19,124][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:23:19,630][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:23:20,135][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:23:20,640][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:23:21,146][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:23:21,651][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:23:22,159][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:23:22,663][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:23:23,168][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:23:23,673][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:23:24,180][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:23:24,686][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:23:25,191][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:23:25,697][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:23:26,203][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:23:26,709][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:23:27,215][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:23:27,721][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:23:28,229][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:23:28,736][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:23:29,242][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:23:29,752][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:23:30,260][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:23:30,765][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:23:31,271][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:23:31,778][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:23:32,283][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:23:32,790][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10762 tokens. +[2026-03-25 23:23:33,529][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 60.66%, Block Peak % of device VRAM: 62.27%, ΔTime: 00:00:33 +[2026-03-25 23:23:34,312][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:23:34,315][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:23:34,317][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:23:35,082][__main__][INFO] - Iteration 159 took 53s (32.60% Gen, 65.98% Train). Generation: 17s, Training: 35s. Estimated remaining time: 42h 38m 34s. Estimated total time: 44h 56m 29s. Time estimates for 10 more iterations: 8m 59s, 100 more iterations: 1h 29m 52s, 500 more iterations: 7h 29m 24s. +[2026-03-25 23:23:35,084][__main__][INFO] - Starting iteration 159. +[2026-03-25 23:23:35,487][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 15 and human policies 1. +[2026-03-25 23:23:35,487][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:23:51,668][__main__][INFO] - Number of regex retries in iteration 159: 0 +[2026-03-25 23:23:51,670][__main__][INFO] - agents played in iteration 159 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:23:52,543][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:23:52,568][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:23:52,593][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:23:52,618][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:23:52,619][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:23:52,620][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:23:53,478][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:23:53,962][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:23:54,482][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:23:54,997][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:23:55,508][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:23:56,045][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:23:56,582][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:23:57,096][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:23:57,611][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:23:58,125][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:23:58,636][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:23:59,145][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:23:59,658][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:24:00,169][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:24:00,680][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:24:01,192][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:24:01,702][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:24:02,218][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:24:02,729][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:24:03,237][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:24:03,748][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:24:04,261][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:24:04,773][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:24:05,286][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:24:05,799][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:24:06,312][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:24:06,823][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:24:07,335][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:24:07,846][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:24:08,357][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:24:08,866][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:24:09,379][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:24:09,893][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:24:10,405][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:24:10,917][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:24:11,431][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:24:11,943][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:24:12,455][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:24:12,967][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:24:13,477][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:24:13,989][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:24:14,500][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:24:15,010][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:24:15,520][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:24:16,032][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:24:16,544][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:24:17,051][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:24:17,565][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:24:18,077][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:24:18,588][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:24:19,096][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:24:19,607][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:24:20,117][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:24:20,627][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:24:21,138][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:24:21,651][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:24:22,163][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:24:22,675][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:24:23,186][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:24:23,694][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:24:24,205][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:24:24,718][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:24:25,226][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:24:25,738][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:24:26,250][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10778 tokens. +[2026-03-25 23:24:27,108][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 60.63%, Block Peak % of device VRAM: 62.28%, ΔTime: 00:00:33 +[2026-03-25 23:24:28,178][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:24:28,181][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:24:28,183][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:24:29,212][__main__][INFO] - Iteration 160 took 53s (30.12% Gen, 67.96% Train). Generation: 16s, Training: 36s. Estimated remaining time: 42h 27m 28s. Estimated total time: 44h 46m 17s. Time estimates for 10 more iterations: 8m 57s, 100 more iterations: 1h 29m 32s, 500 more iterations: 7h 27m 42s. +[2026-03-25 23:24:29,215][__main__][INFO] - Starting iteration 160. +[2026-03-25 23:24:29,619][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 15 and human policies 1. +[2026-03-25 23:24:29,620][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:24:46,471][__main__][INFO] - Number of regex retries in iteration 160: 0 +[2026-03-25 23:24:46,472][__main__][INFO] - agents played in iteration 160 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:24:47,318][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:24:47,345][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:24:47,373][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:24:47,400][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:24:47,401][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:24:47,402][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:24:48,233][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:24:48,708][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:24:49,221][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:24:49,736][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:24:50,249][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:24:50,791][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:24:51,312][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:24:51,826][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:24:52,342][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:24:52,855][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:24:53,367][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:24:53,881][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:24:54,395][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:24:54,907][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:24:55,420][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:24:55,930][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:24:56,443][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:24:56,958][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:24:57,469][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:24:57,979][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:24:58,488][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:24:59,003][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:24:59,515][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:25:00,029][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:25:00,540][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:25:01,073][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:25:01,591][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:25:02,105][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:25:02,621][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:25:03,134][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:25:03,647][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:25:04,160][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:25:04,672][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:25:05,182][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:25:05,692][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:25:06,202][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:25:06,713][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:25:07,224][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:25:07,735][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:25:08,244][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:25:08,757][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:25:09,268][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:25:09,777][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:25:10,290][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:25:10,801][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:25:11,312][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:25:11,822][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:25:12,333][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:25:12,843][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:25:13,357][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:25:13,868][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:25:14,380][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:25:14,889][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:25:15,399][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:25:15,913][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:25:16,424][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:25:16,935][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:25:17,444][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:25:17,956][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:25:18,467][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:25:18,997][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:25:19,511][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:25:20,022][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:25:20,532][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:25:21,044][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10732 tokens. +[2026-03-25 23:25:21,860][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 60.66%, Block Peak % of device VRAM: 62.28%, ΔTime: 00:00:33 +[2026-03-25 23:25:22,748][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:25:22,752][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:25:22,753][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:25:24,733][__main__][INFO] - Iteration 161 took 55s (30.58% Gen, 65.83% Train). Generation: 16s, Training: 36s. Estimated remaining time: 43h 35m 59s. Estimated total time: 45h 55m 43s. Time estimates for 10 more iterations: 9m 11s, 100 more iterations: 1h 31m 51s, 500 more iterations: 7h 39m 17s. +[2026-03-25 23:25:24,737][__main__][INFO] - Starting iteration 161. +[2026-03-25 23:25:25,138][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 16 and human policies 1. +[2026-03-25 23:25:25,139][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:25:33,239][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 1 y book, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 23:25:41,568][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 23:25:42,443][__main__][INFO] - Number of regex retries in iteration 161: 2 +[2026-03-25 23:25:42,445][__main__][INFO] - agents played in iteration 161 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:25:43,334][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:25:43,361][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:25:43,388][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:25:43,414][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:25:43,415][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:25:43,416][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:25:44,230][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:25:44,698][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:25:45,220][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:25:45,731][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:25:46,243][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:25:46,765][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:25:47,277][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:25:47,788][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:25:48,302][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:25:48,817][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:25:49,330][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:25:49,851][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:25:50,357][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:25:50,870][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:25:51,380][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:25:51,900][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:25:52,409][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:25:52,921][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:25:53,432][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:25:53,950][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:25:54,461][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:25:54,972][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:25:55,485][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:25:56,004][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:25:56,516][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:25:57,032][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:25:57,540][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:25:58,051][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:25:58,588][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:25:59,098][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:25:59,607][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:26:00,117][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:26:00,629][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:26:01,140][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:26:01,656][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:26:02,171][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:26:02,681][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:26:03,194][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:26:03,707][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:26:04,218][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:26:04,732][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:26:05,268][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:26:05,782][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:26:06,295][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:26:06,810][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:26:07,324][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:26:07,840][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:26:08,356][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:26:08,872][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:26:09,384][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:26:09,897][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:26:10,413][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:26:10,926][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:26:11,444][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:26:11,954][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:26:12,466][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:26:12,979][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:26:13,491][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:26:14,002][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:26:14,511][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:26:15,056][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:26:15,565][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:26:16,095][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:26:16,609][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:26:17,117][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10776 tokens. +[2026-03-25 23:26:18,077][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 60.57%, Block Peak % of device VRAM: 62.40%, ΔTime: 00:00:33 +[2026-03-25 23:26:18,965][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:26:18,968][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:26:18,970][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:26:19,902][__main__][INFO] - Iteration 162 took 54s (31.60% Gen, 66.70% Train). Generation: 17s, Training: 36s. Estimated remaining time: 43h 17m 33s. Estimated total time: 45h 38m 12s. Time estimates for 10 more iterations: 9m 7s, 100 more iterations: 1h 31m 16s, 500 more iterations: 7h 36m 22s. +[2026-03-25 23:26:19,905][__main__][INFO] - Starting iteration 162. +[2026-03-25 23:26:20,307][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 16 and human policies 1. +[2026-03-25 23:26:20,308][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:26:37,219][__main__][INFO] - Number of regex retries in iteration 162: 0 +[2026-03-25 23:26:37,221][__main__][INFO] - agents played in iteration 162 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:26:38,118][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:26:38,145][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:26:38,172][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:26:38,199][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:26:38,200][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:26:38,201][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:26:39,065][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:26:39,537][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:26:40,051][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:26:40,566][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:26:41,080][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:26:41,589][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:26:42,100][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:26:42,619][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:26:43,151][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:26:43,675][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:26:44,188][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:26:44,706][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:26:45,240][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:26:45,736][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:26:46,249][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:26:46,759][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:26:47,270][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:26:47,780][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:26:48,293][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:26:48,802][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:26:49,311][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:26:49,820][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:26:50,329][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:26:50,847][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:26:51,369][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:26:51,881][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:26:52,386][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:26:52,894][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:26:53,403][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:26:53,911][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:26:54,418][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:26:54,926][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:26:55,432][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:26:55,947][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:26:56,454][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:26:56,963][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:26:57,476][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:26:57,989][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:26:58,516][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:26:59,025][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:26:59,535][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:27:00,046][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:27:00,563][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:27:01,072][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:27:01,581][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:27:02,088][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:27:02,597][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:27:03,104][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:27:03,634][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:27:04,141][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:27:04,660][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:27:05,164][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:27:05,668][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:27:06,173][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:27:06,678][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:27:07,183][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:27:07,688][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:27:08,190][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:27:08,697][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:27:09,201][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:27:09,702][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:27:10,205][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:27:10,708][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:27:11,211][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:27:11,713][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10780 tokens. +[2026-03-25 23:27:12,383][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.41%, ΔTime: 00:00:33 +[2026-03-25 23:27:13,140][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:27:13,142][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:27:13,144][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:27:13,925][__main__][INFO] - Iteration 163 took 53s (31.54% Gen, 67.00% Train). Generation: 16s, Training: 35s. Estimated remaining time: 42h 19m 22s. Estimated total time: 44h 40m 55s. Time estimates for 10 more iterations: 8m 56s, 100 more iterations: 1h 29m 21s, 500 more iterations: 7h 26m 49s. +[2026-03-25 23:27:13,929][__main__][INFO] - Starting iteration 163. +[2026-03-25 23:27:14,334][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 16 and human policies 1. +[2026-03-25 23:27:14,335][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:27:26,700][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 23:27:30,235][__main__][INFO] - Number of regex retries in iteration 163: 1 +[2026-03-25 23:27:30,236][__main__][INFO] - agents played in iteration 163 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:27:31,010][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:27:31,034][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:27:31,057][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:27:31,080][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:27:31,081][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:27:31,082][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:27:31,727][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:27:32,185][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:27:32,694][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:27:33,195][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:27:33,697][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:27:34,198][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:27:34,699][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:27:35,199][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:27:35,702][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:27:36,208][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:27:36,715][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:27:37,218][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:27:37,718][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:27:38,219][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:27:38,736][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:27:39,242][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:27:39,764][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:27:40,263][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:27:40,770][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:27:41,270][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:27:41,770][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:27:42,268][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:27:42,766][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:27:43,264][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:27:43,762][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:27:44,260][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:27:44,760][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:27:45,261][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:27:45,761][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:27:46,259][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:27:46,759][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:27:47,258][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:27:47,759][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:27:48,259][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:27:48,761][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:27:49,261][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:27:49,760][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:27:50,259][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:27:50,758][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:27:51,259][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:27:51,758][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:27:52,260][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:27:52,761][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:27:53,263][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:27:53,761][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:27:54,261][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:27:54,761][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:27:55,262][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:27:55,763][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:27:56,264][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:27:56,763][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:27:57,264][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:27:57,763][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:27:58,262][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:27:58,763][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:27:59,261][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:27:59,772][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:28:00,272][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:28:00,771][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:28:01,271][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:28:01,771][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:28:02,270][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:28:02,767][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:28:03,267][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:28:03,766][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10762 tokens. +[2026-03-25 23:28:04,380][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 60.60%, Block Peak % of device VRAM: 62.39%, ΔTime: 00:00:32 +[2026-03-25 23:28:05,116][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:28:05,118][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:28:05,119][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:28:05,951][__main__][INFO] - Iteration 164 took 51s (30.81% Gen, 67.58% Train). Generation: 15s, Training: 34s. Estimated remaining time: 40h 38m 26s. Estimated total time: 43h 0m 51s. Time estimates for 10 more iterations: 8m 36s, 100 more iterations: 1h 26m 1s, 500 more iterations: 7h 10m 8s. +[2026-03-25 23:28:05,953][__main__][INFO] - Starting iteration 164. +[2026-03-25 23:28:06,353][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 16 and human policies 1. +[2026-03-25 23:28:06,354][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:28:12,151][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 23:28:15,222][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 23:28:21,792][__main__][INFO] - Number of regex retries in iteration 164: 2 +[2026-03-25 23:28:21,793][__main__][INFO] - agents played in iteration 164 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:28:22,574][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:28:22,597][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:28:22,621][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:28:22,645][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:28:22,646][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:28:22,646][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:28:23,363][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:28:23,826][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:28:24,334][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:28:24,842][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:28:25,354][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:28:25,857][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:28:26,373][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:28:26,875][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:28:27,376][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:28:27,882][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:28:28,381][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:28:28,899][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:28:29,385][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:28:29,903][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:28:30,403][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:28:30,903][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:28:31,402][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:28:31,901][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:28:32,401][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:28:32,900][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:28:33,397][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:28:33,895][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:28:34,395][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:28:34,894][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:28:35,392][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:28:35,911][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:28:36,410][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:28:36,914][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:28:37,415][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:28:37,915][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:28:38,415][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:28:38,914][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:28:39,413][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:28:39,916][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:28:40,418][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:28:40,939][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:28:41,439][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:28:41,940][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:28:42,441][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:28:42,941][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:28:43,443][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:28:43,943][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:28:44,442][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:28:44,942][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:28:45,442][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:28:45,940][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:28:46,449][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:28:46,942][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:28:47,447][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:28:47,947][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:28:48,450][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:28:48,949][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:28:49,449][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:28:49,951][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:28:50,449][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:28:50,945][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:28:51,443][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:28:51,943][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:28:52,442][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:28:52,940][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:28:53,440][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:28:53,938][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:28:54,437][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:28:54,936][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:28:55,455][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10769 tokens. +[2026-03-25 23:28:56,079][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.25%, Current % of VRAM taken: 60.68%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:32 +[2026-03-25 23:28:56,830][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:28:56,832][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:28:56,834][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:28:57,587][__main__][INFO] - Iteration 165 took 51s (30.13% Gen, 68.39% Train). Generation: 15s, Training: 35s. Estimated remaining time: 40h 18m 25s. Estimated total time: 42h 41m 42s. Time estimates for 10 more iterations: 8m 32s, 100 more iterations: 1h 25m 23s, 500 more iterations: 7h 6m 57s. +[2026-03-25 23:28:57,589][__main__][INFO] - Starting iteration 165. +[2026-03-25 23:28:57,989][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 16 and human policies 1. +[2026-03-25 23:28:57,990][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:29:14,429][__main__][INFO] - Number of regex retries in iteration 165: 0 +[2026-03-25 23:29:14,430][__main__][INFO] - agents played in iteration 165 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:29:15,201][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:29:15,225][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:29:15,249][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:29:15,272][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:29:15,273][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:29:15,274][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:29:16,122][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:29:16,588][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:29:17,095][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:29:17,613][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:29:18,124][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:29:18,642][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:29:19,152][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:29:19,669][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:29:20,179][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:29:20,689][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:29:21,194][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:29:21,698][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:29:22,200][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:29:22,703][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:29:23,206][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:29:23,711][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:29:24,222][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:29:24,735][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:29:25,242][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:29:25,746][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:29:26,258][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:29:26,778][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:29:27,283][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:29:27,799][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:29:28,308][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:29:28,824][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:29:29,331][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:29:29,836][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:29:30,338][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:29:30,838][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:29:31,338][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:29:31,838][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:29:32,339][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:29:32,845][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:29:33,356][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:29:33,863][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:29:34,373][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:29:34,876][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:29:35,376][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:29:35,895][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:29:36,406][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:29:36,926][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:29:37,436][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:29:37,952][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:29:38,465][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:29:38,970][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:29:39,472][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:29:39,972][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:29:40,473][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:29:40,974][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:29:41,476][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:29:41,981][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:29:42,484][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:29:42,998][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:29:43,510][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:29:44,015][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:29:44,518][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:29:45,033][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:29:45,543][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:29:46,062][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:29:46,571][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:29:47,092][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:29:47,600][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:29:48,102][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:29:48,601][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10759 tokens. +[2026-03-25 23:29:49,223][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 60.50%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:33 +[2026-03-25 23:29:49,969][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:29:49,971][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:29:49,972][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:29:50,733][__main__][INFO] - Iteration 166 took 52s (31.17% Gen, 67.39% Train). Generation: 16s, Training: 35s. Estimated remaining time: 41h 33m 2s. Estimated total time: 43h 57m 13s. Time estimates for 10 more iterations: 8m 47s, 100 more iterations: 1h 27m 54s, 500 more iterations: 7h 19m 32s. +[2026-03-25 23:29:50,735][__main__][INFO] - Starting iteration 166. +[2026-03-25 23:29:51,138][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 16 and human policies 1. +[2026-03-25 23:29:51,138][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:30:07,956][__main__][INFO] - Number of regex retries in iteration 166: 0 +[2026-03-25 23:30:07,957][__main__][INFO] - agents played in iteration 166 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:30:08,770][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:30:08,794][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:30:08,818][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:30:08,842][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:30:08,842][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:30:08,843][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:30:09,437][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:30:09,895][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:30:10,407][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:30:10,915][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:30:11,426][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:30:11,930][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:30:12,432][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:30:12,951][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:30:13,465][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:30:13,981][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:30:14,493][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:30:15,011][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:30:15,521][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:30:16,025][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:30:16,529][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:30:17,032][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:30:17,533][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:30:18,030][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:30:18,534][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:30:19,038][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:30:19,545][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:30:20,055][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:30:20,564][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:30:21,072][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:30:21,574][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:30:22,078][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:30:22,596][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:30:23,102][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:30:23,620][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:30:24,132][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:30:24,650][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:30:25,159][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:30:25,661][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:30:26,159][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:30:26,661][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:30:27,161][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:30:27,662][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:30:28,162][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:30:28,668][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:30:29,178][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:30:29,688][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:30:30,201][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:30:30,706][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:30:31,208][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:30:31,728][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:30:32,238][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:30:32,760][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:30:33,266][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:30:33,785][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:30:34,292][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:30:34,794][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:30:35,297][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:30:35,800][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:30:36,301][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:30:36,804][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:30:37,305][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:30:37,809][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:30:38,318][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:30:38,827][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:30:39,336][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:30:39,838][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:30:40,340][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:30:40,856][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:30:41,362][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:30:41,878][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10726 tokens. +[2026-03-25 23:30:42,758][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 60.65%, Block Peak % of device VRAM: 62.49%, ΔTime: 00:00:33 +[2026-03-25 23:30:43,558][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:30:43,560][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:30:43,562][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:30:44,270][__main__][INFO] - Iteration 167 took 53s (31.65% Gen, 67.01% Train). Generation: 16s, Training: 35s. Estimated remaining time: 41h 51m 34s. Estimated total time: 44h 16m 38s. Time estimates for 10 more iterations: 8m 51s, 100 more iterations: 1h 28m 33s, 500 more iterations: 7h 22m 46s. +[2026-03-25 23:30:44,272][__main__][INFO] - Starting iteration 167. +[2026-03-25 23:30:44,681][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 16 and human policies 1. +[2026-03-25 23:30:44,681][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:31:00,849][__main__][INFO] - Number of regex retries in iteration 167: 0 +[2026-03-25 23:31:00,851][__main__][INFO] - agents played in iteration 167 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:31:01,760][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:31:01,787][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:31:01,813][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:31:01,839][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:31:01,840][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:31:01,841][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:31:02,465][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:31:02,923][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:31:03,430][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:31:03,935][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:31:04,438][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:31:04,940][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:31:05,445][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:31:05,951][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:31:06,460][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:31:06,975][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:31:07,480][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:31:07,982][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:31:08,501][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:31:09,014][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:31:09,534][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:31:10,045][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:31:10,568][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:31:11,077][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:31:11,580][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:31:12,083][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:31:12,586][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:31:13,090][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:31:13,592][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:31:14,095][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:31:14,598][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:31:15,101][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:31:15,616][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:31:16,129][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:31:16,635][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:31:17,137][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:31:17,653][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:31:18,166][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:31:18,682][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:31:19,193][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:31:19,711][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:31:20,220][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:31:20,729][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:31:21,231][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:31:21,733][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:31:22,237][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:31:22,739][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:31:23,241][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:31:23,747][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:31:24,254][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:31:24,763][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:31:25,270][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:31:25,773][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:31:26,275][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:31:26,781][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:31:27,298][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:31:27,813][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:31:28,329][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:31:28,844][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:31:29,355][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:31:29,862][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:31:30,362][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:31:30,863][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:31:31,364][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:31:31,865][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:31:32,368][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:31:32,870][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:31:33,375][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:31:33,885][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:31:34,393][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:31:34,900][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10732 tokens. +[2026-03-25 23:31:35,524][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 60.63%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:33 +[2026-03-25 23:31:36,382][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:31:36,384][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:31:36,387][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:31:37,198][__main__][INFO] - Iteration 168 took 52s (30.79% Gen, 67.66% Train). Generation: 16s, Training: 35s. Estimated remaining time: 41h 19m 56s. Estimated total time: 43h 45m 53s. Time estimates for 10 more iterations: 8m 45s, 100 more iterations: 1h 27m 31s, 500 more iterations: 7h 17m 38s. +[2026-03-25 23:31:37,201][__main__][INFO] - Starting iteration 168. +[2026-03-25 23:31:37,610][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 16 and human policies 1. +[2026-03-25 23:31:37,611][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:31:53,227][__main__][INFO] - Number of regex retries in iteration 168: 0 +[2026-03-25 23:31:53,228][__main__][INFO] - agents played in iteration 168 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:31:54,010][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:31:54,034][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:31:54,057][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:31:54,080][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:31:54,080][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:31:54,081][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:31:54,799][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:31:55,274][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:31:55,795][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:31:56,311][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:31:56,823][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:31:57,342][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:31:57,851][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:31:58,353][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:31:58,854][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:31:59,358][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:31:59,862][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:32:00,362][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:32:00,863][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:32:01,367][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:32:01,879][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:32:02,388][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:32:02,898][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:32:03,400][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:32:03,899][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:32:04,419][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:32:04,927][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:32:05,442][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:32:05,955][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:32:06,473][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:32:06,979][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:32:07,482][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:32:07,985][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:32:08,482][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:32:08,981][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:32:09,481][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:32:09,981][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:32:10,483][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:32:10,987][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:32:11,493][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:32:12,000][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:32:12,500][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:32:12,998][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:32:13,517][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:32:14,027][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:32:14,542][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:32:15,052][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:32:15,568][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:32:16,074][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:32:16,579][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:32:17,082][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:32:17,582][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:32:18,081][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:32:18,580][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:32:19,078][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:32:19,580][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:32:20,084][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:32:20,597][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:32:21,108][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:32:21,627][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:32:22,142][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:32:22,662][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:32:23,173][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:32:23,688][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:32:24,199][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:32:24,701][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:32:25,202][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:32:25,705][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:32:26,202][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:32:26,703][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:32:27,205][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10764 tokens. +[2026-03-25 23:32:27,907][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 60.63%, Block Peak % of device VRAM: 62.27%, ΔTime: 00:00:33 +[2026-03-25 23:32:28,656][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:32:28,658][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:32:28,660][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:32:29,412][__main__][INFO] - Iteration 169 took 51s (30.15% Gen, 68.40% Train). Generation: 15s, Training: 35s. Estimated remaining time: 40h 43m 18s. Estimated total time: 43h 10m 7s. Time estimates for 10 more iterations: 8m 38s, 100 more iterations: 1h 26m 20s, 500 more iterations: 7h 11m 41s. +[2026-03-25 23:32:29,414][__main__][INFO] - Starting iteration 169. +[2026-03-25 23:32:29,817][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 16 and human policies 1. +[2026-03-25 23:32:29,818][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:32:33,394][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 23:32:40,394][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 23:32:46,229][__main__][INFO] - Number of regex retries in iteration 169: 2 +[2026-03-25 23:32:46,230][__main__][INFO] - agents played in iteration 169 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:32:47,001][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:32:47,025][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:32:47,048][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:32:47,070][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:32:47,071][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:32:47,072][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:32:47,705][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:32:48,167][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:32:48,683][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:32:49,195][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:32:49,700][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:32:50,203][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:32:50,722][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:32:51,246][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:32:51,766][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:32:52,278][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:32:52,798][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:32:53,310][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:32:53,821][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:32:54,321][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:32:54,821][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:32:55,322][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:32:55,819][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:32:56,320][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:32:56,823][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:32:57,330][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:32:57,843][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:32:58,358][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:32:58,864][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:32:59,366][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:32:59,880][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:33:00,397][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:33:00,912][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:33:01,429][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:33:01,946][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:33:02,457][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:33:02,962][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:33:03,464][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:33:03,965][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:33:04,464][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:33:04,963][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:33:05,462][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:33:05,962][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:33:06,469][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:33:06,976][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:33:07,484][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:33:07,987][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:33:08,485][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:33:08,989][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:33:09,505][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:33:10,013][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:33:10,529][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:33:11,041][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:33:11,559][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:33:12,069][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:33:12,571][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:33:13,069][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:33:13,570][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:33:14,070][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:33:14,570][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:33:15,069][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:33:15,578][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:33:16,091][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:33:16,601][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:33:17,108][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:33:17,610][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:33:18,114][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:33:18,630][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:33:19,137][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:33:19,656][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:33:20,163][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10687 tokens. +[2026-03-25 23:33:21,010][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 60.63%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:33 +[2026-03-25 23:33:21,772][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:33:21,774][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:33:21,776][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:33:22,495][__main__][INFO] - Iteration 170 took 52s (31.15% Gen, 67.48% Train). Generation: 16s, Training: 35s. Estimated remaining time: 41h 26m 13s. Estimated total time: 43h 53m 56s. Time estimates for 10 more iterations: 8m 46s, 100 more iterations: 1h 27m 47s, 500 more iterations: 7h 18m 59s. +[2026-03-25 23:33:22,497][__main__][INFO] - Starting iteration 170. +[2026-03-25 23:33:22,897][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 16 and human policies 1. +[2026-03-25 23:33:22,897][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:33:40,520][__main__][INFO] - Number of regex retries in iteration 170: 0 +[2026-03-25 23:33:40,521][__main__][INFO] - agents played in iteration 170 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:33:41,297][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:33:41,322][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:33:41,345][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:33:41,369][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:33:41,370][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:33:41,370][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:33:41,947][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:33:42,404][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:33:42,910][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:33:43,417][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:33:43,925][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:33:44,436][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:33:44,942][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:33:45,442][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:33:45,942][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:33:46,458][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:33:46,972][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:33:47,490][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:33:48,003][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:33:48,521][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:33:49,031][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:33:49,534][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:33:50,036][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:33:50,538][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:33:51,035][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:33:51,537][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:33:52,039][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:33:52,546][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:33:53,057][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:33:53,568][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:33:54,075][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:33:54,576][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:33:55,084][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:33:55,600][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:33:56,105][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:33:56,622][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:33:57,137][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:33:57,654][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:33:58,166][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:33:58,668][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:33:59,169][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:33:59,666][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:34:00,188][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:34:00,692][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:34:01,193][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:34:01,704][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:34:02,219][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:34:02,728][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:34:03,237][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:34:03,740][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:34:04,243][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:34:04,762][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:34:05,271][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:34:05,787][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:34:06,300][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:34:06,817][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:34:07,324][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:34:07,827][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:34:08,328][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:34:08,829][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:34:09,330][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:34:09,834][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:34:10,336][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:34:10,843][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:34:11,356][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:34:11,866][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:34:12,376][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:34:12,879][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:34:13,384][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:34:13,905][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:34:14,413][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10757 tokens. +[2026-03-25 23:34:15,283][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 60.63%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:33 +[2026-03-25 23:34:16,159][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:34:16,161][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:34:16,162][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:34:17,465][__main__][INFO] - Iteration 171 took 54s (32.30% Gen, 65.31% Train). Generation: 17s, Training: 35s. Estimated remaining time: 42h 59m 51s. Estimated total time: 45h 28m 28s. Time estimates for 10 more iterations: 9m 5s, 100 more iterations: 1h 30m 56s, 500 more iterations: 7h 34m 44s. +[2026-03-25 23:34:17,468][__main__][INFO] - Starting iteration 171. +[2026-03-25 23:34:17,867][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 17 and human policies 1. +[2026-03-25 23:34:17,868][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:34:32,570][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 23:34:33,431][__main__][INFO] - Number of regex retries in iteration 171: 1 +[2026-03-25 23:34:33,433][__main__][INFO] - agents played in iteration 171 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:34:34,489][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:34:34,517][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:34:34,546][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:34:34,572][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:34:34,573][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:34:34,574][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:34:35,283][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:34:35,741][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:34:36,249][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:34:36,753][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:34:37,255][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:34:37,758][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:34:38,262][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:34:38,767][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:34:39,282][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:34:39,787][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:34:40,296][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:34:40,797][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:34:41,298][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:34:41,818][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:34:42,322][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:34:42,835][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:34:43,339][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:34:43,859][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:34:44,363][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:34:44,865][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:34:45,364][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:34:45,868][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:34:46,372][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:34:46,874][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:34:47,379][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:34:47,885][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:34:48,395][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:34:48,907][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:34:49,425][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:34:49,927][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:34:50,432][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:34:50,955][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:34:51,466][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:34:51,984][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:34:52,493][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:34:53,012][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:34:53,519][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:34:54,023][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:34:54,524][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:34:55,026][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:34:55,526][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:34:56,049][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:34:56,554][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:34:57,061][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:34:57,571][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:34:58,078][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:34:58,588][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:34:59,089][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:34:59,591][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:35:00,111][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:35:00,619][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:35:01,134][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:35:01,641][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:35:02,155][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:35:02,663][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:35:03,169][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:35:03,672][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:35:04,174][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:35:04,674][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:35:05,173][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:35:05,672][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:35:06,174][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:35:06,684][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:35:07,190][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:35:07,702][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10687 tokens. +[2026-03-25 23:35:08,332][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 60.63%, Block Peak % of device VRAM: 62.36%, ΔTime: 00:00:33 +[2026-03-25 23:35:09,162][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:35:09,165][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:35:09,167][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:35:09,956][__main__][INFO] - Iteration 172 took 52s (29.88% Gen, 68.60% Train). Generation: 15s, Training: 35s. Estimated remaining time: 40h 54m 59s. Estimated total time: 43h 24m 29s. Time estimates for 10 more iterations: 8m 40s, 100 more iterations: 1h 26m 48s, 500 more iterations: 7h 14m 4s. +[2026-03-25 23:35:09,960][__main__][INFO] - Starting iteration 172. +[2026-03-25 23:35:10,378][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 17 and human policies 1. +[2026-03-25 23:35:10,379][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:35:27,398][__main__][INFO] - Number of regex retries in iteration 172: 0 +[2026-03-25 23:35:27,399][__main__][INFO] - agents played in iteration 172 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:35:28,307][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:35:28,332][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:35:28,359][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:35:28,383][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:35:28,385][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:35:28,386][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:35:29,259][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:35:29,732][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:35:30,248][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:35:30,759][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:35:31,262][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:35:31,765][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:35:32,266][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:35:32,767][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:35:33,271][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:35:33,773][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:35:34,282][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:35:34,789][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:35:35,297][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:35:35,818][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:35:36,336][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:35:36,858][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:35:37,369][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:35:37,892][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:35:38,402][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:35:38,904][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:35:39,406][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:35:39,908][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:35:40,410][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:35:40,913][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:35:41,420][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:35:41,927][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:35:42,430][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:35:42,934][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:35:43,436][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:35:43,939][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:35:44,440][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:35:44,947][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:35:45,456][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:35:45,962][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:35:46,461][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:35:46,975][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:35:47,492][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:35:48,005][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:35:48,519][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:35:49,031][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:35:49,543][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:35:50,046][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:35:50,548][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:35:51,053][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:35:51,553][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:35:52,053][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:35:52,555][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:35:53,058][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:35:53,560][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:35:54,069][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:35:54,573][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:35:55,075][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:35:55,580][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:35:56,097][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:35:56,609][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:35:57,123][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:35:57,639][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:35:58,150][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:35:58,661][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:35:59,163][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:35:59,664][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:36:00,166][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:36:00,667][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:36:01,170][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:36:01,669][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10683 tokens. +[2026-03-25 23:36:02,390][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.58%, Block Peak % of device VRAM: 62.24%, ΔTime: 00:00:33 +[2026-03-25 23:36:03,198][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:36:03,201][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:36:03,202][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:36:03,946][__main__][INFO] - Iteration 173 took 53s (31.77% Gen, 66.84% Train). Generation: 17s, Training: 35s. Estimated remaining time: 42h 8m 3s. Estimated total time: 44h 38m 27s. Time estimates for 10 more iterations: 8m 55s, 100 more iterations: 1h 29m 16s, 500 more iterations: 7h 26m 24s. +[2026-03-25 23:36:03,948][__main__][INFO] - Starting iteration 173. +[2026-03-25 23:36:04,347][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 17 and human policies 1. +[2026-03-25 23:36:04,347][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:36:17,008][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 23:36:21,402][__main__][INFO] - Number of regex retries in iteration 173: 1 +[2026-03-25 23:36:21,403][__main__][INFO] - agents played in iteration 173 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:36:22,284][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:36:22,308][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:36:22,332][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:36:22,355][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:36:22,356][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:36:22,356][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:36:22,961][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:36:23,420][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:36:23,945][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:36:24,456][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:36:24,976][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:36:25,496][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:36:26,016][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:36:26,524][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:36:27,026][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:36:27,527][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:36:28,029][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:36:28,529][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:36:29,033][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:36:29,534][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:36:30,039][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:36:30,548][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:36:31,061][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:36:31,571][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:36:32,074][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:36:32,581][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:36:33,101][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:36:33,604][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:36:34,119][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:36:34,628][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:36:35,142][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:36:35,651][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:36:36,153][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:36:36,653][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:36:37,149][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:36:37,649][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:36:38,148][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:36:38,650][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:36:39,153][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:36:39,665][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:36:40,171][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:36:40,685][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:36:41,182][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:36:41,683][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:36:42,200][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:36:42,704][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:36:43,221][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:36:43,732][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:36:44,252][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:36:44,760][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:36:45,263][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:36:45,764][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:36:46,265][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:36:46,765][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:36:47,266][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:36:47,765][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:36:48,269][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:36:48,777][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:36:49,282][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:36:49,794][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:36:50,295][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:36:50,796][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:36:51,314][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:36:51,827][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:36:52,350][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:36:52,860][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:36:53,381][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:36:53,890][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:36:54,396][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:36:54,900][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:36:55,401][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10754 tokens. +[2026-03-25 23:36:56,023][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.30%, Current % of VRAM taken: 60.72%, Block Peak % of device VRAM: 62.40%, ΔTime: 00:00:33 +[2026-03-25 23:36:56,772][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:36:56,774][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:36:56,776][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:36:57,537][__main__][INFO] - Iteration 174 took 53s (32.07% Gen, 66.50% Train). Generation: 17s, Training: 35s. Estimated remaining time: 41h 48m 14s. Estimated total time: 44h 19m 31s. Time estimates for 10 more iterations: 8m 51s, 100 more iterations: 1h 28m 39s, 500 more iterations: 7h 23m 15s. +[2026-03-25 23:36:57,539][__main__][INFO] - Starting iteration 174. +[2026-03-25 23:36:57,945][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 17 and human policies 1. +[2026-03-25 23:36:57,946][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:37:04,279][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 23:37:14,475][__main__][INFO] - Number of regex retries in iteration 174: 1 +[2026-03-25 23:37:14,476][__main__][INFO] - agents played in iteration 174 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:37:15,270][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:37:15,294][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:37:15,317][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:37:15,340][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:37:15,340][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:37:15,341][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:37:15,976][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:37:16,433][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:37:16,951][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:37:17,461][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:37:17,966][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:37:18,467][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:37:18,982][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:37:19,496][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:37:20,017][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:37:20,534][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:37:21,055][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:37:21,583][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:37:22,093][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:37:22,599][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:37:23,101][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:37:23,601][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:37:24,102][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:37:24,602][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:37:25,106][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:37:25,609][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:37:26,123][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:37:26,635][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:37:27,138][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:37:27,639][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:37:28,156][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:37:28,667][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:37:29,186][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:37:29,697][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:37:30,216][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:37:30,727][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:37:31,230][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:37:31,733][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:37:32,233][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:37:32,732][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:37:33,231][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:37:33,727][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:37:34,230][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:37:34,735][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:37:35,242][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:37:35,753][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:37:36,258][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:37:36,759][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:37:37,270][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:37:37,787][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:37:38,303][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:37:38,814][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:37:39,329][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:37:39,839][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:37:40,348][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:37:40,851][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:37:41,355][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:37:41,857][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:37:42,359][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:37:42,860][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:37:43,363][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:37:43,869][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:37:44,378][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:37:44,887][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:37:45,392][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:37:45,894][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:37:46,407][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:37:46,925][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:37:47,441][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:37:47,960][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:37:48,479][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10727 tokens. +[2026-03-25 23:37:49,279][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.29%, Current % of VRAM taken: 60.72%, Block Peak % of device VRAM: 62.49%, ΔTime: 00:00:33 +[2026-03-25 23:37:50,039][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:37:50,041][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:37:50,042][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:37:50,763][__main__][INFO] - Iteration 175 took 52s (31.30% Gen, 67.34% Train). Generation: 16s, Training: 35s. Estimated remaining time: 41h 28m 46s. Estimated total time: 44h 0m 56s. Time estimates for 10 more iterations: 8m 48s, 100 more iterations: 1h 28m 1s, 500 more iterations: 7h 20m 9s. +[2026-03-25 23:37:50,765][__main__][INFO] - Starting iteration 175. +[2026-03-25 23:37:51,165][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 17 and human policies 1. +[2026-03-25 23:37:51,166][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:37:55,229][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 23:38:08,461][__main__][INFO] - Number of regex retries in iteration 175: 1 +[2026-03-25 23:38:08,461][__main__][INFO] - agents played in iteration 175 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:38:09,248][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.59%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:38:09,271][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.59%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:38:09,294][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.59%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:38:09,317][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.59%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:38:09,318][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:38:09,318][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:38:09,907][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:38:10,366][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:38:10,873][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:38:11,382][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:38:11,898][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:38:12,411][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:38:12,922][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:38:13,423][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:38:13,927][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:38:14,447][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:38:14,958][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:38:15,477][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:38:15,985][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:38:16,502][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:38:17,011][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:38:17,516][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:38:18,018][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:38:18,517][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:38:19,013][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:38:19,511][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:38:20,012][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:38:20,515][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:38:21,032][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:38:21,539][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:38:22,052][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:38:22,557][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:38:23,060][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:38:23,577][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:38:24,087][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:38:24,603][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:38:25,111][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:38:25,627][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:38:26,137][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:38:26,641][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:38:27,143][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:38:27,644][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:38:28,144][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:38:28,645][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:38:29,144][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:38:29,647][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:38:30,155][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:38:30,664][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:38:31,177][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:38:31,680][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:38:32,182][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:38:32,702][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:38:33,214][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:38:33,735][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:38:34,247][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:38:34,768][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:38:35,276][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:38:35,781][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:38:36,283][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:38:36,783][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:38:37,281][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:38:37,780][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:38:38,280][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:38:38,780][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:38:39,286][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:38:39,794][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:38:40,304][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:38:40,808][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:38:41,312][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:38:41,824][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:38:42,340][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10749 tokens. +[2026-03-25 23:38:43,302][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 60.63%, Block Peak % of device VRAM: 62.28%, ΔTime: 00:00:33 +[2026-03-25 23:38:44,183][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:38:44,186][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:38:44,187][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:38:44,941][__main__][INFO] - Iteration 176 took 53s (32.16% Gen, 66.43% Train). Generation: 17s, Training: 35s. Estimated remaining time: 42h 15m 45s. Estimated total time: 44h 48m 50s. Time estimates for 10 more iterations: 8m 57s, 100 more iterations: 1h 29m 37s, 500 more iterations: 7h 28m 8s. +[2026-03-25 23:38:44,944][__main__][INFO] - Starting iteration 176. +[2026-03-25 23:38:45,345][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 17 and human policies 1. +[2026-03-25 23:38:45,345][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:39:00,982][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 1球 did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 23:39:01,960][__main__][INFO] - Number of regex retries in iteration 176: 1 +[2026-03-25 23:39:01,961][__main__][INFO] - agents played in iteration 176 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:39:02,912][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:39:02,936][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:39:02,959][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:39:02,982][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:39:02,983][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:39:02,984][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:39:03,884][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:39:04,349][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:39:04,857][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:39:05,363][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:39:05,860][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:39:06,362][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:39:06,863][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:39:07,365][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:39:07,872][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:39:08,381][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:39:08,883][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:39:09,383][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:39:09,886][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:39:10,400][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:39:10,908][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:39:11,425][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:39:11,939][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:39:12,457][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:39:12,962][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:39:13,465][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:39:13,969][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:39:14,470][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:39:14,970][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:39:15,471][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:39:15,972][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:39:16,476][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:39:16,988][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:39:17,495][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:39:18,003][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:39:18,504][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:39:19,004][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:39:19,520][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:39:20,027][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:39:20,549][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:39:21,058][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:39:21,579][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:39:22,088][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:39:22,594][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:39:23,096][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:39:23,597][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:39:24,097][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:39:24,596][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:39:25,097][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:39:25,601][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:39:26,113][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:39:26,620][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:39:27,131][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:39:27,635][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:39:28,135][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:39:28,654][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:39:29,161][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:39:29,677][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:39:30,183][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:39:30,698][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:39:31,207][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:39:31,711][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:39:32,213][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:39:32,714][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:39:33,215][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:39:33,714][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:39:34,215][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:39:34,716][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:39:35,228][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:39:35,734][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:39:36,248][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10720 tokens. +[2026-03-25 23:39:36,903][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.29%, Current % of VRAM taken: 60.72%, Block Peak % of device VRAM: 62.49%, ΔTime: 00:00:33 +[2026-03-25 23:39:37,737][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:39:37,739][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:39:37,741][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:39:38,542][__main__][INFO] - Iteration 177 took 53s (31.23% Gen, 67.26% Train). Generation: 16s, Training: 35s. Estimated remaining time: 41h 45m 56s. Estimated total time: 44h 19m 55s. Time estimates for 10 more iterations: 8m 51s, 100 more iterations: 1h 28m 39s, 500 more iterations: 7h 23m 19s. +[2026-03-25 23:39:38,545][__main__][INFO] - Starting iteration 177. +[2026-03-25 23:39:38,964][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 17 and human policies 1. +[2026-03-25 23:39:38,965][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:39:56,146][__main__][INFO] - Number of regex retries in iteration 177: 0 +[2026-03-25 23:39:56,147][__main__][INFO] - agents played in iteration 177 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:39:57,045][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.60%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:39:57,073][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.60%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:39:57,103][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.60%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:39:57,131][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.60%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:39:57,133][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:39:57,134][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:39:57,996][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:39:58,482][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:39:58,997][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:39:59,503][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:40:00,011][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:40:00,512][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:40:01,011][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:40:01,509][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:40:02,012][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:40:02,520][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:40:03,025][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:40:03,534][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:40:04,049][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:40:04,550][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:40:05,050][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:40:05,571][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:40:06,080][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:40:06,600][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:40:07,113][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:40:07,634][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:40:08,143][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:40:08,648][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:40:09,156][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:40:09,660][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:40:10,160][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:40:10,659][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:40:11,159][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:40:11,663][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:40:12,169][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:40:12,674][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:40:13,185][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:40:13,687][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:40:14,186][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:40:14,702][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:40:15,213][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:40:15,733][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:40:16,242][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:40:16,761][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:40:17,268][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:40:17,773][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:40:18,272][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:40:18,779][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:40:19,277][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:40:19,779][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:40:20,281][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:40:20,784][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:40:21,285][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:40:21,796][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:40:22,305][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:40:22,811][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:40:23,312][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:40:23,828][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:40:24,340][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:40:24,859][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:40:25,370][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:40:25,890][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:40:26,402][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:40:26,906][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:40:27,405][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:40:27,905][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:40:28,403][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:40:28,902][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:40:29,401][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:40:29,903][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:40:30,407][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10692 tokens. +[2026-03-25 23:40:31,178][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 60.61%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:33 +[2026-03-25 23:40:32,027][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:40:32,029][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:40:32,031][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:40:32,698][__main__][INFO] - Iteration 178 took 53s (31.98% Gen, 66.78% Train). Generation: 17s, Training: 35s. Estimated remaining time: 42h 11m 49s. Estimated total time: 44h 46m 42s. Time estimates for 10 more iterations: 8m 57s, 100 more iterations: 1h 29m 33s, 500 more iterations: 7h 27m 47s. +[2026-03-25 23:40:32,700][__main__][INFO] - Starting iteration 178. +[2026-03-25 23:40:33,122][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 17 and human policies 1. +[2026-03-25 23:40:33,123][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:40:51,772][__main__][INFO] - Number of regex retries in iteration 178: 0 +[2026-03-25 23:40:51,773][__main__][INFO] - agents played in iteration 178 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:40:52,808][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:40:52,835][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:40:52,862][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:40:52,889][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:40:52,890][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:40:52,891][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:40:53,749][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:40:54,227][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:40:54,740][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:40:55,241][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:40:55,738][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:40:56,240][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:40:56,741][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:40:57,240][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:40:57,744][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:40:58,249][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:40:58,761][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:40:59,274][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:40:59,782][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:41:00,281][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:41:00,794][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:41:01,310][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:41:01,826][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:41:02,345][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:41:02,862][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:41:03,371][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:41:03,881][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:41:04,382][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:41:04,883][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:41:05,380][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:41:05,880][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:41:06,383][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:41:06,882][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:41:07,389][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:41:07,895][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:41:08,406][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:41:08,906][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:41:09,403][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:41:09,903][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:41:10,421][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:41:10,934][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:41:11,453][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:41:11,968][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:41:12,483][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:41:12,990][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:41:13,493][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:41:13,994][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:41:14,490][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:41:14,991][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:41:15,491][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:41:15,990][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:41:16,498][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:41:17,012][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:41:17,521][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:41:18,030][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:41:18,532][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:41:19,038][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:41:19,551][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:41:20,059][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:41:20,577][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:41:21,095][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:41:21,606][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:41:22,115][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:41:22,618][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:41:23,118][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:41:23,620][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:41:24,118][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:41:24,622][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:41:25,126][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:41:25,632][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:41:26,143][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10657 tokens. +[2026-03-25 23:41:26,939][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 60.62%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:33 +[2026-03-25 23:41:27,697][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:41:27,700][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:41:27,701][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:41:28,530][__main__][INFO] - Iteration 179 took 55s (33.66% Gen, 64.84% Train). Generation: 18s, Training: 35s. Estimated remaining time: 43h 34m 38s. Estimated total time: 46h 10m 26s. Time estimates for 10 more iterations: 9m 14s, 100 more iterations: 1h 32m 20s, 500 more iterations: 7h 41m 44s. +[2026-03-25 23:41:28,533][__main__][INFO] - Starting iteration 179. +[2026-03-25 23:41:28,943][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 17 and human policies 1. +[2026-03-25 23:41:28,944][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:41:44,713][__main__][INFO] - Number of regex retries in iteration 179: 0 +[2026-03-25 23:41:44,714][__main__][INFO] - agents played in iteration 179 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:41:45,586][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:41:45,611][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:41:45,635][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:41:45,658][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:41:45,658][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:41:45,659][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:41:46,287][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:41:46,745][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:41:47,272][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:41:47,780][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:41:48,301][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:41:48,809][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:41:49,328][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:41:49,837][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:41:50,341][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:41:50,842][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:41:51,342][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:41:51,841][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:41:52,338][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:41:52,839][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:41:53,345][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:41:53,853][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:41:54,360][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:41:54,876][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:41:55,377][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:41:55,879][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:41:56,401][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:41:56,911][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:41:57,434][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:41:57,946][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:41:58,466][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:41:58,973][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:41:59,477][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:41:59,979][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:42:00,482][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:42:00,984][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:42:01,485][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:42:01,993][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:42:02,498][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:42:03,004][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:42:03,513][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:42:04,024][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:42:04,546][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:42:05,058][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:42:05,576][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:42:06,093][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:42:06,610][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:42:07,117][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:42:07,620][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:42:08,121][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:42:08,620][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:42:09,122][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:42:09,624][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:42:10,130][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:42:10,636][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:42:11,138][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:42:11,641][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:42:12,142][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:42:12,643][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:42:13,150][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:42:13,660][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:42:14,164][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:42:14,667][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:42:15,169][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:42:15,687][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:42:16,196][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:42:16,716][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:42:17,226][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:42:17,744][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:42:18,250][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:42:18,755][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10798 tokens. +[2026-03-25 23:42:19,421][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 60.65%, Block Peak % of device VRAM: 62.41%, ΔTime: 00:00:33 +[2026-03-25 23:42:20,178][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:42:20,180][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:42:20,182][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:42:20,911][__main__][INFO] - Iteration 180 took 51s (30.34% Gen, 68.25% Train). Generation: 15s, Training: 35s. Estimated remaining time: 40h 41m 48s. Estimated total time: 43h 18m 28s. Time estimates for 10 more iterations: 8m 39s, 100 more iterations: 1h 26m 36s, 500 more iterations: 7h 13m 4s. +[2026-03-25 23:42:20,914][__main__][INFO] - Starting iteration 180. +[2026-03-25 23:42:21,314][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 17 and human policies 1. +[2026-03-25 23:42:21,314][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:42:38,386][__main__][INFO] - Number of regex retries in iteration 180: 0 +[2026-03-25 23:42:38,387][__main__][INFO] - agents played in iteration 180 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:42:39,164][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:42:39,189][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:42:39,213][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:42:39,236][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:42:39,237][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:42:39,237][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:42:39,900][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:42:40,363][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:42:40,873][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:42:41,432][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:42:41,941][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:42:42,441][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:42:42,963][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:42:43,469][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:42:43,986][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:42:44,549][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:42:45,063][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:42:45,572][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:42:46,071][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:42:46,575][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:42:47,071][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:42:47,566][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:42:48,067][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:42:48,567][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:42:49,069][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:42:49,576][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:42:50,081][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:42:50,592][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:42:51,093][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:42:51,594][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:42:52,110][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:42:52,622][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:42:53,143][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:42:53,653][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:42:54,173][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:42:54,681][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:42:55,184][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:42:55,684][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:42:56,184][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:42:56,680][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:42:57,180][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:42:57,679][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:42:58,180][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:42:58,690][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:42:59,192][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:42:59,700][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:43:00,199][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:43:00,696][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:43:01,218][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:43:01,729][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:43:02,249][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:43:02,758][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:43:03,280][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:43:03,785][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:43:04,286][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:43:04,788][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:43:05,291][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:43:05,787][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:43:06,288][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:43:06,787][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:43:07,292][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:43:07,793][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:43:08,300][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:43:08,813][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:43:09,315][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:43:09,816][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:43:10,335][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:43:10,844][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:43:11,360][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:43:11,869][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:43:12,384][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10714 tokens. +[2026-03-25 23:43:13,131][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 60.63%, Block Peak % of device VRAM: 62.49%, ΔTime: 00:00:33 +[2026-03-25 23:43:13,783][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:43:13,785][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:43:13,786][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:43:15,219][__main__][INFO] - Iteration 181 took 53s (31.67% Gen, 65.67% Train). Generation: 17s, Training: 35s. Estimated remaining time: 42h 17m 43s. Estimated total time: 44h 55m 17s. Time estimates for 10 more iterations: 8m 59s, 100 more iterations: 1h 29m 50s, 500 more iterations: 7h 29m 12s. +[2026-03-25 23:43:15,222][__main__][INFO] - Starting iteration 181. +[2026-03-25 23:43:15,623][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 18 and human policies 1. +[2026-03-25 23:43:15,624][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:43:45,945][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 23:43:47,990][__main__][INFO] - Number of regex retries in iteration 181: 1 +[2026-03-25 23:43:47,991][__main__][INFO] - agents played in iteration 181 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:43:48,943][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:43:48,972][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:43:49,002][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:43:49,031][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:43:49,033][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:43:49,034][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:43:49,836][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:43:50,297][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:43:50,805][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:43:51,309][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:43:51,811][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:43:52,315][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:43:52,818][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:43:53,323][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:43:53,826][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:43:54,337][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:43:54,853][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:43:55,361][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:43:55,863][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:43:56,382][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:43:56,895][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:43:57,415][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:43:57,928][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:43:58,452][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:43:58,962][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:43:59,466][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:43:59,970][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:44:00,473][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:44:00,976][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:44:01,478][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:44:01,979][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:44:02,484][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:44:02,989][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:44:03,497][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:44:04,007][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:44:04,511][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:44:05,013][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:44:05,528][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:44:06,037][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:44:06,555][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:44:07,066][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:44:07,583][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:44:08,093][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:44:08,603][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:44:09,103][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:44:09,606][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:44:10,108][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:44:10,610][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:44:11,110][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:44:11,615][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:44:12,117][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:44:12,626][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:44:13,138][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:44:13,642][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:44:14,144][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:44:14,656][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:44:15,170][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:44:15,684][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:44:16,196][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:44:16,715][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:44:17,230][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:44:17,739][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:44:18,241][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:44:18,741][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:44:19,243][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:44:19,744][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:44:20,248][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:44:20,753][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:44:21,260][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:44:21,768][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:44:22,279][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10752 tokens. +[2026-03-25 23:44:22,983][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 60.56%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:33 +[2026-03-25 23:44:23,749][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:44:23,752][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:44:23,754][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:44:24,560][__main__][INFO] - Iteration 182 took 1m 8s (46.95% Gen, 51.88% Train). Generation: 32s, Training: 35s. Estimated remaining time: 54h 48m 8s. Estimated total time: 57h 26m 52s. Time estimates for 10 more iterations: 11m 29s, 100 more iterations: 1h 54m 53s, 500 more iterations: 9h 34m 28s. +[2026-03-25 23:44:24,562][__main__][INFO] - Starting iteration 182. +[2026-03-25 23:44:24,983][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 18 and human policies 1. +[2026-03-25 23:44:24,984][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:44:38,737][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 23:44:50,404][__main__][INFO] - Number of regex retries in iteration 182: 1 +[2026-03-25 23:44:50,405][__main__][INFO] - agents played in iteration 182 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:44:51,184][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:44:51,208][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:44:51,231][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:44:51,254][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:44:51,255][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:44:51,255][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:44:52,156][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:44:52,620][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:44:53,147][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:44:53,662][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:44:54,183][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:44:54,694][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:44:55,197][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:44:55,703][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:44:56,208][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:44:56,712][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:44:57,216][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:44:57,718][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:44:58,226][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:44:58,740][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:44:59,246][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:44:59,763][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:45:00,266][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:45:00,773][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:45:01,295][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:45:01,805][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:45:02,323][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:45:02,833][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:45:03,350][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:45:03,860][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:45:04,362][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:45:04,865][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:45:05,368][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:45:05,870][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:45:06,374][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:45:06,878][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:45:07,383][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:45:07,892][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:45:08,398][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:45:08,907][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:45:09,410][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:45:09,911][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:45:10,432][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:45:10,943][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:45:11,464][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:45:11,968][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:45:12,485][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:45:12,992][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:45:13,502][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:45:14,005][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:45:14,509][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:45:15,010][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:45:15,512][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:45:16,014][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:45:16,519][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:45:17,023][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:45:17,531][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:45:18,041][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:45:18,562][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:45:19,072][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:45:19,590][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:45:20,101][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:45:20,619][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:45:21,131][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:45:21,637][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:45:22,140][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:45:22,642][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:45:23,144][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:45:23,644][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:45:24,149][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:45:24,658][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10725 tokens. +[2026-03-25 23:45:25,329][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.25%, Current % of VRAM taken: 60.67%, Block Peak % of device VRAM: 62.49%, ΔTime: 00:00:33 +[2026-03-25 23:45:26,072][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:45:26,074][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:45:26,075][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:45:26,824][__main__][INFO] - Iteration 183 took 1m 1s (41.10% Gen, 57.68% Train). Generation: 25s, Training: 35s. Estimated remaining time: 48h 52m 23s. Estimated total time: 51h 32m 10s. Time estimates for 10 more iterations: 10m 18s, 100 more iterations: 1h 43m 4s, 500 more iterations: 8h 35m 21s. +[2026-03-25 23:45:26,826][__main__][INFO] - Starting iteration 183. +[2026-03-25 23:45:27,229][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 18 and human policies 1. +[2026-03-25 23:45:27,230][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:45:31,195][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 23:45:43,939][__main__][INFO] - Number of regex retries in iteration 183: 1 +[2026-03-25 23:45:43,940][__main__][INFO] - agents played in iteration 183 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:45:44,750][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:45:44,774][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:45:44,797][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:45:44,819][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:45:44,820][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:45:44,820][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:45:45,568][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:45:46,041][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:45:46,550][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:45:47,053][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:45:47,572][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:45:48,083][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:45:48,603][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:45:49,110][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:45:49,629][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:45:50,138][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:45:50,644][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:45:51,145][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:45:51,647][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:45:52,146][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:45:52,648][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:45:53,152][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:45:53,660][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:45:54,172][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:45:54,683][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:45:55,193][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:45:55,699][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:45:56,202][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:45:56,731][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:45:57,252][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:45:57,768][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:45:58,284][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:45:58,801][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:45:59,309][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:45:59,813][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:46:00,314][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:46:00,812][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:46:01,314][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:46:01,815][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:46:02,318][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:46:02,821][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:46:03,326][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:46:03,838][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:46:04,347][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:46:04,857][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:46:05,359][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:46:05,865][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:46:06,382][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:46:06,886][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:46:07,403][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:46:07,915][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:46:08,430][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:46:08,937][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:46:09,442][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:46:09,946][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:46:10,451][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:46:10,956][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:46:11,460][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:46:11,965][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:46:12,473][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:46:12,984][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:46:13,491][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:46:14,000][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:46:14,502][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:46:15,005][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:46:15,520][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:46:16,025][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:46:16,542][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:46:17,045][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:46:17,561][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:46:18,067][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10748 tokens. +[2026-03-25 23:46:18,749][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 60.61%, Block Peak % of device VRAM: 62.41%, ΔTime: 00:00:33 +[2026-03-25 23:46:19,498][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:46:19,500][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:46:19,502][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:46:20,206][__main__][INFO] - Iteration 184 took 52s (31.54% Gen, 67.12% Train). Generation: 16s, Training: 35s. Estimated remaining time: 41h 28m 14s. Estimated total time: 44h 8m 54s. Time estimates for 10 more iterations: 8m 49s, 100 more iterations: 1h 28m 17s, 500 more iterations: 7h 21m 29s. +[2026-03-25 23:46:20,208][__main__][INFO] - Starting iteration 184. +[2026-03-25 23:46:20,609][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 18 and human policies 1. +[2026-03-25 23:46:20,610][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:46:26,700][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 23:46:37,565][__main__][INFO] - Number of regex retries in iteration 184: 1 +[2026-03-25 23:46:37,566][__main__][INFO] - agents played in iteration 184 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:46:38,358][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:46:38,381][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:46:38,405][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:46:38,427][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:46:38,428][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:46:38,428][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:46:39,053][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:46:39,511][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:46:40,018][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:46:40,526][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:46:41,034][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:46:41,548][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:46:42,051][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:46:42,554][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:46:43,069][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:46:43,581][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:46:44,101][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:46:44,610][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:46:45,131][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:46:45,638][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:46:46,143][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:46:46,647][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:46:47,147][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:46:47,650][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:46:48,152][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:46:48,654][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:46:49,158][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:46:49,664][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:46:50,178][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:46:50,690][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:46:51,195][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:46:51,697][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:46:52,212][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:46:52,724][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:46:53,240][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:46:53,753][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:46:54,274][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:46:54,782][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:46:55,285][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:46:55,787][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:46:56,293][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:46:56,795][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:46:57,296][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:46:57,797][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:46:58,299][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:46:58,803][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:46:59,306][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:46:59,812][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:47:00,318][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:47:00,829][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:47:01,331][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:47:01,836][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:47:02,350][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:47:02,862][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:47:03,380][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:47:03,891][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:47:04,408][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:47:04,917][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:47:05,421][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:47:05,923][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:47:06,425][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:47:06,926][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:47:07,427][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:47:07,927][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:47:08,428][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:47:08,933][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:47:09,445][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:47:09,951][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:47:10,455][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:47:10,958][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:47:11,467][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10703 tokens. +[2026-03-25 23:47:12,385][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.57%, Block Peak % of device VRAM: 62.24%, ΔTime: 00:00:33 +[2026-03-25 23:47:13,246][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:47:13,248][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:47:13,250][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:47:14,122][__main__][INFO] - Iteration 185 took 53s (31.69% Gen, 66.68% Train). Generation: 16s, Training: 35s. Estimated remaining time: 41h 54m 4s. Estimated total time: 44h 35m 38s. Time estimates for 10 more iterations: 8m 55s, 100 more iterations: 1h 29m 11s, 500 more iterations: 7h 25m 56s. +[2026-03-25 23:47:14,124][__main__][INFO] - Starting iteration 185. +[2026-03-25 23:47:14,527][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 18 and human policies 1. +[2026-03-25 23:47:14,528][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:47:30,566][__main__][INFO] - Number of regex retries in iteration 185: 0 +[2026-03-25 23:47:30,567][__main__][INFO] - agents played in iteration 185 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:47:31,553][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:47:31,578][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:47:31,602][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:47:31,626][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:47:31,627][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:47:31,627][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:47:32,528][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:47:32,993][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:47:33,500][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:47:34,006][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:47:34,511][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:47:35,014][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:47:35,519][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:47:36,023][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:47:36,529][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:47:37,038][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:47:37,545][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:47:38,059][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:47:38,561][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:47:39,060][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:47:39,579][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:47:40,085][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:47:40,602][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:47:41,110][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:47:41,626][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:47:42,134][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:47:42,639][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:47:43,140][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:47:43,640][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:47:44,143][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:47:44,644][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:47:45,147][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:47:45,652][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:47:46,159][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:47:46,667][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:47:47,178][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:47:47,681][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:47:48,183][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:47:48,703][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:47:49,215][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:47:49,737][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:47:50,247][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:47:50,768][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:47:51,276][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:47:51,781][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:47:52,284][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:47:52,786][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:47:53,286][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:47:53,784][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:47:54,284][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:47:54,785][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:47:55,295][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:47:55,801][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:47:56,314][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:47:56,816][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:47:57,318][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:47:57,837][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:47:58,344][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:47:58,864][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:47:59,377][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:47:59,895][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:48:00,401][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:48:00,905][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:48:01,408][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:48:01,908][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:48:02,407][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:48:02,907][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:48:03,408][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:48:03,911][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:48:04,415][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:48:04,924][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10759 tokens. +[2026-03-25 23:48:05,689][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.26%, Current % of VRAM taken: 60.68%, Block Peak % of device VRAM: 62.41%, ΔTime: 00:00:33 +[2026-03-25 23:48:06,428][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:48:06,431][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:48:06,432][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:48:07,246][__main__][INFO] - Iteration 186 took 52s (30.42% Gen, 68.03% Train). Generation: 16s, Training: 35s. Estimated remaining time: 41h 13m 32s. Estimated total time: 43h 55m 59s. Time estimates for 10 more iterations: 8m 47s, 100 more iterations: 1h 27m 51s, 500 more iterations: 7h 19m 19s. +[2026-03-25 23:48:07,249][__main__][INFO] - Starting iteration 186. +[2026-03-25 23:48:07,655][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 18 and human policies 1. +[2026-03-25 23:48:07,655][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:48:24,015][__main__][INFO] - Number of regex retries in iteration 186: 0 +[2026-03-25 23:48:24,016][__main__][INFO] - agents played in iteration 186 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:48:24,792][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:48:24,815][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:48:24,838][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:48:24,861][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:48:24,862][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:48:24,862][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:48:25,723][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:48:26,188][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:48:26,715][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:48:27,227][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:48:27,745][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:48:28,256][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:48:28,760][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:48:29,263][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:48:29,763][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:48:30,263][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:48:30,763][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:48:31,264][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:48:31,768][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:48:32,276][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:48:32,781][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:48:33,294][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:48:33,810][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:48:34,325][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:48:34,842][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:48:35,354][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:48:35,871][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:48:36,377][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:48:36,880][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:48:37,381][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:48:37,883][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:48:38,385][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:48:38,888][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:48:39,394][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:48:39,899][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:48:40,403][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:48:40,905][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:48:41,409][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:48:41,912][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:48:42,421][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:48:42,930][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:48:43,438][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:48:43,941][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:48:44,450][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:48:44,967][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:48:45,474][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:48:45,991][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:48:46,504][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:48:47,022][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:48:47,529][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:48:48,032][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:48:48,534][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:48:49,035][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:48:49,535][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:48:50,034][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:48:50,533][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:48:51,038][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:48:51,549][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:48:52,058][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:48:52,557][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:48:53,059][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:48:53,575][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:48:54,083][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:48:54,599][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:48:55,109][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:48:55,625][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:48:56,131][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:48:56,632][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:48:57,132][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:48:57,630][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:48:58,128][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10683 tokens. +[2026-03-25 23:48:58,773][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 60.60%, Block Peak % of device VRAM: 62.40%, ΔTime: 00:00:33 +[2026-03-25 23:48:59,530][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:48:59,532][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:48:59,535][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:49:00,335][__main__][INFO] - Iteration 187 took 52s (31.06% Gen, 67.42% Train). Generation: 16s, Training: 35s. Estimated remaining time: 41h 10m 41s. Estimated total time: 43h 54m 1s. Time estimates for 10 more iterations: 8m 46s, 100 more iterations: 1h 27m 48s, 500 more iterations: 7h 19m 0s. +[2026-03-25 23:49:00,337][__main__][INFO] - Starting iteration 187. +[2026-03-25 23:49:00,742][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 18 and human policies 1. +[2026-03-25 23:49:00,743][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:49:18,378][__main__][INFO] - Number of regex retries in iteration 187: 0 +[2026-03-25 23:49:18,379][__main__][INFO] - agents played in iteration 187 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:49:19,267][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:49:19,292][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:49:19,319][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:49:19,342][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:49:19,343][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:49:19,344][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:49:19,958][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:49:20,412][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:49:20,937][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:49:21,446][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:49:21,970][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:49:22,478][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:49:22,999][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:49:23,506][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:49:24,009][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:49:24,508][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:49:25,006][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:49:25,504][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:49:26,002][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:49:26,500][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:49:27,003][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:49:27,508][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:49:28,013][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:49:28,519][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:49:29,017][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:49:29,516][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:49:30,034][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:49:30,541][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:49:31,058][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:49:31,567][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:49:32,080][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:49:32,585][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:49:33,089][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:49:33,590][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:49:34,089][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:49:34,587][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:49:35,082][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:49:35,582][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:49:36,083][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:49:36,587][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:49:37,092][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:49:37,602][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:49:38,101][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:49:38,600][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:49:39,116][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:49:39,624][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:49:40,142][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:49:40,656][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:49:41,176][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:49:41,686][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:49:42,192][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:49:42,692][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:49:43,187][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:49:43,682][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:49:44,182][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:49:44,682][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:49:45,184][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:49:45,689][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:49:46,200][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:49:46,712][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:49:47,215][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:49:47,714][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:49:48,228][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:49:48,739][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:49:49,255][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:49:49,767][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:49:50,286][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:49:50,796][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:49:51,303][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:49:51,805][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:49:52,306][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10698 tokens. +[2026-03-25 23:49:52,961][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 60.60%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:33 +[2026-03-25 23:49:53,710][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:49:53,712][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:49:53,714][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:49:54,581][__main__][INFO] - Iteration 188 took 53s (32.76% Gen, 65.63% Train). Generation: 17s, Training: 35s. Estimated remaining time: 42h 7m 46s. Estimated total time: 44h 52m 0s. Time estimates for 10 more iterations: 8m 58s, 100 more iterations: 1h 29m 44s, 500 more iterations: 7h 28m 40s. +[2026-03-25 23:49:54,583][__main__][INFO] - Starting iteration 188. +[2026-03-25 23:49:54,995][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 18 and human policies 1. +[2026-03-25 23:49:54,996][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:50:08,948][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 23:50:11,712][__main__][INFO] - Number of regex retries in iteration 188: 1 +[2026-03-25 23:50:11,712][__main__][INFO] - agents played in iteration 188 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:50:12,504][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:50:12,528][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:50:12,552][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:50:12,575][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:50:12,576][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:50:12,576][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:50:13,226][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:50:13,693][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:50:14,209][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:50:14,718][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:50:15,220][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:50:15,725][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:50:16,243][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:50:16,752][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:50:17,269][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:50:17,784][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:50:18,303][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:50:18,813][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:50:19,316][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:50:19,817][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:50:20,319][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:50:20,820][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:50:21,324][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:50:21,823][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:50:22,330][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:50:22,842][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:50:23,352][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:50:23,859][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:50:24,361][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:50:24,864][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:50:25,382][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:50:25,887][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:50:26,402][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:50:26,914][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:50:27,430][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:50:27,938][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:50:28,439][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:50:28,938][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:50:29,441][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:50:29,945][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:50:30,450][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:50:30,957][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:50:31,466][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:50:31,977][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:50:32,484][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:50:32,995][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:50:33,496][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:50:33,996][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:50:34,512][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:50:35,022][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:50:35,539][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:50:36,044][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:50:36,557][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:50:37,063][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:50:37,561][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:50:38,061][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:50:38,560][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:50:39,060][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:50:39,559][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:50:40,061][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:50:40,564][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:50:41,070][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:50:41,575][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:50:42,086][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:50:42,587][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:50:43,086][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:50:43,604][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:50:44,114][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:50:44,629][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:50:45,136][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:50:45,654][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10759 tokens. +[2026-03-25 23:50:46,410][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 60.63%, Block Peak % of device VRAM: 62.41%, ΔTime: 00:00:33 +[2026-03-25 23:50:47,151][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:50:47,153][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:50:47,155][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:50:48,001][__main__][INFO] - Iteration 189 took 53s (31.54% Gen, 66.86% Train). Generation: 16s, Training: 35s. Estimated remaining time: 41h 25m 13s. Estimated total time: 44h 10m 21s. Time estimates for 10 more iterations: 8m 50s, 100 more iterations: 1h 28m 20s, 500 more iterations: 7h 21m 43s. +[2026-03-25 23:50:48,004][__main__][INFO] - Starting iteration 189. +[2026-03-25 23:50:48,407][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 18 and human policies 1. +[2026-03-25 23:50:48,408][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:51:04,857][__main__][INFO] - Number of regex retries in iteration 189: 0 +[2026-03-25 23:51:04,858][__main__][INFO] - agents played in iteration 189 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:51:05,642][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:51:05,665][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:51:05,689][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:51:05,712][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:51:05,712][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:51:05,713][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:51:06,335][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:51:06,789][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:51:07,296][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:51:07,797][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:51:08,302][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:51:08,811][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:51:09,316][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:51:09,828][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:51:10,328][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:51:10,824][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:51:11,344][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:51:11,857][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:51:12,376][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:51:12,903][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:51:13,424][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:51:13,929][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:51:14,433][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:51:14,935][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:51:15,434][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:51:15,933][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:51:16,433][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:51:16,934][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:51:17,438][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:51:17,944][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:51:18,448][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:51:18,960][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:51:19,461][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:51:19,962][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:51:20,481][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:51:20,990][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:51:21,508][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:51:22,015][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:51:22,532][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:51:23,039][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:51:23,541][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:51:24,042][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:51:24,538][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:51:25,037][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:51:25,536][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:51:26,035][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:51:26,538][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:51:27,044][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:51:27,549][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:51:28,060][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:51:28,560][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:51:29,059][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:51:29,575][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:51:30,089][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:51:30,608][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:51:31,119][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:51:31,637][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:51:32,149][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:51:32,655][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:51:33,157][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:51:33,658][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:51:34,158][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:51:34,657][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:51:35,153][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:51:35,653][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:51:36,157][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:51:36,665][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:51:37,173][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:51:37,674][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:51:38,174][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:51:38,684][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10714 tokens. +[2026-03-25 23:51:39,565][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 60.60%, Block Peak % of device VRAM: 62.35%, ΔTime: 00:00:33 +[2026-03-25 23:51:40,420][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:51:40,423][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:51:40,425][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:51:41,193][__main__][INFO] - Iteration 190 took 52s (31.16% Gen, 67.38% Train). Generation: 16s, Training: 35s. Estimated remaining time: 41h 13m 20s. Estimated total time: 43h 59m 21s. Time estimates for 10 more iterations: 8m 47s, 100 more iterations: 1h 27m 58s, 500 more iterations: 7h 19m 53s. +[2026-03-25 23:51:41,196][__main__][INFO] - Starting iteration 190. +[2026-03-25 23:51:41,600][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 18 and human policies 1. +[2026-03-25 23:51:41,600][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:51:45,858][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 23:51:47,039][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 23:51:51,070][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 23:51:59,266][__main__][INFO] - Number of regex retries in iteration 190: 3 +[2026-03-25 23:51:59,267][__main__][INFO] - agents played in iteration 190 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:52:00,310][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:52:00,334][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:52:00,359][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:52:00,384][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:52:00,386][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:52:00,386][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:52:01,075][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:52:01,533][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:52:02,037][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:52:02,543][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:52:03,045][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:52:03,553][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:52:04,058][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:52:04,567][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:52:05,077][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:52:05,587][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:52:06,084][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:52:06,590][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:52:07,108][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:52:07,622][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:52:08,140][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:52:08,659][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:52:09,175][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:52:09,680][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:52:10,184][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:52:10,683][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:52:11,182][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:52:11,683][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:52:12,182][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:52:12,683][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:52:13,184][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:52:13,695][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:52:14,207][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:52:14,717][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:52:15,220][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:52:15,728][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:52:16,246][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:52:16,756][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:52:17,274][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:52:17,789][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:52:18,303][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:52:18,811][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:52:19,314][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:52:19,815][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:52:20,314][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:52:20,813][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:52:21,312][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:52:21,812][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:52:22,315][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:52:22,825][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:52:23,331][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:52:23,842][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:52:24,342][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:52:24,840][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:52:25,358][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:52:25,865][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:52:26,381][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:52:26,887][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:52:27,399][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:52:27,904][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:52:28,404][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:52:28,900][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:52:29,399][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:52:29,897][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:52:30,397][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:52:30,897][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:52:31,401][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:52:31,908][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:52:32,412][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:52:32,923][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:52:33,424][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10764 tokens. +[2026-03-25 23:52:34,194][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 60.64%, Block Peak % of device VRAM: 62.34%, ΔTime: 00:00:33 +[2026-03-25 23:52:35,045][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:52:35,048][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:52:35,050][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:52:36,559][__main__][INFO] - Iteration 191 took 54s (32.15% Gen, 65.11% Train). Generation: 17s, Training: 35s. Estimated remaining time: 43h 1m 5s. Estimated total time: 45h 48m 1s. Time estimates for 10 more iterations: 9m 9s, 100 more iterations: 1h 31m 36s, 500 more iterations: 7h 38m 0s. +[2026-03-25 23:52:36,563][__main__][INFO] - Starting iteration 191. +[2026-03-25 23:52:36,969][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 19 and human policies 1. +[2026-03-25 23:52:36,970][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:52:41,592][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 23:52:55,749][__main__][INFO] - Number of regex retries in iteration 191: 1 +[2026-03-25 23:52:55,749][__main__][INFO] - agents played in iteration 191 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:52:56,538][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:52:56,563][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:52:56,586][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:52:56,609][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:52:56,610][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:52:56,611][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:52:57,238][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:52:57,695][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:52:58,197][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:52:58,698][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:52:59,203][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:52:59,709][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:53:00,218][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:53:00,728][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:53:01,235][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:53:01,737][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:53:02,258][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:53:02,768][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:53:03,288][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:53:03,797][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:53:04,320][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:53:04,827][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:53:05,331][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:53:05,835][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:53:06,336][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:53:06,835][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:53:07,336][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:53:07,837][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:53:08,340][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:53:08,845][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:53:09,351][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:53:09,862][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:53:10,364][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:53:10,864][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:53:11,380][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:53:11,891][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:53:12,412][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:53:12,923][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:53:13,444][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:53:13,951][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:53:14,451][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:53:14,955][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:53:15,456][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:53:15,957][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:53:16,454][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:53:16,956][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:53:17,461][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:53:17,966][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:53:18,475][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:53:18,987][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:53:19,488][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:53:20,007][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:53:20,532][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:53:21,042][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:53:21,561][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:53:22,069][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:53:22,586][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:53:23,101][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:53:23,604][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:53:24,107][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:53:24,606][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:53:25,105][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:53:25,602][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:53:26,104][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:53:26,606][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:53:27,112][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:53:27,619][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:53:28,128][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:53:28,633][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:53:29,135][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:53:29,654][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10692 tokens. +[2026-03-25 23:53:30,558][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 60.61%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:33 +[2026-03-25 23:53:31,424][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:53:31,427][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:53:31,430][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:53:32,229][__main__][INFO] - Iteration 192 took 55s (33.98% Gen, 64.57% Train). Generation: 18s, Training: 35s. Estimated remaining time: 43h 15m 9s. Estimated total time: 46h 3m 1s. Time estimates for 10 more iterations: 9m 12s, 100 more iterations: 1h 32m 6s, 500 more iterations: 7h 40m 30s. +[2026-03-25 23:53:32,232][__main__][INFO] - Starting iteration 192. +[2026-03-25 23:53:32,635][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 19 and human policies 1. +[2026-03-25 23:53:32,636][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:53:48,866][__main__][INFO] - Number of regex retries in iteration 192: 0 +[2026-03-25 23:53:48,867][__main__][INFO] - agents played in iteration 192 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:53:49,792][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:53:49,816][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:53:49,839][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:53:49,863][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:53:49,864][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:53:49,865][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:53:50,764][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:53:51,233][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:53:51,739][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:53:52,241][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:53:52,743][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:53:53,245][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:53:53,749][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:53:54,252][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:53:54,758][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:53:55,269][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:53:55,779][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:53:56,289][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:53:56,795][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:53:57,303][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:53:57,820][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:53:58,327][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:53:58,846][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:53:59,362][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:53:59,878][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:54:00,387][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:54:00,891][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:54:01,390][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:54:01,891][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:54:02,390][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:54:02,890][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:54:03,388][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:54:03,891][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:54:04,401][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:54:04,908][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:54:05,413][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:54:05,911][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:54:06,415][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:54:06,930][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:54:07,440][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:54:07,959][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:54:08,473][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:54:08,992][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:54:09,500][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:54:10,011][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:54:10,513][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:54:11,014][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:54:11,513][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:54:12,016][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:54:12,517][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:54:13,023][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:54:13,531][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:54:14,036][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:54:14,545][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:54:15,045][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:54:15,570][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:54:16,093][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:54:16,599][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:54:17,116][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:54:17,631][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:54:18,148][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:54:18,658][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:54:19,163][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:54:19,663][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:54:20,160][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:54:20,659][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:54:21,157][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:54:21,658][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:54:22,163][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:54:22,675][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:54:23,182][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10732 tokens. +[2026-03-25 23:54:23,902][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 60.61%, Block Peak % of device VRAM: 62.40%, ΔTime: 00:00:33 +[2026-03-25 23:54:24,671][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:54:24,673][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:54:24,675][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:54:25,507][__main__][INFO] - Iteration 193 took 52s (30.70% Gen, 67.72% Train). Generation: 16s, Training: 35s. Estimated remaining time: 41h 14m 52s. Estimated total time: 44h 3m 38s. Time estimates for 10 more iterations: 8m 48s, 100 more iterations: 1h 28m 7s, 500 more iterations: 7h 20m 36s. +[2026-03-25 23:54:25,509][__main__][INFO] - Starting iteration 193. +[2026-03-25 23:54:25,922][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 19 and human policies 1. +[2026-03-25 23:54:25,923][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:54:43,184][__main__][INFO] - Number of regex retries in iteration 193: 0 +[2026-03-25 23:54:43,185][__main__][INFO] - agents played in iteration 193 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:54:44,108][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:54:44,132][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:54:44,155][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:54:44,179][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:54:44,181][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:54:44,182][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:54:45,077][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:54:45,544][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:54:46,061][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:54:46,565][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:54:47,071][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:54:47,574][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:54:48,072][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:54:48,570][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:54:49,068][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:54:49,571][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:54:50,077][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:54:50,588][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:54:51,098][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:54:51,604][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:54:52,103][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:54:52,616][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:54:53,133][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:54:53,654][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:54:54,167][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:54:54,685][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:54:55,205][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:54:55,714][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:54:56,215][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:54:56,720][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:54:57,221][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:54:57,723][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:54:58,225][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:54:58,726][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:54:59,231][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:54:59,742][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:55:00,250][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:55:00,767][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:55:01,280][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:55:01,797][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:55:02,310][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:55:02,830][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:55:03,341][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:55:03,844][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:55:04,343][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:55:04,843][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:55:05,344][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:55:05,847][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:55:06,347][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:55:06,856][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:55:07,359][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:55:07,862][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:55:08,366][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:55:08,867][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:55:09,369][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:55:09,882][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:55:10,386][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:55:10,886][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:55:11,383][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:55:11,900][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:55:12,409][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:55:12,928][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:55:13,437][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:55:13,950][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:55:14,460][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:55:14,965][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:55:15,475][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:55:15,977][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:55:16,479][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:55:16,981][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:55:17,485][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10584 tokens. +[2026-03-25 23:55:18,198][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.25%, Current % of VRAM taken: 60.68%, Block Peak % of device VRAM: 62.22%, ΔTime: 00:00:33 +[2026-03-25 23:55:18,993][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:55:18,995][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:55:18,997][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:55:19,726][__main__][INFO] - Iteration 194 took 53s (32.08% Gen, 66.56% Train). Generation: 17s, Training: 35s. Estimated remaining time: 42h 0m 35s. Estimated total time: 44h 50m 15s. Time estimates for 10 more iterations: 8m 58s, 100 more iterations: 1h 29m 40s, 500 more iterations: 7h 28m 22s. +[2026-03-25 23:55:19,728][__main__][INFO] - Starting iteration 194. +[2026-03-25 23:55:20,129][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 19 and human policies 1. +[2026-03-25 23:55:20,130][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:55:24,405][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 23:55:37,718][__main__][INFO] - Number of regex retries in iteration 194: 1 +[2026-03-25 23:55:37,719][__main__][INFO] - agents played in iteration 194 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:55:38,497][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:55:38,520][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:55:38,544][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:55:38,567][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:55:38,567][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:55:38,568][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:55:39,464][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:55:39,921][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:55:40,450][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:55:40,958][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:55:41,477][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:55:41,984][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:55:42,487][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:55:42,986][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:55:43,492][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:55:43,991][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:55:44,492][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:55:44,994][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:55:45,498][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:55:46,004][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:55:46,509][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:55:47,018][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:55:47,520][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:55:48,023][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:55:48,544][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:55:49,053][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:55:49,569][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:55:50,076][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:55:50,594][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:55:51,099][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:55:51,599][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:55:52,103][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:55:52,605][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:55:53,105][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:55:53,605][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:55:54,103][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:55:54,600][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:55:55,103][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:55:55,602][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:55:56,100][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:55:56,597][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:55:57,095][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:55:57,593][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:55:58,092][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:55:58,589][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:55:59,084][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:55:59,582][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:56:00,100][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:56:00,602][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:56:01,102][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:56:01,601][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:56:02,101][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:56:02,600][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:56:03,097][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:56:03,599][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:56:04,104][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:56:04,602][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:56:05,102][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:56:05,601][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:56:06,100][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:56:06,600][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:56:07,100][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:56:07,599][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:56:08,097][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:56:08,592][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:56:09,091][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:56:09,591][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:56:10,088][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:56:10,583][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:56:11,082][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:56:11,580][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10688 tokens. +[2026-03-25 23:56:12,228][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 60.62%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 +[2026-03-25 23:56:12,981][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:56:12,983][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:56:12,985][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:56:13,721][__main__][INFO] - Iteration 195 took 53s (32.82% Gen, 65.80% Train). Generation: 17s, Training: 35s. Estimated remaining time: 41h 49m 2s. Estimated total time: 44h 39m 35s. Time estimates for 10 more iterations: 8m 55s, 100 more iterations: 1h 29m 19s, 500 more iterations: 7h 26m 35s. +[2026-03-25 23:56:13,723][__main__][INFO] - Starting iteration 195. +[2026-03-25 23:56:14,121][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 19 and human policies 1. +[2026-03-25 23:56:14,122][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:56:19,673][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 30 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 23:56:30,571][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 23:56:32,005][__main__][INFO] - Number of regex retries in iteration 195: 2 +[2026-03-25 23:56:32,006][__main__][INFO] - agents played in iteration 195 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:56:32,806][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:56:32,830][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:56:32,853][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:56:32,876][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:56:32,876][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:56:32,877][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:56:33,497][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:56:33,953][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:56:34,459][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:56:34,967][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:56:35,469][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:56:35,973][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:56:36,475][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:56:36,979][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:56:37,483][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:56:37,987][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:56:38,489][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:56:38,993][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:56:39,497][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:56:40,004][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:56:40,509][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:56:41,012][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:56:41,517][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:56:42,023][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:56:42,531][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:56:43,037][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:56:43,540][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:56:44,044][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:56:44,547][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:56:45,050][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:56:45,557][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:56:46,060][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:56:46,562][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:56:47,064][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:56:47,566][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:56:48,067][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:56:48,569][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:56:49,070][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:56:49,574][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:56:50,080][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:56:50,586][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:56:51,098][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:56:51,603][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:56:52,108][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:56:52,613][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:56:53,118][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:56:53,622][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:56:54,125][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:56:54,629][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:56:55,129][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:56:55,633][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:56:56,137][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:56:56,640][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:56:57,145][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:56:57,646][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:56:58,153][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:56:58,659][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:56:59,165][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:56:59,670][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:57:00,175][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:57:00,680][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:57:01,185][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:57:01,688][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:57:02,192][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:57:02,696][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:57:03,200][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:57:03,703][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:57:04,208][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:57:04,713][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:57:05,218][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:57:05,722][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10660 tokens. +[2026-03-25 23:57:06,419][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 60.63%, Block Peak % of device VRAM: 62.33%, ΔTime: 00:00:32 +[2026-03-25 23:57:07,174][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:57:07,176][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:57:07,178][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:57:07,906][__main__][INFO] - Iteration 196 took 53s (33.25% Gen, 65.39% Train). Generation: 17s, Training: 35s. Estimated remaining time: 41h 57m 47s. Estimated total time: 44h 49m 15s. Time estimates for 10 more iterations: 8m 57s, 100 more iterations: 1h 29m 38s, 500 more iterations: 7h 28m 12s. +[2026-03-25 23:57:07,908][__main__][INFO] - Starting iteration 196. +[2026-03-25 23:57:08,311][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 19 and human policies 1. +[2026-03-25 23:57:08,312][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:57:11,947][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 23:57:24,892][__main__][INFO] - Number of regex retries in iteration 196: 1 +[2026-03-25 23:57:24,893][__main__][INFO] - agents played in iteration 196 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:57:25,669][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:57:25,692][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:57:25,716][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:57:25,739][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:57:25,739][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:57:25,740][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:57:26,759][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:57:27,213][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:57:27,724][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:57:28,234][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:57:28,739][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:57:29,243][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:57:29,744][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:57:30,250][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:57:30,755][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:57:31,259][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:57:31,766][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:57:32,276][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:57:32,780][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:57:33,284][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:57:33,787][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:57:34,288][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:57:34,787][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:57:35,283][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:57:35,778][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:57:36,278][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:57:36,776][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:57:37,271][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:57:37,769][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:57:38,265][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:57:38,762][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:57:39,281][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:57:39,783][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:57:40,281][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:57:40,777][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:57:41,277][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:57:41,776][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:57:42,276][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:57:42,775][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:57:43,277][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:57:43,776][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:57:44,272][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:57:44,773][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:57:45,273][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:57:45,772][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:57:46,271][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:57:46,770][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:57:47,275][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:57:47,776][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:57:48,276][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:57:48,772][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:57:49,271][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:57:49,767][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:57:50,265][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:57:50,764][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:57:51,263][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:57:51,763][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:57:52,261][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:57:52,760][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:57:53,258][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:57:53,758][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:57:54,257][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:57:54,756][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:57:55,263][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:57:55,760][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:57:56,258][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:57:56,757][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:57:57,255][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:57:57,756][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:57:58,255][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:57:58,750][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10647 tokens. +[2026-03-25 23:57:59,422][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.58%, Block Peak % of device VRAM: 62.36%, ΔTime: 00:00:32 +[2026-03-25 23:58:00,154][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:58:00,156][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:58:00,158][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:58:00,993][__main__][INFO] - Iteration 197 took 52s (31.47% Gen, 66.94% Train). Generation: 16s, Training: 35s. Estimated remaining time: 41h 1m 47s. Estimated total time: 43h 54m 8s. Time estimates for 10 more iterations: 8m 46s, 100 more iterations: 1h 27m 48s, 500 more iterations: 7h 19m 1s. +[2026-03-25 23:58:00,995][__main__][INFO] - Starting iteration 197. +[2026-03-25 23:58:01,394][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 19 and human policies 1. +[2026-03-25 23:58:01,395][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:58:05,389][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 23:58:16,645][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-25 23:58:18,093][__main__][INFO] - Number of regex retries in iteration 197: 2 +[2026-03-25 23:58:18,094][__main__][INFO] - agents played in iteration 197 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:58:18,866][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:58:18,889][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:58:18,912][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:58:18,935][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:58:18,936][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:58:18,936][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:58:19,556][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:58:20,018][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:58:20,523][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:58:21,025][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:58:21,524][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:58:22,024][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:58:22,524][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:58:23,024][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:58:23,525][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:58:24,023][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:58:24,522][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:58:25,021][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:58:25,528][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:58:26,027][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:58:26,527][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:58:27,025][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:58:27,524][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:58:28,024][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:58:28,525][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:58:29,024][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:58:29,521][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:58:30,021][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:58:30,521][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:58:31,020][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:58:31,518][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:58:32,016][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:58:32,512][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:58:33,008][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:58:33,506][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:58:34,004][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:58:34,503][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:58:35,001][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:58:35,498][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:58:35,996][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:58:36,497][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:58:36,997][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:58:37,494][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:58:37,993][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:58:38,487][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:58:38,985][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:58:39,481][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:58:39,980][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:58:40,481][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:58:40,981][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:58:41,479][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:58:41,977][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:58:42,476][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:58:42,975][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:58:43,477][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:58:43,977][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:58:44,477][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:58:44,977][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:58:45,476][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:58:45,977][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:58:46,476][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:58:46,976][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:58:47,475][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:58:47,975][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:58:48,474][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:58:48,972][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:58:49,469][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:58:49,969][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:58:50,467][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:58:50,967][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:58:51,464][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10691 tokens. +[2026-03-25 23:58:52,116][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.58%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 +[2026-03-25 23:58:52,847][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:58:52,849][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:58:52,851][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:58:53,711][__main__][INFO] - Iteration 198 took 52s (31.92% Gen, 66.43% Train). Generation: 16s, Training: 34s. Estimated remaining time: 40h 42m 38s. Estimated total time: 43h 35m 51s. Time estimates for 10 more iterations: 8m 43s, 100 more iterations: 1h 27m 11s, 500 more iterations: 7h 15m 58s. +[2026-03-25 23:58:53,713][__main__][INFO] - Starting iteration 198. +[2026-03-25 23:58:54,112][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 19 and human policies 1. +[2026-03-25 23:58:54,112][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-25 23:59:10,315][__main__][INFO] - Number of regex retries in iteration 198: 0 +[2026-03-25 23:59:10,316][__main__][INFO] - agents played in iteration 198 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-25 23:59:11,161][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:59:11,185][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:59:11,208][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:59:11,231][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-25 23:59:11,231][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-25 23:59:11,232][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-25 23:59:11,852][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-25 23:59:12,313][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-25 23:59:12,816][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-25 23:59:13,317][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-25 23:59:13,816][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-25 23:59:14,316][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-25 23:59:14,815][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-25 23:59:15,315][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-25 23:59:15,817][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-25 23:59:16,316][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-25 23:59:16,815][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-25 23:59:17,314][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-25 23:59:17,813][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-25 23:59:18,312][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-25 23:59:18,811][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-25 23:59:19,311][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-25 23:59:19,811][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-25 23:59:20,311][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-25 23:59:20,811][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-25 23:59:21,314][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-25 23:59:21,816][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-25 23:59:22,316][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-25 23:59:22,817][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-25 23:59:23,317][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-25 23:59:23,818][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-25 23:59:24,318][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-25 23:59:24,817][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-25 23:59:25,313][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-25 23:59:25,812][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-25 23:59:26,314][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-25 23:59:26,815][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-25 23:59:27,316][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-25 23:59:27,815][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-25 23:59:28,313][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-25 23:59:28,812][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-25 23:59:29,308][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-25 23:59:29,805][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-25 23:59:30,304][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-25 23:59:30,803][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-25 23:59:31,301][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-25 23:59:31,802][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-25 23:59:32,310][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-25 23:59:32,810][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-25 23:59:33,310][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-25 23:59:33,808][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-25 23:59:34,307][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-25 23:59:34,805][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-25 23:59:35,303][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-25 23:59:35,803][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-25 23:59:36,301][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-25 23:59:36,800][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-25 23:59:37,298][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-25 23:59:37,796][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-25 23:59:38,296][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-25 23:59:38,796][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-25 23:59:39,294][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-25 23:59:39,792][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-25 23:59:40,291][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-25 23:59:40,788][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-25 23:59:41,288][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-25 23:59:41,788][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-25 23:59:42,285][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-25 23:59:42,785][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-25 23:59:43,283][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-25 23:59:43,783][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10691 tokens. +[2026-03-25 23:59:44,457][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.27%, Current % of VRAM taken: 60.69%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:32 +[2026-03-25 23:59:45,193][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-25 23:59:45,195][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-25 23:59:45,197][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-25 23:59:46,030][__main__][INFO] - Iteration 199 took 51s (31.21% Gen, 67.18% Train). Generation: 16s, Training: 34s. Estimated remaining time: 40h 21m 51s. Estimated total time: 43h 15m 56s. Time estimates for 10 more iterations: 8m 39s, 100 more iterations: 1h 26m 31s, 500 more iterations: 7h 12m 39s. +[2026-03-25 23:59:46,032][__main__][INFO] - Starting iteration 199. +[2026-03-25 23:59:46,432][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 19 and human policies 1. +[2026-03-25 23:59:46,432][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:00:02,228][__main__][INFO] - Number of regex retries in iteration 199: 0 +[2026-03-26 00:00:02,229][__main__][INFO] - agents played in iteration 199 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:00:03,015][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:00:03,038][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:00:03,061][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:00:03,084][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:00:03,085][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:00:03,085][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:00:03,706][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:00:04,158][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:00:04,664][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:00:05,161][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:00:05,660][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:00:06,159][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:00:06,660][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:00:07,159][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:00:07,659][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:00:08,158][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:00:08,657][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:00:09,154][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:00:09,650][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:00:10,148][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:00:10,648][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:00:11,146][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:00:11,645][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:00:12,143][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:00:12,641][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:00:13,140][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:00:13,637][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:00:14,136][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:00:14,635][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:00:15,134][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:00:15,632][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:00:16,130][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:00:16,628][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:00:17,129][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:00:17,629][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:00:18,130][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:00:18,629][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:00:19,124][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:00:19,623][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:00:20,120][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:00:20,615][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:00:21,111][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:00:21,607][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:00:22,103][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:00:22,601][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:00:23,098][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:00:23,596][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:00:24,093][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:00:24,591][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:00:25,089][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:00:25,585][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:00:26,082][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:00:26,578][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:00:27,075][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:00:27,575][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:00:28,074][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:00:28,573][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:00:29,069][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:00:29,565][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:00:30,064][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:00:30,562][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:00:31,062][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:00:31,559][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:00:32,059][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:00:32,559][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:00:33,058][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:00:33,557][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:00:34,058][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:00:34,558][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:00:35,057][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:00:35,557][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10617 tokens. +[2026-03-26 00:00:36,209][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 60.62%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 +[2026-03-26 00:00:36,928][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:00:36,930][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:00:36,932][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:00:37,577][__main__][INFO] - Iteration 200 took 51s (30.88% Gen, 67.85% Train). Generation: 15s, Training: 34s. Estimated remaining time: 39h 42m 20s. Estimated total time: 42h 37m 17s. Time estimates for 10 more iterations: 8m 31s, 100 more iterations: 1h 25m 14s, 500 more iterations: 7h 6m 12s. +[2026-03-26 00:00:37,579][__main__][INFO] - Starting iteration 200. +[2026-03-26 00:00:37,983][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 19 and human policies 1. +[2026-03-26 00:00:37,983][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:00:54,975][__main__][INFO] - Number of regex retries in iteration 200: 0 +[2026-03-26 00:00:54,976][__main__][INFO] - agents played in iteration 200 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:00:55,758][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:00:55,781][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:00:55,805][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:00:55,828][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:00:55,828][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:00:55,829][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:00:56,445][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:00:56,900][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:00:57,405][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:00:57,904][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:00:58,403][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:00:58,901][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:00:59,401][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:00:59,901][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:01:00,398][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:01:00,898][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:01:01,400][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:01:01,903][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:01:02,404][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:01:02,903][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:01:03,402][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:01:03,901][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:01:04,401][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:01:04,901][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:01:05,401][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:01:05,900][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:01:06,400][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:01:06,899][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:01:07,398][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:01:07,898][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:01:08,400][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:01:08,899][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:01:09,425][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:01:09,925][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:01:10,424][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:01:10,925][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:01:11,427][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:01:11,923][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:01:12,423][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:01:12,921][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:01:13,419][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:01:13,918][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:01:14,418][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:01:14,918][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:01:15,418][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:01:15,917][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:01:16,415][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:01:16,912][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:01:17,415][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:01:17,911][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:01:18,420][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:01:18,920][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:01:19,419][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:01:19,917][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:01:20,416][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:01:20,917][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:01:21,417][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:01:21,917][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:01:22,416][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:01:22,918][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:01:23,419][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:01:23,920][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:01:24,419][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:01:24,923][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:01:25,423][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:01:25,923][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:01:26,423][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:01:26,924][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:01:27,423][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:01:27,923][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:01:28,423][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10671 tokens. +[2026-03-26 00:01:29,074][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.30%, Current % of VRAM taken: 60.72%, Block Peak % of device VRAM: 62.40%, ΔTime: 00:00:32 +[2026-03-26 00:01:29,807][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:01:29,809][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:01:29,811][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:01:31,260][__main__][INFO] - Iteration 201 took 53s (31.89% Gen, 65.38% Train). Generation: 16s, Training: 34s. Estimated remaining time: 41h 28m 3s. Estimated total time: 44h 23m 54s. Time estimates for 10 more iterations: 8m 52s, 100 more iterations: 1h 28m 47s, 500 more iterations: 7h 23m 59s. +[2026-03-26 00:01:31,263][__main__][INFO] - Starting iteration 201. +[2026-03-26 00:01:31,674][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 20 and human policies 1. +[2026-03-26 00:01:31,675][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:01:47,069][__main__][INFO] - Number of regex retries in iteration 201: 0 +[2026-03-26 00:01:47,070][__main__][INFO] - agents played in iteration 201 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:01:47,847][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:01:47,871][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:01:47,894][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:01:47,917][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:01:47,917][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:01:47,918][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:01:48,531][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:01:48,990][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:01:49,495][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:01:49,994][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:01:50,493][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:01:50,993][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:01:51,496][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:01:51,998][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:01:52,498][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:01:52,995][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:01:53,494][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:01:53,992][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:01:54,490][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:01:54,989][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:01:55,488][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:01:55,984][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:01:56,481][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:01:56,978][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:01:57,476][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:01:57,973][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:01:58,471][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:01:58,971][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:01:59,469][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:01:59,966][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:02:00,464][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:02:00,967][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:02:01,464][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:02:01,966][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:02:02,466][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:02:02,965][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:02:03,465][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:02:03,966][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:02:04,467][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:02:04,966][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:02:05,466][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:02:05,964][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:02:06,465][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:02:06,966][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:02:07,466][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:02:07,965][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:02:08,464][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:02:08,964][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:02:09,464][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:02:09,963][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:02:10,462][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:02:10,962][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:02:11,462][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:02:11,963][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:02:12,463][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:02:12,964][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:02:13,465][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:02:13,965][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:02:14,464][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:02:14,964][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:02:15,461][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:02:15,962][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:02:16,461][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:02:16,961][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:02:17,459][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:02:17,959][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:02:18,460][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:02:18,958][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:02:19,457][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:02:19,957][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:02:20,457][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10653 tokens. +[2026-03-26 00:02:21,104][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 60.62%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:32 +[2026-03-26 00:02:21,840][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:02:21,842][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:02:21,844][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:02:22,556][__main__][INFO] - Iteration 202 took 50s (30.26% Gen, 68.34% Train). Generation: 15s, Training: 34s. Estimated remaining time: 39h 27m 24s. Estimated total time: 42h 24m 7s. Time estimates for 10 more iterations: 8m 28s, 100 more iterations: 1h 24m 48s, 500 more iterations: 7h 4m 1s. +[2026-03-26 00:02:22,558][__main__][INFO] - Starting iteration 202. +[2026-03-26 00:02:22,958][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 20 and human policies 1. +[2026-03-26 00:02:22,958][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:02:29,346][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 00:02:40,526][__main__][INFO] - Number of regex retries in iteration 202: 1 +[2026-03-26 00:02:40,527][__main__][INFO] - agents played in iteration 202 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:02:41,353][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:02:41,376][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:02:41,399][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:02:41,422][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:02:41,422][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:02:41,423][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:02:42,038][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:02:42,495][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:02:43,001][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:02:43,500][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:02:43,999][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:02:44,504][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:02:45,006][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:02:45,508][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:02:46,006][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:02:46,506][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:02:47,005][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:02:47,505][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:02:48,005][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:02:48,504][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:02:49,003][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:02:49,501][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:02:50,019][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:02:50,521][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:02:51,020][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:02:51,516][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:02:52,015][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:02:52,514][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:02:53,012][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:02:53,513][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:02:54,010][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:02:54,507][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:02:55,004][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:02:55,503][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:02:56,001][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:02:56,499][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:02:56,994][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:02:57,492][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:02:57,987][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:02:58,486][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:02:58,985][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:02:59,486][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:02:59,986][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:03:00,486][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:03:00,985][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:03:01,484][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:03:01,984][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:03:02,481][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:03:02,981][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:03:03,480][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:03:03,979][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:03:04,474][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:03:04,973][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:03:05,470][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:03:05,970][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:03:06,472][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:03:06,969][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:03:07,467][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:03:07,966][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:03:08,466][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:03:08,963][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:03:09,462][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:03:09,958][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:03:10,458][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:03:10,958][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:03:11,458][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:03:11,959][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:03:12,459][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:03:12,958][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:03:13,457][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:03:13,958][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10647 tokens. +[2026-03-26 00:03:14,612][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 60.61%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:32 +[2026-03-26 00:03:15,348][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:03:15,350][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:03:15,354][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:03:16,072][__main__][INFO] - Iteration 203 took 53s (33.08% Gen, 65.57% Train). Generation: 17s, Training: 34s. Estimated remaining time: 41h 18m 9s. Estimated total time: 44h 15m 44s. Time estimates for 10 more iterations: 8m 51s, 100 more iterations: 1h 28m 31s, 500 more iterations: 7h 22m 37s. +[2026-03-26 00:03:16,075][__main__][INFO] - Starting iteration 203. +[2026-03-26 00:03:16,484][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 20 and human policies 1. +[2026-03-26 00:03:16,485][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:03:35,845][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 1 y book, 10 z balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 00:03:39,377][__main__][INFO] - Number of regex retries in iteration 203: 1 +[2026-03-26 00:03:39,378][__main__][INFO] - agents played in iteration 203 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:03:40,165][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:03:40,188][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:03:40,212][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:03:40,235][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:03:40,236][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:03:40,236][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:03:40,885][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:03:41,347][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:03:41,863][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:03:42,369][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:03:42,874][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:03:43,375][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:03:43,872][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:03:44,368][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:03:44,866][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:03:45,364][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:03:45,864][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:03:46,365][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:03:46,867][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:03:47,367][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:03:47,864][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:03:48,365][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:03:48,873][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:03:49,376][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:03:49,878][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:03:50,385][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:03:50,886][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:03:51,385][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:03:51,895][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:03:52,398][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:03:52,899][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:03:53,400][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:03:53,901][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:03:54,400][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:03:54,900][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:03:55,409][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:03:55,910][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:03:56,409][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:03:56,910][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:03:57,408][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:03:57,907][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:03:58,404][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:03:58,904][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:03:59,404][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:03:59,902][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:04:00,401][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:04:00,900][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:04:01,406][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:04:01,908][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:04:02,405][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:04:02,905][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:04:03,404][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:04:03,904][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:04:04,404][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:04:04,902][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:04:05,400][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:04:05,900][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:04:06,398][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:04:06,898][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:04:07,405][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:04:07,904][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:04:08,401][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:04:08,901][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:04:09,401][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:04:09,900][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:04:10,399][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:04:10,898][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:04:11,395][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:04:11,894][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:04:12,391][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:04:12,891][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10597 tokens. +[2026-03-26 00:04:13,555][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 60.60%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 +[2026-03-26 00:04:14,305][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:04:14,308][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:04:14,310][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:04:15,144][__main__][INFO] - Iteration 204 took 58s (39.03% Gen, 59.55% Train). Generation: 22s, Training: 34s. Estimated remaining time: 45h 54m 26s. Estimated total time: 48h 53m 1s. Time estimates for 10 more iterations: 9m 46s, 100 more iterations: 1h 37m 46s, 500 more iterations: 8h 8m 50s. +[2026-03-26 00:04:15,146][__main__][INFO] - Starting iteration 204. +[2026-03-26 00:04:15,550][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 20 and human policies 1. +[2026-03-26 00:04:15,551][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:04:19,046][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 00:04:31,741][__main__][INFO] - Number of regex retries in iteration 204: 1 +[2026-03-26 00:04:31,742][__main__][INFO] - agents played in iteration 204 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:04:32,513][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:04:32,537][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:04:32,560][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:04:32,583][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:04:32,583][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:04:32,584][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:04:33,211][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:04:33,662][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:04:34,165][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:04:34,662][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:04:35,161][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:04:35,664][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:04:36,162][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:04:36,665][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:04:37,162][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:04:37,689][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:04:38,189][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:04:38,690][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:04:39,191][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:04:39,690][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:04:40,190][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:04:40,691][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:04:41,192][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:04:41,691][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:04:42,191][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:04:42,691][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:04:43,191][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:04:43,689][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:04:44,188][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:04:44,685][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:04:45,185][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:04:45,689][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:04:46,190][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:04:46,690][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:04:47,190][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:04:47,692][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:04:48,193][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:04:48,694][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:04:49,194][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:04:49,694][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:04:50,193][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:04:50,694][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:04:51,196][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:04:51,696][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:04:52,196][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:04:52,694][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:04:53,194][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:04:53,693][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:04:54,192][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:04:54,693][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:04:55,193][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:04:55,693][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:04:56,191][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:04:56,698][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:04:57,198][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:04:57,697][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:04:58,198][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:04:58,698][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:04:59,199][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:04:59,699][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:05:00,198][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:05:00,701][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:05:01,203][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:05:01,702][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:05:02,203][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:05:02,702][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:05:03,197][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:05:03,694][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:05:04,191][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:05:04,691][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:05:05,190][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10631 tokens. +[2026-03-26 00:05:05,851][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 60.61%, Block Peak % of device VRAM: 62.38%, ΔTime: 00:00:32 +[2026-03-26 00:05:06,604][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:05:06,606][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:05:06,608][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:05:07,326][__main__][INFO] - Iteration 205 took 51s (31.27% Gen, 67.34% Train). Generation: 16s, Training: 34s. Estimated remaining time: 40h 9m 20s. Estimated total time: 43h 8m 47s. Time estimates for 10 more iterations: 8m 37s, 100 more iterations: 1h 26m 17s, 500 more iterations: 7h 11m 27s. +[2026-03-26 00:05:07,328][__main__][INFO] - Starting iteration 205. +[2026-03-26 00:05:07,737][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 20 and human policies 1. +[2026-03-26 00:05:07,738][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:05:24,724][__main__][INFO] - Number of regex retries in iteration 205: 0 +[2026-03-26 00:05:24,725][__main__][INFO] - agents played in iteration 205 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:05:25,529][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:05:25,553][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:05:25,576][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:05:25,599][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:05:25,599][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:05:25,600][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:05:26,221][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:05:26,677][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:05:27,181][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:05:27,679][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:05:28,179][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:05:28,679][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:05:29,181][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:05:29,681][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:05:30,180][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:05:30,677][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:05:31,173][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:05:31,673][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:05:32,173][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:05:32,672][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:05:33,170][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:05:33,665][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:05:34,167][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:05:34,667][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:05:35,167][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:05:35,664][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:05:36,170][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:05:36,666][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:05:37,166][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:05:37,669][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:05:38,168][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:05:38,667][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:05:39,167][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:05:39,663][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:05:40,163][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:05:40,663][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:05:41,161][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:05:41,660][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:05:42,159][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:05:42,659][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:05:43,160][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:05:43,660][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:05:44,159][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:05:44,658][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:05:45,156][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:05:45,657][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:05:46,158][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:05:46,658][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:05:47,157][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:05:47,658][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:05:48,155][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:05:48,658][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:05:49,159][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:05:49,659][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:05:50,159][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:05:50,654][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:05:51,153][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:05:51,652][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:05:52,148][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:05:52,644][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:05:53,140][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:05:53,639][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:05:54,139][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:05:54,635][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:05:55,131][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:05:55,627][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:05:56,124][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:05:56,620][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:05:57,117][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:05:57,615][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:05:58,113][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10630 tokens. +[2026-03-26 00:05:58,768][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 60.53%, Block Peak % of device VRAM: 62.24%, ΔTime: 00:00:32 +[2026-03-26 00:05:59,540][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:05:59,542][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:05:59,544][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:06:00,277][__main__][INFO] - Iteration 206 took 52s (32.33% Gen, 66.27% Train). Generation: 16s, Training: 34s. Estimated remaining time: 40h 46m 40s. Estimated total time: 43h 47m 0s. Time estimates for 10 more iterations: 8m 45s, 100 more iterations: 1h 27m 34s, 500 more iterations: 7h 17m 50s. +[2026-03-26 00:06:00,279][__main__][INFO] - Starting iteration 206. +[2026-03-26 00:06:00,679][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 20 and human policies 1. +[2026-03-26 00:06:00,680][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:06:18,007][__main__][INFO] - Number of regex retries in iteration 206: 0 +[2026-03-26 00:06:18,008][__main__][INFO] - agents played in iteration 206 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:06:18,784][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:06:18,807][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:06:18,831][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:06:18,854][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:06:18,854][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:06:18,855][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:06:19,483][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:06:19,936][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:06:20,440][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:06:20,940][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:06:21,438][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:06:21,938][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:06:22,437][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:06:22,935][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:06:23,434][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:06:23,938][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:06:24,437][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:06:24,938][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:06:25,438][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:06:25,936][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:06:26,431][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:06:26,927][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:06:27,425][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:06:27,924][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:06:28,424][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:06:28,921][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:06:29,420][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:06:29,920][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:06:30,419][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:06:30,917][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:06:31,413][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:06:31,915][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:06:32,416][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:06:32,912][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:06:33,412][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:06:33,909][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:06:34,409][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:06:34,905][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:06:35,404][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:06:35,908][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:06:36,408][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:06:36,931][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:06:37,429][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:06:37,926][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:06:38,422][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:06:38,922][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:06:39,423][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:06:39,920][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:06:40,420][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:06:40,920][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:06:41,419][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:06:41,916][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:06:42,418][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:06:42,917][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:06:43,416][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:06:43,916][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:06:44,419][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:06:44,927][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:06:45,432][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:06:45,937][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:06:46,442][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:06:46,946][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:06:47,451][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:06:47,954][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:06:48,450][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:06:48,949][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:06:49,449][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:06:49,957][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:06:50,461][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:06:50,960][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:06:51,460][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10593 tokens. +[2026-03-26 00:06:52,132][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 60.64%, Block Peak % of device VRAM: 62.37%, ΔTime: 00:00:32 +[2026-03-26 00:06:52,892][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:06:52,894][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:06:52,896][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:06:53,630][__main__][INFO] - Iteration 207 took 52s (32.72% Gen, 65.89% Train). Generation: 17s, Training: 34s. Estimated remaining time: 41h 6m 22s. Estimated total time: 44h 7m 35s. Time estimates for 10 more iterations: 8m 49s, 100 more iterations: 1h 28m 15s, 500 more iterations: 7h 21m 15s. +[2026-03-26 00:06:53,633][__main__][INFO] - Starting iteration 207. +[2026-03-26 00:06:54,033][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 20 and human policies 1. +[2026-03-26 00:06:54,034][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:07:10,991][__main__][INFO] - Number of regex retries in iteration 207: 0 +[2026-03-26 00:07:10,992][__main__][INFO] - agents played in iteration 207 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:07:11,806][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:07:11,830][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:07:11,853][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:07:11,875][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:07:11,876][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:07:11,877][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:07:12,512][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:07:12,968][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:07:13,468][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:07:13,967][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:07:14,464][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:07:14,962][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:07:15,461][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:07:15,960][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:07:16,458][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:07:16,961][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:07:17,461][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:07:17,960][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:07:18,459][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:07:18,961][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:07:19,460][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:07:19,960][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:07:20,460][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:07:20,962][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:07:21,462][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:07:21,961][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:07:22,462][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:07:22,963][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:07:23,464][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:07:23,965][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:07:24,464][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:07:24,964][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:07:25,464][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:07:25,964][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:07:26,464][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:07:26,964][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:07:27,468][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:07:27,967][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:07:28,466][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:07:28,967][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:07:29,466][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:07:29,966][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:07:30,465][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:07:30,965][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:07:31,465][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:07:31,966][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:07:32,466][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:07:32,962][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:07:33,463][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:07:33,962][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:07:34,474][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:07:34,976][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:07:35,477][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:07:35,978][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:07:36,481][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:07:36,979][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:07:37,479][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:07:37,980][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:07:38,478][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:07:38,975][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:07:39,475][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:07:39,975][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:07:40,475][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:07:40,975][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:07:41,501][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:07:42,002][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:07:42,502][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:07:43,002][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:07:43,501][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:07:44,001][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:07:44,502][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10593 tokens. +[2026-03-26 00:07:45,162][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.27%, Current % of VRAM taken: 60.70%, Block Peak % of device VRAM: 62.45%, ΔTime: 00:00:32 +[2026-03-26 00:07:45,934][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:07:45,936][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:07:45,938][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:07:46,674][__main__][INFO] - Iteration 208 took 52s (32.21% Gen, 66.39% Train). Generation: 16s, Training: 34s. Estimated remaining time: 40h 49m 58s. Estimated total time: 43h 52m 5s. Time estimates for 10 more iterations: 8m 46s, 100 more iterations: 1h 27m 44s, 500 more iterations: 7h 18m 40s. +[2026-03-26 00:07:46,676][__main__][INFO] - Starting iteration 208. +[2026-03-26 00:07:47,074][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 20 and human policies 1. +[2026-03-26 00:07:47,075][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:08:03,764][__main__][INFO] - Number of regex retries in iteration 208: 0 +[2026-03-26 00:08:03,765][__main__][INFO] - agents played in iteration 208 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:08:04,544][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:08:04,571][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:08:04,597][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:08:04,622][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:08:04,623][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:08:04,623][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:08:05,252][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:08:05,705][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:08:06,212][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:08:06,713][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:08:07,213][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:08:07,721][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:08:08,217][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:08:08,715][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:08:09,212][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:08:09,713][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:08:10,214][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:08:10,717][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:08:11,218][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:08:11,716][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:08:12,216][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:08:12,718][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:08:13,218][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:08:13,717][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:08:14,213][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:08:14,709][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:08:15,206][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:08:15,702][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:08:16,200][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:08:16,705][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:08:17,203][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:08:17,734][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:08:18,236][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:08:18,747][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:08:19,249][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:08:19,749][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:08:20,248][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:08:20,745][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:08:21,246][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:08:21,745][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:08:22,244][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:08:22,745][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:08:23,242][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:08:23,740][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:08:24,242][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:08:24,742][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:08:25,241][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:08:25,741][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:08:26,241][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:08:26,740][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:08:27,241][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:08:27,740][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:08:28,241][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:08:28,742][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:08:29,242][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:08:29,740][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:08:30,241][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:08:30,737][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:08:31,232][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:08:31,730][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:08:32,228][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:08:32,727][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:08:33,225][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:08:33,726][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:08:34,227][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:08:34,736][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:08:35,238][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:08:35,739][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:08:36,238][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:08:36,739][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:08:37,247][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10581 tokens. +[2026-03-26 00:08:37,912][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 60.63%, Block Peak % of device VRAM: 62.27%, ΔTime: 00:00:32 +[2026-03-26 00:08:38,665][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:08:38,667][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:08:38,669][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:08:39,407][__main__][INFO] - Iteration 209 took 52s (31.89% Gen, 66.69% Train). Generation: 16s, Training: 34s. Estimated remaining time: 40h 33m 40s. Estimated total time: 43h 36m 39s. Time estimates for 10 more iterations: 8m 43s, 100 more iterations: 1h 27m 13s, 500 more iterations: 7h 16m 6s. +[2026-03-26 00:08:39,409][__main__][INFO] - Starting iteration 209. +[2026-03-26 00:08:39,809][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 20 and human policies 1. +[2026-03-26 00:08:39,809][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:08:46,572][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 00:08:57,789][__main__][INFO] - Number of regex retries in iteration 209: 1 +[2026-03-26 00:08:57,790][__main__][INFO] - agents played in iteration 209 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:08:58,576][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:08:58,600][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:08:58,624][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:08:58,647][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:08:58,648][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:08:58,649][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:08:59,281][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:08:59,736][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:09:00,242][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:09:00,743][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:09:01,242][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:09:01,750][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:09:02,252][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:09:02,756][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:09:03,260][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:09:03,761][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:09:04,261][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:09:04,761][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:09:05,262][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:09:05,760][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:09:06,262][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:09:06,764][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:09:07,264][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:09:07,764][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:09:08,273][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:09:08,773][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:09:09,270][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:09:09,768][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:09:10,267][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:09:10,768][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:09:11,265][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:09:11,766][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:09:12,263][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:09:12,762][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:09:13,260][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:09:13,760][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:09:14,264][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:09:14,763][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:09:15,261][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:09:15,756][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:09:16,258][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:09:16,763][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:09:17,260][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:09:17,762][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:09:18,262][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:09:18,763][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:09:19,263][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:09:19,763][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:09:20,263][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:09:20,765][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:09:21,264][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:09:21,762][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:09:22,264][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:09:22,764][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:09:23,263][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:09:23,783][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:09:24,294][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:09:24,795][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:09:25,297][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:09:25,801][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:09:26,301][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:09:26,801][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:09:27,303][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:09:27,809][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:09:28,307][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:09:28,805][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:09:29,304][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:09:29,804][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:09:30,303][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:09:30,802][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:09:31,301][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10601 tokens. +[2026-03-26 00:09:31,980][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 60.63%, Block Peak % of device VRAM: 62.36%, ΔTime: 00:00:32 +[2026-03-26 00:09:32,722][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:09:32,724][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:09:32,725][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:09:33,457][__main__][INFO] - Iteration 210 took 53s (33.52% Gen, 65.12% Train). Generation: 17s, Training: 34s. Estimated remaining time: 41h 38m 34s. Estimated total time: 44h 42m 27s. Time estimates for 10 more iterations: 8m 56s, 100 more iterations: 1h 29m 24s, 500 more iterations: 7h 27m 4s. +[2026-03-26 00:09:33,460][__main__][INFO] - Starting iteration 210. +[2026-03-26 00:09:33,868][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 20 and human policies 1. +[2026-03-26 00:09:33,868][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:09:38,058][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 00:09:46,246][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 00:09:51,129][__main__][INFO] - Number of regex retries in iteration 210: 2 +[2026-03-26 00:09:51,130][__main__][INFO] - agents played in iteration 210 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:09:51,923][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:09:51,946][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:09:51,970][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:09:51,994][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:09:51,995][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:09:51,995][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:09:52,629][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:09:53,084][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:09:53,595][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:09:54,102][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:09:54,602][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:09:55,100][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:09:55,604][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:09:56,115][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:09:56,616][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:09:57,117][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:09:57,618][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:09:58,117][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:09:58,617][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:09:59,120][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:09:59,620][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:10:00,119][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:10:00,616][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:10:01,117][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:10:01,617][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:10:02,119][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:10:02,618][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:10:03,118][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:10:03,618][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:10:04,116][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:10:04,615][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:10:05,111][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:10:05,606][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:10:06,104][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:10:06,603][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:10:07,103][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:10:07,600][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:10:08,098][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:10:08,598][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:10:09,096][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:10:09,596][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:10:10,096][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:10:10,598][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:10:11,095][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:10:11,595][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:10:12,095][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:10:12,590][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:10:13,090][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:10:13,585][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:10:14,082][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:10:14,581][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:10:15,080][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:10:15,581][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:10:16,081][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:10:16,580][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:10:17,080][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:10:17,577][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:10:18,078][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:10:18,580][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:10:19,082][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:10:19,583][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:10:20,085][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:10:20,583][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:10:21,084][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:10:21,583][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:10:22,083][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:10:22,583][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:10:23,082][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:10:23,582][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:10:24,082][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:10:24,583][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10538 tokens. +[2026-03-26 00:10:25,244][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 60.64%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 +[2026-03-26 00:10:25,995][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:10:25,997][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:10:25,999][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:10:27,315][__main__][INFO] - Iteration 211 took 53s (32.29% Gen, 65.24% Train). Generation: 17s, Training: 34s. Estimated remaining time: 41h 27m 37s. Estimated total time: 44h 32m 24s. Time estimates for 10 more iterations: 8m 54s, 100 more iterations: 1h 29m 4s, 500 more iterations: 7h 25m 24s. +[2026-03-26 00:10:27,319][__main__][INFO] - Starting iteration 211. +[2026-03-26 00:10:27,722][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 21 and human policies 1. +[2026-03-26 00:10:27,723][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:10:31,710][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 00:10:45,409][__main__][INFO] - Number of regex retries in iteration 211: 1 +[2026-03-26 00:10:45,410][__main__][INFO] - agents played in iteration 211 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:10:46,190][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:10:46,213][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:10:46,237][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:10:46,260][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:10:46,261][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:10:46,261][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:10:46,902][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:10:47,364][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:10:47,871][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:10:48,370][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:10:48,866][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:10:49,370][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:10:49,870][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:10:50,369][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:10:50,869][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:10:51,379][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:10:51,879][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:10:52,379][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:10:52,880][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:10:53,378][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:10:53,877][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:10:54,385][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:10:54,882][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:10:55,383][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:10:55,883][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:10:56,385][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:10:56,886][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:10:57,386][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:10:57,885][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:10:58,393][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:10:58,892][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:10:59,392][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:10:59,889][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:11:00,390][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:11:00,893][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:11:01,389][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:11:01,890][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:11:02,390][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:11:02,889][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:11:03,399][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:11:03,898][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:11:04,398][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:11:04,894][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:11:05,396][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:11:05,896][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:11:06,398][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:11:06,898][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:11:07,398][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:11:07,897][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:11:08,396][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:11:08,894][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:11:09,391][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:11:09,890][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:11:10,394][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:11:10,894][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:11:11,392][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:11:11,893][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:11:12,390][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:11:12,890][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:11:13,387][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:11:13,885][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:11:14,383][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:11:14,881][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:11:15,377][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:11:15,873][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:11:16,370][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:11:16,867][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:11:17,365][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:11:17,866][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:11:18,362][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:11:18,867][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10476 tokens. +[2026-03-26 00:11:19,548][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 60.53%, Block Peak % of device VRAM: 62.16%, ΔTime: 00:00:32 +[2026-03-26 00:11:20,295][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:11:20,297][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:11:20,299][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:11:21,031][__main__][INFO] - Iteration 212 took 53s (33.18% Gen, 65.45% Train). Generation: 17s, Training: 34s. Estimated remaining time: 41h 19m 46s. Estimated total time: 44h 25m 27s. Time estimates for 10 more iterations: 8m 53s, 100 more iterations: 1h 28m 50s, 500 more iterations: 7h 24m 14s. +[2026-03-26 00:11:21,033][__main__][INFO] - Starting iteration 212. +[2026-03-26 00:11:21,435][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 21 and human policies 1. +[2026-03-26 00:11:21,436][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:11:24,696][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 00:11:35,072][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 00:11:37,794][__main__][INFO] - Number of regex retries in iteration 212: 2 +[2026-03-26 00:11:37,794][__main__][INFO] - agents played in iteration 212 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:11:38,574][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:11:38,598][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:11:38,621][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:11:38,643][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:11:38,644][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:11:38,645][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:11:39,280][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:11:39,734][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:11:40,238][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:11:40,736][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:11:41,236][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:11:41,731][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:11:42,227][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:11:42,723][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:11:43,223][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:11:43,721][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:11:44,217][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:11:44,715][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:11:45,211][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:11:45,706][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:11:46,204][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:11:46,702][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:11:47,201][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:11:47,707][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:11:48,205][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:11:48,704][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:11:49,203][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:11:49,702][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:11:50,202][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:11:50,705][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:11:51,206][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:11:51,705][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:11:52,204][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:11:52,703][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:11:53,203][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:11:53,702][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:11:54,199][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:11:54,699][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:11:55,200][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:11:55,698][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:11:56,224][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:11:56,720][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:11:57,220][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:11:57,719][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:11:58,218][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:11:58,717][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:11:59,220][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:11:59,720][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:12:00,220][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:12:00,721][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:12:01,219][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:12:01,719][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:12:02,216][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:12:02,713][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:12:03,213][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:12:03,713][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:12:04,209][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:12:04,705][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:12:05,201][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:12:05,700][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:12:06,198][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:12:06,699][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:12:07,196][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:12:07,692][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:12:08,187][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:12:08,682][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:12:09,179][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:12:09,677][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:12:10,174][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:12:10,671][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:12:11,168][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10588 tokens. +[2026-03-26 00:12:11,829][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.23%, ΔTime: 00:00:32 +[2026-03-26 00:12:12,576][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:12:12,579][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:12:12,580][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:12:13,389][__main__][INFO] - Iteration 213 took 51s (31.49% Gen, 66.95% Train). Generation: 16s, Training: 34s. Estimated remaining time: 40h 11m 11s. Estimated total time: 43h 17m 44s. Time estimates for 10 more iterations: 8m 39s, 100 more iterations: 1h 26m 35s, 500 more iterations: 7h 12m 57s. +[2026-03-26 00:12:13,391][__main__][INFO] - Starting iteration 213. +[2026-03-26 00:12:13,790][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 21 and human policies 1. +[2026-03-26 00:12:13,791][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:12:31,382][__main__][INFO] - Number of regex retries in iteration 213: 0 +[2026-03-26 00:12:31,383][__main__][INFO] - agents played in iteration 213 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:12:32,167][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:12:32,190][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:12:32,213][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:12:32,236][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:12:32,237][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:12:32,237][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:12:32,869][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:12:33,321][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:12:33,822][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:12:34,320][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:12:34,821][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:12:35,316][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:12:35,813][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:12:36,307][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:12:36,803][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:12:37,303][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:12:37,800][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:12:38,302][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:12:38,801][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:12:39,300][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:12:39,800][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:12:40,298][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:12:40,796][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:12:41,291][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:12:41,788][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:12:42,284][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:12:42,783][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:12:43,280][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:12:43,779][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:12:44,275][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:12:44,773][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:12:45,270][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:12:45,785][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:12:46,285][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:12:46,783][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:12:47,281][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:12:47,781][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:12:48,276][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:12:48,776][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:12:49,271][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:12:49,768][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:12:50,268][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:12:50,768][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:12:51,267][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:12:51,765][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:12:52,264][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:12:52,763][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:12:53,264][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:12:53,769][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:12:54,283][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:12:54,789][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:12:55,292][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:12:55,797][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:12:56,303][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:12:56,808][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:12:57,311][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:12:57,811][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:12:58,310][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:12:58,810][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:12:59,308][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:12:59,807][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:13:00,307][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:13:00,805][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:13:01,301][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:13:01,800][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:13:02,301][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:13:02,799][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:13:03,297][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:13:03,797][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:13:04,294][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:13:04,793][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10421 tokens. +[2026-03-26 00:13:05,461][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 60.61%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:32 +[2026-03-26 00:13:06,217][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:13:06,219][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:13:06,221][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:13:06,962][__main__][INFO] - Iteration 214 took 53s (33.08% Gen, 65.52% Train). Generation: 17s, Training: 34s. Estimated remaining time: 41h 11m 12s. Estimated total time: 44h 18m 38s. Time estimates for 10 more iterations: 8m 51s, 100 more iterations: 1h 28m 37s, 500 more iterations: 7h 23m 6s. +[2026-03-26 00:13:06,965][__main__][INFO] - Starting iteration 214. +[2026-03-26 00:13:07,366][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 21 and human policies 1. +[2026-03-26 00:13:07,366][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:13:20,298][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 00:13:21,153][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 1 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 00:13:22,940][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 1, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 00:13:24,677][__main__][INFO] - Number of regex retries in iteration 214: 3 +[2026-03-26 00:13:24,678][__main__][INFO] - agents played in iteration 214 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:13:25,474][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:13:25,497][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:13:25,520][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:13:25,543][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:13:25,544][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:13:25,544][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:13:26,174][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:13:26,631][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:13:27,136][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:13:27,635][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:13:28,139][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:13:28,638][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:13:29,141][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:13:29,640][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:13:30,137][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:13:30,633][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:13:31,129][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:13:31,624][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:13:32,124][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:13:32,621][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:13:33,120][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:13:33,617][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:13:34,114][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:13:34,613][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:13:35,112][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:13:35,609][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:13:36,107][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:13:36,608][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:13:37,107][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:13:37,606][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:13:38,105][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:13:38,603][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:13:39,098][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:13:39,598][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:13:40,097][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:13:40,598][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:13:41,099][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:13:41,599][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:13:42,097][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:13:42,595][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:13:43,091][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:13:43,588][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:13:44,085][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:13:44,583][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:13:45,081][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:13:45,580][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:13:46,076][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:13:46,574][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:13:47,072][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:13:47,571][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:13:48,070][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:13:48,566][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:13:49,064][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:13:49,563][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:13:50,063][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:13:50,560][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:13:51,062][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:13:51,561][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:13:52,060][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:13:52,560][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:13:53,060][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:13:53,559][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:13:54,059][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:13:54,559][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:13:55,057][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:13:55,556][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:13:56,054][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:13:56,552][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:13:57,050][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:13:57,545][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:13:58,044][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10562 tokens. +[2026-03-26 00:13:58,700][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.58%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 +[2026-03-26 00:13:59,456][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:13:59,458][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:13:59,460][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:14:00,146][__main__][INFO] - Iteration 215 took 52s (32.80% Gen, 65.90% Train). Generation: 17s, Training: 34s. Estimated remaining time: 40h 50m 45s. Estimated total time: 43h 59m 5s. Time estimates for 10 more iterations: 8m 47s, 100 more iterations: 1h 27m 58s, 500 more iterations: 7h 19m 50s. +[2026-03-26 00:14:00,151][__main__][INFO] - Starting iteration 215. +[2026-03-26 00:14:00,554][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 21 and human policies 1. +[2026-03-26 00:14:00,555][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:14:17,338][__main__][INFO] - Number of regex retries in iteration 215: 0 +[2026-03-26 00:14:17,339][__main__][INFO] - agents played in iteration 215 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:14:18,121][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:14:18,147][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:14:18,172][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:14:18,196][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:14:18,196][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:14:18,197][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:14:18,834][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:14:19,293][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:14:19,794][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:14:20,296][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:14:20,797][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:14:21,298][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:14:21,798][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:14:22,296][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:14:22,797][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:14:23,294][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:14:23,802][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:14:24,304][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:14:24,804][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:14:25,302][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:14:25,808][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:14:26,309][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:14:26,806][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:14:27,307][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:14:27,807][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:14:28,312][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:14:28,813][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:14:29,314][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:14:29,815][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:14:30,315][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:14:30,815][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:14:31,314][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:14:31,816][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:14:32,317][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:14:32,837][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:14:33,342][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:14:33,844][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:14:34,344][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:14:34,845][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:14:35,347][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:14:35,848][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:14:36,348][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:14:36,848][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:14:37,348][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:14:37,846][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:14:38,345][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:14:38,845][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:14:39,343][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:14:39,840][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:14:40,336][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:14:40,831][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:14:41,327][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:14:41,824][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:14:42,322][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:14:42,819][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:14:43,319][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:14:43,814][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:14:44,309][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:14:44,807][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:14:45,304][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:14:45,800][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:14:46,297][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:14:46,795][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:14:47,289][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:14:47,786][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:14:48,283][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:14:48,783][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:14:49,279][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:14:49,774][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:14:50,271][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:14:50,767][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10471 tokens. +[2026-03-26 00:14:51,443][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.39%, ΔTime: 00:00:32 +[2026-03-26 00:14:52,194][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:14:52,197][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:14:52,199][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:14:52,932][__main__][INFO] - Iteration 216 took 52s (32.04% Gen, 66.55% Train). Generation: 16s, Training: 34s. Estimated remaining time: 40h 29m 48s. Estimated total time: 43h 39m 1s. Time estimates for 10 more iterations: 8m 43s, 100 more iterations: 1h 27m 18s, 500 more iterations: 7h 16m 30s. +[2026-03-26 00:14:52,934][__main__][INFO] - Starting iteration 216. +[2026-03-26 00:14:53,338][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 21 and human policies 1. +[2026-03-26 00:14:53,339][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:15:04,062][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 00:15:10,786][__main__][INFO] - Number of regex retries in iteration 216: 1 +[2026-03-26 00:15:10,787][__main__][INFO] - agents played in iteration 216 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:15:11,564][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:15:11,590][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:15:11,616][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:15:11,641][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:15:11,641][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:15:11,642][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:15:12,283][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:15:12,747][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:15:13,252][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:15:13,753][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:15:14,255][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:15:14,764][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:15:15,266][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:15:15,768][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:15:16,268][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:15:16,769][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:15:17,269][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:15:17,775][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:15:18,279][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:15:18,778][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:15:19,278][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:15:19,778][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:15:20,281][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:15:20,778][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:15:21,286][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:15:21,786][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:15:22,284][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:15:22,783][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:15:23,282][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:15:23,781][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:15:24,278][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:15:24,779][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:15:25,279][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:15:25,778][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:15:26,286][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:15:26,787][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:15:27,288][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:15:27,789][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:15:28,289][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:15:28,790][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:15:29,288][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:15:29,786][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:15:30,287][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:15:30,785][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:15:31,282][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:15:31,783][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:15:32,295][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:15:32,792][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:15:33,289][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:15:33,791][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:15:34,287][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:15:34,785][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:15:35,287][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:15:35,784][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:15:36,283][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:15:36,780][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:15:37,277][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:15:37,780][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:15:38,280][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:15:38,777][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:15:39,275][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:15:39,775][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:15:40,270][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:15:40,778][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:15:41,281][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:15:41,782][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:15:42,280][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:15:42,781][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:15:43,280][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:15:43,781][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:15:44,281][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10462 tokens. +[2026-03-26 00:15:44,945][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 60.62%, Block Peak % of device VRAM: 62.27%, ΔTime: 00:00:32 +[2026-03-26 00:15:45,754][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:15:45,756][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:15:45,758][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:15:46,419][__main__][INFO] - Iteration 217 took 53s (32.87% Gen, 65.88% Train). Generation: 17s, Training: 34s. Estimated remaining time: 41h 3m 58s. Estimated total time: 44h 14m 4s. Time estimates for 10 more iterations: 8m 50s, 100 more iterations: 1h 28m 28s, 500 more iterations: 7h 22m 20s. +[2026-03-26 00:15:46,421][__main__][INFO] - Starting iteration 217. +[2026-03-26 00:15:46,820][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 21 and human policies 1. +[2026-03-26 00:15:46,820][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:16:00,097][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 00:16:04,495][__main__][INFO] - Number of regex retries in iteration 217: 1 +[2026-03-26 00:16:04,495][__main__][INFO] - agents played in iteration 217 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:16:05,271][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:16:05,299][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:16:05,325][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:16:05,350][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:16:05,351][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:16:05,352][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:16:05,996][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:16:06,456][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:16:06,960][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:16:07,459][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:16:07,968][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:16:08,468][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:16:08,969][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:16:09,468][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:16:09,974][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:16:10,472][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:16:10,972][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:16:11,470][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:16:11,968][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:16:12,467][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:16:12,970][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:16:13,469][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:16:13,970][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:16:14,471][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:16:14,971][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:16:15,474][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:16:15,982][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:16:16,482][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:16:16,984][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:16:17,482][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:16:17,983][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:16:18,483][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:16:18,981][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:16:19,481][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:16:19,983][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:16:20,483][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:16:20,982][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:16:21,483][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:16:21,983][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:16:22,483][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:16:22,983][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:16:23,483][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:16:23,982][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:16:24,482][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:16:24,982][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:16:25,482][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:16:25,980][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:16:26,479][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:16:26,975][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:16:27,471][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:16:27,970][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:16:28,470][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:16:28,974][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:16:29,472][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:16:29,970][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:16:30,467][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:16:30,965][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:16:31,465][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:16:31,967][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:16:32,463][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:16:32,961][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:16:33,459][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:16:33,955][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:16:34,456][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:16:34,955][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:16:35,454][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:16:35,953][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:16:36,454][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:16:36,955][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:16:37,456][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:16:37,955][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10476 tokens. +[2026-03-26 00:16:38,638][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 60.57%, Block Peak % of device VRAM: 62.27%, ΔTime: 00:00:32 +[2026-03-26 00:16:39,396][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:16:39,399][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:16:39,400][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:16:40,177][__main__][INFO] - Iteration 218 took 53s (33.13% Gen, 65.42% Train). Generation: 17s, Training: 34s. Estimated remaining time: 41h 16m 55s. Estimated total time: 44h 27m 55s. Time estimates for 10 more iterations: 8m 53s, 100 more iterations: 1h 28m 55s, 500 more iterations: 7h 24m 39s. +[2026-03-26 00:16:40,180][__main__][INFO] - Starting iteration 218. +[2026-03-26 00:16:40,581][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 21 and human policies 1. +[2026-03-26 00:16:40,582][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:16:48,674][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 00:16:58,203][__main__][INFO] - Number of regex retries in iteration 218: 1 +[2026-03-26 00:16:58,204][__main__][INFO] - agents played in iteration 218 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:16:58,973][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:16:58,999][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:16:59,023][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:16:59,047][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:16:59,048][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:16:59,049][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:16:59,699][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:17:00,154][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:17:00,659][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:17:01,158][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:17:01,668][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:17:02,169][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:17:02,667][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:17:03,166][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:17:03,667][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:17:04,172][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:17:04,672][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:17:05,181][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:17:05,682][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:17:06,186][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:17:06,687][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:17:07,205][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:17:07,709][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:17:08,205][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:17:08,710][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:17:09,212][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:17:09,713][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:17:10,213][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:17:10,712][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:17:11,213][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:17:11,711][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:17:12,210][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:17:12,708][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:17:13,213][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:17:13,710][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:17:14,208][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:17:14,704][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:17:15,203][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:17:15,703][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:17:16,200][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:17:16,697][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:17:17,198][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:17:17,698][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:17:18,195][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:17:18,700][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:17:19,196][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:17:19,698][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:17:20,199][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:17:20,700][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:17:21,200][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:17:21,699][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:17:22,199][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:17:22,696][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:17:23,197][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:17:23,700][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:17:24,202][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:17:24,700][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:17:25,202][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:17:25,703][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:17:26,202][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:17:26,702][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:17:27,202][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:17:27,710][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:17:28,207][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:17:28,705][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:17:29,202][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:17:29,702][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:17:30,200][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:17:30,699][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:17:31,196][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:17:31,693][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10449 tokens. +[2026-03-26 00:17:32,378][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 60.54%, Block Peak % of device VRAM: 62.28%, ΔTime: 00:00:32 +[2026-03-26 00:17:33,126][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:17:33,128][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:17:33,130][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:17:33,853][__main__][INFO] - Iteration 219 took 53s (33.08% Gen, 65.56% Train). Generation: 17s, Training: 34s. Estimated remaining time: 41h 11m 44s. Estimated total time: 44h 23m 37s. Time estimates for 10 more iterations: 8m 52s, 100 more iterations: 1h 28m 47s, 500 more iterations: 7h 23m 56s. +[2026-03-26 00:17:33,856][__main__][INFO] - Starting iteration 219. +[2026-03-26 00:17:34,253][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 21 and human policies 1. +[2026-03-26 00:17:34,254][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:17:47,028][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 1, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 00:17:50,813][__main__][INFO] - Number of regex retries in iteration 219: 1 +[2026-03-26 00:17:50,814][__main__][INFO] - agents played in iteration 219 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:17:51,577][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:17:51,601][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:17:51,624][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:17:51,647][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:17:51,648][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:17:51,648][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:17:52,277][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:17:52,727][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:17:53,233][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:17:53,729][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:17:54,230][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:17:54,727][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:17:55,228][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:17:55,726][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:17:56,225][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:17:56,724][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:17:57,222][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:17:57,721][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:17:58,219][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:17:58,719][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:17:59,218][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:17:59,717][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:18:00,215][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:18:00,711][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:18:01,208][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:18:01,713][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:18:02,222][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:18:02,723][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:18:03,220][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:18:03,721][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:18:04,218][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:18:04,716][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:18:05,218][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:18:05,719][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:18:06,222][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:18:06,724][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:18:07,224][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:18:07,724][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:18:08,224][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:18:08,723][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:18:09,222][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:18:09,718][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:18:10,216][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:18:10,716][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:18:11,215][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:18:11,713][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:18:12,212][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:18:12,711][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:18:13,209][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:18:13,709][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:18:14,207][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:18:14,705][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:18:15,202][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:18:15,702][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:18:16,202][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:18:16,702][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:18:17,202][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:18:17,702][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:18:18,202][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:18:18,701][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:18:19,201][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:18:19,701][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:18:20,200][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:18:20,698][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:18:21,200][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:18:21,697][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:18:22,197][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:18:22,693][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:18:23,194][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:18:23,691][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:18:24,191][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10421 tokens. +[2026-03-26 00:18:24,862][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.58%, Block Peak % of device VRAM: 62.17%, ΔTime: 00:00:32 +[2026-03-26 00:18:25,618][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:18:25,620][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:18:25,622][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:18:26,340][__main__][INFO] - Iteration 220 took 52s (31.79% Gen, 66.82% Train). Generation: 16s, Training: 34s. Estimated remaining time: 40h 11m 37s. Estimated total time: 43h 24m 23s. Time estimates for 10 more iterations: 8m 40s, 100 more iterations: 1h 26m 48s, 500 more iterations: 7h 14m 3s. +[2026-03-26 00:18:26,342][__main__][INFO] - Starting iteration 220. +[2026-03-26 00:18:26,744][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 21 and human policies 1. +[2026-03-26 00:18:26,744][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:18:40,028][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 1, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 00:18:43,240][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 00:18:44,090][__main__][INFO] - Number of regex retries in iteration 220: 2 +[2026-03-26 00:18:44,091][__main__][INFO] - agents played in iteration 220 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:18:44,874][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:18:44,897][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:18:44,920][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:18:44,943][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:18:44,944][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:18:44,944][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:18:45,587][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:18:46,043][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:18:46,547][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:18:47,046][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:18:47,545][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:18:48,043][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:18:48,543][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:18:49,054][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:18:49,555][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:18:50,053][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:18:50,552][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:18:51,050][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:18:51,550][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:18:52,048][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:18:52,548][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:18:53,047][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:18:53,546][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:18:54,046][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:18:54,547][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:18:55,048][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:18:55,544][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:18:56,043][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:18:56,540][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:18:57,041][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:18:57,541][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:18:58,041][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:18:58,542][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:18:59,041][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:18:59,539][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:19:00,039][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:19:00,538][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:19:01,037][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:19:01,537][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:19:02,038][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:19:02,543][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:19:03,040][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:19:03,541][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:19:04,041][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:19:04,545][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:19:05,048][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:19:05,547][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:19:06,045][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:19:06,546][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:19:07,045][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:19:07,545][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:19:08,043][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:19:08,545][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:19:09,043][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:19:09,541][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:19:10,041][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:19:10,540][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:19:11,039][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:19:11,539][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:19:12,040][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:19:12,541][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:19:13,041][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:19:13,540][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:19:14,039][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:19:14,535][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:19:15,034][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:19:15,533][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:19:16,031][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:19:16,530][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:19:17,033][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:19:17,534][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10506 tokens. +[2026-03-26 00:19:18,203][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 60.61%, Block Peak % of device VRAM: 62.23%, ΔTime: 00:00:32 +[2026-03-26 00:19:18,945][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:19:18,948][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:19:18,949][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:19:20,572][__main__][INFO] - Iteration 221 took 53s (32.22% Gen, 64.76% Train). Generation: 17s, Training: 34s. Estimated remaining time: 41h 37m 47s. Estimated total time: 44h 51m 27s. Time estimates for 10 more iterations: 8m 58s, 100 more iterations: 1h 29m 42s, 500 more iterations: 7h 28m 34s. +[2026-03-26 00:19:20,576][__main__][INFO] - Starting iteration 221. +[2026-03-26 00:19:20,978][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 22 and human policies 1. +[2026-03-26 00:19:20,979][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:19:38,537][__main__][INFO] - Number of regex retries in iteration 221: 0 +[2026-03-26 00:19:38,538][__main__][INFO] - agents played in iteration 221 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:19:39,324][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:19:39,347][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:19:39,371][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:19:39,394][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:19:39,394][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:19:39,395][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:19:40,027][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:19:40,490][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:19:40,996][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:19:41,494][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:19:41,993][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:19:42,489][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:19:42,988][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:19:43,486][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:19:43,984][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:19:44,487][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:19:44,987][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:19:45,483][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:19:45,981][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:19:46,481][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:19:46,980][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:19:47,491][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:19:47,991][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:19:48,491][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:19:48,993][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:19:49,493][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:19:49,999][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:19:50,501][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:19:51,002][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:19:51,518][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:19:52,018][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:19:52,522][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:19:53,022][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:19:53,522][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:19:54,023][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:19:54,522][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:19:55,022][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:19:55,523][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:19:56,022][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:19:56,522][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:19:57,021][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:19:57,522][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:19:58,023][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:19:58,522][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:19:59,022][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:19:59,521][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:20:00,021][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:20:00,521][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:20:01,022][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:20:01,522][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:20:02,022][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:20:02,523][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:20:03,023][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:20:03,524][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:20:04,024][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:20:04,523][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:20:05,022][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:20:05,522][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:20:06,022][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:20:06,520][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:20:07,020][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:20:07,521][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:20:08,021][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:20:08,523][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:20:09,021][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:20:09,523][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:20:10,022][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:20:10,522][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:20:11,021][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:20:11,521][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:20:12,019][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10516 tokens. +[2026-03-26 00:20:12,765][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 60.67%, Block Peak % of device VRAM: 62.36%, ΔTime: 00:00:32 +[2026-03-26 00:20:13,516][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:20:13,518][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:20:13,520][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:20:14,345][__main__][INFO] - Iteration 222 took 53s (32.90% Gen, 65.55% Train). Generation: 17s, Training: 34s. Estimated remaining time: 41h 13m 48s. Estimated total time: 44h 28m 22s. Time estimates for 10 more iterations: 8m 53s, 100 more iterations: 1h 28m 56s, 500 more iterations: 7h 24m 43s. +[2026-03-26 00:20:14,347][__main__][INFO] - Starting iteration 222. +[2026-03-26 00:20:14,744][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 22 and human policies 1. +[2026-03-26 00:20:14,745][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:20:30,746][__main__][INFO] - Number of regex retries in iteration 222: 0 +[2026-03-26 00:20:30,747][__main__][INFO] - agents played in iteration 222 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:20:31,507][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:20:31,530][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:20:31,554][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:20:31,576][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:20:31,577][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:20:31,578][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:20:32,205][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:20:32,663][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:20:33,167][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:20:33,667][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:20:34,166][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:20:34,666][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:20:35,165][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:20:35,664][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:20:36,164][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:20:36,663][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:20:37,163][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:20:37,662][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:20:38,160][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:20:38,660][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:20:39,160][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:20:39,655][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:20:40,152][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:20:40,650][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:20:41,148][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:20:41,647][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:20:42,147][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:20:42,643][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:20:43,141][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:20:43,642][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:20:44,140][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:20:44,637][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:20:45,133][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:20:45,629][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:20:46,128][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:20:46,624][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:20:47,124][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:20:47,621][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:20:48,119][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:20:48,620][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:20:49,119][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:20:49,616][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:20:50,114][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:20:50,613][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:20:51,112][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:20:51,610][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:20:52,106][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:20:52,611][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:20:53,110][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:20:53,635][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:20:54,135][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:20:54,631][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:20:55,131][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:20:55,629][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:20:56,128][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:20:56,627][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:20:57,126][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:20:57,624][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:20:58,121][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:20:58,618][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:20:59,117][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:20:59,621][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:21:00,121][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:21:00,622][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:21:01,121][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:21:01,619][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:21:02,121][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:21:02,624][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:21:03,123][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:21:03,622][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:21:04,121][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10524 tokens. +[2026-03-26 00:21:04,780][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 60.61%, Block Peak % of device VRAM: 62.44%, ΔTime: 00:00:32 +[2026-03-26 00:21:05,532][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:21:05,535][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:21:05,536][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:21:06,265][__main__][INFO] - Iteration 223 took 51s (31.06% Gen, 67.52% Train). Generation: 16s, Training: 34s. Estimated remaining time: 39h 40m 38s. Estimated total time: 42h 56m 4s. Time estimates for 10 more iterations: 8m 35s, 100 more iterations: 1h 25m 52s, 500 more iterations: 7h 9m 20s. +[2026-03-26 00:21:06,267][__main__][INFO] - Starting iteration 223. +[2026-03-26 00:21:06,665][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 22 and human policies 1. +[2026-03-26 00:21:06,666][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:21:13,142][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 00:21:26,017][__main__][INFO] - Number of regex retries in iteration 223: 1 +[2026-03-26 00:21:26,017][__main__][INFO] - agents played in iteration 223 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:21:26,782][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:21:26,805][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:21:26,828][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:21:26,851][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:21:26,852][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:21:26,852][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:21:27,482][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:21:27,937][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:21:28,442][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:21:28,944][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:21:29,444][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:21:29,945][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:21:30,443][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:21:30,941][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:21:31,443][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:21:31,942][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:21:32,441][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:21:32,940][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:21:33,438][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:21:33,934][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:21:34,434][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:21:34,934][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:21:35,434][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:21:35,929][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:21:36,426][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:21:36,923][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:21:37,420][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:21:37,916][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:21:38,412][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:21:38,909][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:21:39,406][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:21:39,902][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:21:40,398][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:21:40,894][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:21:41,389][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:21:41,886][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:21:42,383][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:21:42,879][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:21:43,375][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:21:43,871][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:21:44,368][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:21:44,865][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:21:45,361][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:21:45,861][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:21:46,361][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:21:46,856][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:21:47,355][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:21:47,849][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:21:48,345][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:21:48,842][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:21:49,341][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:21:49,838][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:21:50,333][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:21:50,829][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:21:51,325][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:21:51,823][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:21:52,323][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:21:52,821][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:21:53,319][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:21:53,821][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:21:54,320][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:21:54,820][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:21:55,319][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:21:55,817][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:21:56,315][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:21:56,812][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:21:57,310][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:21:57,809][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:21:58,308][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:21:58,805][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:21:59,303][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10507 tokens. +[2026-03-26 00:21:59,965][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 60.56%, Block Peak % of device VRAM: 62.18%, ΔTime: 00:00:32 +[2026-03-26 00:22:00,710][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:22:00,712][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:22:00,714][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:22:01,439][__main__][INFO] - Iteration 224 took 54s (35.33% Gen, 63.34% Train). Generation: 19s, Training: 34s. Estimated remaining time: 42h 22m 22s. Estimated total time: 45h 38m 43s. Time estimates for 10 more iterations: 9m 7s, 100 more iterations: 1h 31m 17s, 500 more iterations: 7h 36m 27s. +[2026-03-26 00:22:01,441][__main__][INFO] - Starting iteration 224. +[2026-03-26 00:22:01,842][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 22 and human policies 1. +[2026-03-26 00:22:01,843][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:22:05,278][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 00:22:17,896][__main__][INFO] - Number of regex retries in iteration 224: 1 +[2026-03-26 00:22:17,897][__main__][INFO] - agents played in iteration 224 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:22:18,675][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:22:18,698][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:22:18,721][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:22:18,744][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:22:18,744][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:22:18,745][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:22:19,375][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:22:19,833][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:22:20,339][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:22:20,840][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:22:21,340][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:22:21,840][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:22:22,341][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:22:22,839][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:22:23,337][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:22:23,833][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:22:24,331][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:22:24,830][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:22:25,329][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:22:25,828][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:22:26,327][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:22:26,827][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:22:27,327][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:22:27,823][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:22:28,319][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:22:28,821][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:22:29,321][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:22:29,821][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:22:30,315][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:22:30,812][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:22:31,308][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:22:31,809][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:22:32,305][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:22:32,806][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:22:33,306][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:22:33,805][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:22:34,301][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:22:34,797][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:22:35,293][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:22:35,793][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:22:36,294][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:22:36,795][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:22:37,295][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:22:37,793][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:22:38,294][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:22:38,794][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:22:39,294][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:22:39,790][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:22:40,285][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:22:40,784][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:22:41,281][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:22:41,780][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:22:42,276][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:22:42,773][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:22:43,272][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:22:43,772][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:22:44,272][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:22:44,774][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:22:45,269][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:22:45,768][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:22:46,268][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:22:46,768][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:22:47,270][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:22:47,769][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:22:48,267][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:22:48,766][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:22:49,266][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:22:49,765][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:22:50,265][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:22:50,763][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:22:51,264][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10471 tokens. +[2026-03-26 00:22:51,930][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.25%, Current % of VRAM taken: 60.68%, Block Peak % of device VRAM: 62.36%, ΔTime: 00:00:32 +[2026-03-26 00:22:52,656][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:22:52,658][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:22:52,660][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:22:53,388][__main__][INFO] - Iteration 225 took 51s (31.15% Gen, 67.44% Train). Generation: 16s, Training: 34s. Estimated remaining time: 39h 40m 7s. Estimated total time: 42h 57m 20s. Time estimates for 10 more iterations: 8m 35s, 100 more iterations: 1h 25m 54s, 500 more iterations: 7h 9m 33s. +[2026-03-26 00:22:53,391][__main__][INFO] - Starting iteration 225. +[2026-03-26 00:22:53,792][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 22 and human policies 1. +[2026-03-26 00:22:53,793][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:23:11,658][__main__][INFO] - Number of regex retries in iteration 225: 0 +[2026-03-26 00:23:11,659][__main__][INFO] - agents played in iteration 225 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:23:12,429][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:23:12,452][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:23:12,475][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:23:12,498][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:23:12,498][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:23:12,499][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:23:13,127][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:23:13,581][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:23:14,083][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:23:14,585][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:23:15,082][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:23:15,578][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:23:16,076][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:23:16,574][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:23:17,071][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:23:17,574][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:23:18,073][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:23:18,571][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:23:19,073][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:23:19,569][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:23:20,067][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:23:20,567][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:23:21,066][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:23:21,561][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:23:22,058][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:23:22,555][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:23:23,051][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:23:23,547][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:23:24,046][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:23:24,543][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:23:25,043][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:23:25,542][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:23:26,040][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:23:26,536][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:23:27,035][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:23:27,530][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:23:28,028][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:23:28,526][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:23:29,021][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:23:29,516][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:23:30,015][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:23:30,510][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:23:31,009][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:23:31,509][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:23:32,007][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:23:32,509][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:23:33,007][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:23:33,506][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:23:34,004][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:23:34,503][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:23:35,024][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:23:35,522][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:23:36,022][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:23:36,520][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:23:37,018][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:23:37,518][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:23:38,014][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:23:38,512][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:23:39,009][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:23:39,507][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:23:40,007][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:23:40,504][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:23:41,003][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:23:41,500][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:23:41,999][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:23:42,498][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:23:42,997][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:23:43,494][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:23:43,995][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:23:44,490][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:23:44,989][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10481 tokens. +[2026-03-26 00:23:45,660][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 60.60%, Block Peak % of device VRAM: 62.22%, ΔTime: 00:00:32 +[2026-03-26 00:23:46,403][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:23:46,405][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:23:46,407][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:23:47,161][__main__][INFO] - Iteration 226 took 53s (33.48% Gen, 65.11% Train). Generation: 17s, Training: 34s. Estimated remaining time: 41h 10m 21s. Estimated total time: 44h 28m 28s. Time estimates for 10 more iterations: 8m 53s, 100 more iterations: 1h 28m 56s, 500 more iterations: 7h 24m 44s. +[2026-03-26 00:23:47,163][__main__][INFO] - Starting iteration 226. +[2026-03-26 00:23:47,563][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 22 and human policies 1. +[2026-03-26 00:23:47,564][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:24:09,891][__main__][INFO] - Number of regex retries in iteration 226: 0 +[2026-03-26 00:24:09,892][__main__][INFO] - agents played in iteration 226 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:24:10,669][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:24:10,693][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:24:10,716][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:24:10,739][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:24:10,739][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:24:10,740][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:24:11,375][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:24:11,836][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:24:12,338][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:24:12,837][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:24:13,333][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:24:13,832][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:24:14,329][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:24:14,827][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:24:15,327][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:24:15,822][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:24:16,320][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:24:16,814][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:24:17,309][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:24:17,805][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:24:18,304][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:24:18,803][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:24:19,299][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:24:19,802][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:24:20,302][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:24:20,802][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:24:21,301][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:24:21,801][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:24:22,301][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:24:22,801][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:24:23,300][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:24:23,796][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:24:24,291][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:24:24,790][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:24:25,288][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:24:25,785][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:24:26,285][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:24:26,782][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:24:27,281][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:24:27,776][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:24:28,274][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:24:28,771][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:24:29,267][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:24:29,763][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:24:30,259][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:24:30,755][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:24:31,254][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:24:31,751][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:24:32,252][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:24:32,747][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:24:33,244][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:24:33,744][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:24:34,241][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:24:34,736][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:24:35,236][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:24:35,732][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:24:36,227][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:24:36,723][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:24:37,221][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:24:37,717][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:24:38,217][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:24:38,714][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:24:39,210][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:24:39,710][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:24:40,207][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:24:40,706][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:24:41,206][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:24:41,705][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:24:42,204][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:24:42,702][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:24:43,201][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10423 tokens. +[2026-03-26 00:24:43,894][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 60.62%, Block Peak % of device VRAM: 62.27%, ΔTime: 00:00:32 +[2026-03-26 00:24:44,638][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:24:44,641][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:24:44,642][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:24:45,341][__main__][INFO] - Iteration 227 took 57s (38.65% Gen, 60.14% Train). Generation: 22s, Training: 34s. Estimated remaining time: 44h 49m 49s. Estimated total time: 48h 8m 54s. Time estimates for 10 more iterations: 9m 37s, 100 more iterations: 1h 36m 17s, 500 more iterations: 8h 1m 29s. +[2026-03-26 00:24:45,344][__main__][INFO] - Starting iteration 227. +[2026-03-26 00:24:45,746][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 22 and human policies 1. +[2026-03-26 00:24:45,746][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:25:02,975][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 1, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 00:25:03,736][__main__][INFO] - Number of regex retries in iteration 227: 1 +[2026-03-26 00:25:03,737][__main__][INFO] - agents played in iteration 227 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:25:04,515][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:25:04,538][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:25:04,561][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:25:04,584][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:25:04,584][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:25:04,585][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:25:05,217][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:25:05,675][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:25:06,182][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:25:06,682][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:25:07,182][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:25:07,682][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:25:08,180][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:25:08,680][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:25:09,181][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:25:09,681][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:25:10,181][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:25:10,684][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:25:11,190][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:25:11,692][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:25:12,195][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:25:12,697][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:25:13,201][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:25:13,704][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:25:14,205][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:25:14,704][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:25:15,203][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:25:15,703][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:25:16,203][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:25:16,703][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:25:17,202][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:25:17,699][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:25:18,197][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:25:18,694][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:25:19,193][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:25:19,690][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:25:20,189][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:25:20,687][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:25:21,185][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:25:21,685][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:25:22,184][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:25:22,684][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:25:23,182][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:25:23,682][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:25:24,182][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:25:24,682][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:25:25,181][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:25:25,677][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:25:26,174][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:25:26,672][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:25:27,167][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:25:27,663][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:25:28,160][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:25:28,661][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:25:29,156][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:25:29,655][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:25:30,154][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:25:30,652][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:25:31,150][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:25:31,649][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:25:32,146][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:25:32,646][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:25:33,145][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:25:33,643][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:25:34,142][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:25:34,643][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:25:35,142][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:25:35,641][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:25:36,137][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:25:36,639][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:25:37,140][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10614 tokens. +[2026-03-26 00:25:37,798][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 60.64%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:32 +[2026-03-26 00:25:38,543][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:25:38,545][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:25:38,547][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:25:39,266][__main__][INFO] - Iteration 228 took 53s (33.61% Gen, 65.04% Train). Generation: 17s, Training: 34s. Estimated remaining time: 41h 16m 4s. Estimated total time: 44h 36m 3s. Time estimates for 10 more iterations: 8m 55s, 100 more iterations: 1h 29m 12s, 500 more iterations: 7h 26m 0s. +[2026-03-26 00:25:39,269][__main__][INFO] - Starting iteration 228. +[2026-03-26 00:25:39,666][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 22 and human policies 1. +[2026-03-26 00:25:39,667][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:25:55,977][__main__][INFO] - Number of regex retries in iteration 228: 0 +[2026-03-26 00:25:55,978][__main__][INFO] - agents played in iteration 228 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:25:56,743][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:25:56,767][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:25:56,790][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:25:56,812][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:25:56,813][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:25:56,813][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:25:57,443][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:25:57,901][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:25:58,404][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:25:58,904][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:25:59,405][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:25:59,903][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:26:00,401][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:26:00,899][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:26:01,396][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:26:01,897][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:26:02,396][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:26:02,896][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:26:03,400][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:26:03,901][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:26:04,402][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:26:04,902][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:26:05,402][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:26:05,902][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:26:06,419][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:26:06,924][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:26:07,423][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:26:07,922][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:26:08,422][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:26:08,921][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:26:09,420][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:26:09,920][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:26:10,418][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:26:10,914][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:26:11,412][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:26:11,910][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:26:12,409][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:26:12,908][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:26:13,406][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:26:13,901][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:26:14,400][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:26:14,895][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:26:15,391][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:26:15,888][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:26:16,384][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:26:16,883][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:26:17,381][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:26:17,877][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:26:18,377][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:26:18,875][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:26:19,371][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:26:19,867][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:26:20,366][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:26:20,864][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:26:21,363][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:26:21,859][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:26:22,355][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:26:22,854][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:26:23,348][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:26:23,845][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:26:24,341][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:26:24,842][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:26:25,337][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:26:25,835][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:26:26,330][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:26:26,828][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:26:27,324][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:26:27,818][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:26:28,316][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:26:28,810][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:26:29,305][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10411 tokens. +[2026-03-26 00:26:29,986][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 60.54%, Block Peak % of device VRAM: 62.40%, ΔTime: 00:00:32 +[2026-03-26 00:26:30,729][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:26:30,731][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:26:30,733][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:26:31,396][__main__][INFO] - Iteration 229 took 51s (31.53% Gen, 67.19% Train). Generation: 16s, Training: 34s. Estimated remaining time: 39h 45m 40s. Estimated total time: 43h 6m 31s. Time estimates for 10 more iterations: 8m 37s, 100 more iterations: 1h 26m 13s, 500 more iterations: 7h 11m 5s. +[2026-03-26 00:26:31,399][__main__][INFO] - Starting iteration 229. +[2026-03-26 00:26:31,797][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 22 and human policies 1. +[2026-03-26 00:26:31,797][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:26:48,850][__main__][INFO] - Number of regex retries in iteration 229: 0 +[2026-03-26 00:26:48,851][__main__][INFO] - agents played in iteration 229 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:26:49,614][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:26:49,638][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:26:49,661][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:26:49,683][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:26:49,684][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:26:49,684][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:26:50,324][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:26:50,782][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:26:51,286][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:26:51,785][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:26:52,282][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:26:52,780][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:26:53,281][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:26:53,780][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:26:54,280][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:26:54,782][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:26:55,277][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:26:55,773][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:26:56,273][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:26:56,774][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:26:57,272][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:26:57,770][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:26:58,270][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:26:58,767][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:26:59,268][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:26:59,766][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:27:00,266][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:27:00,766][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:27:01,264][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:27:01,764][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:27:02,265][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:27:02,761][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:27:03,257][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:27:03,753][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:27:04,251][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:27:04,746][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:27:05,246][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:27:05,743][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:27:06,243][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:27:06,745][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:27:07,241][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:27:07,736][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:27:08,232][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:27:08,727][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:27:09,226][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:27:09,724][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:27:10,222][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:27:10,721][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:27:11,221][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:27:11,719][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:27:12,217][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:27:12,713][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:27:13,212][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:27:13,709][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:27:14,205][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:27:14,703][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:27:15,200][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:27:15,698][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:27:16,198][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:27:16,694][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:27:17,193][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:27:17,693][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:27:18,187][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:27:18,686][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:27:19,184][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:27:19,683][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:27:20,183][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:27:20,679][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:27:21,177][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:27:21,675][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:27:22,172][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10527 tokens. +[2026-03-26 00:27:22,847][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 60.63%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:32 +[2026-03-26 00:27:23,594][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:27:23,596][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:27:23,598][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:27:24,439][__main__][INFO] - Iteration 230 took 52s (32.39% Gen, 66.01% Train). Generation: 17s, Training: 34s. Estimated remaining time: 40h 30m 24s. Estimated total time: 43h 52m 8s. Time estimates for 10 more iterations: 8m 46s, 100 more iterations: 1h 27m 44s, 500 more iterations: 7h 18m 41s. +[2026-03-26 00:27:24,441][__main__][INFO] - Starting iteration 230. +[2026-03-26 00:27:24,847][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 22 and human policies 1. +[2026-03-26 00:27:24,848][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:27:45,369][__main__][INFO] - Number of regex retries in iteration 230: 0 +[2026-03-26 00:27:45,369][__main__][INFO] - agents played in iteration 230 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:27:46,145][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:27:46,168][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:27:46,191][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:27:46,214][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:27:46,214][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:27:46,215][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:27:46,845][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:27:47,300][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:27:47,805][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:27:48,304][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:27:48,804][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:27:49,305][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:27:49,801][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:27:50,300][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:27:50,799][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:27:51,297][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:27:51,796][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:27:52,293][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:27:52,792][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:27:53,291][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:27:53,790][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:27:54,286][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:27:54,784][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:27:55,279][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:27:55,776][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:27:56,272][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:27:56,768][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:27:57,265][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:27:57,765][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:27:58,262][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:27:58,757][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:27:59,254][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:27:59,751][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:28:00,250][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:28:00,750][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:28:01,247][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:28:01,747][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:28:02,249][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:28:02,749][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:28:03,249][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:28:03,747][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:28:04,246][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:28:04,745][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:28:05,244][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:28:05,743][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:28:06,241][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:28:06,739][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:28:07,237][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:28:07,739][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:28:08,238][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:28:08,741][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:28:09,241][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:28:09,742][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:28:10,239][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:28:10,738][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:28:11,236][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:28:11,735][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:28:12,232][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:28:12,731][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:28:13,230][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:28:13,727][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:28:14,225][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:28:14,723][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:28:15,221][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:28:15,723][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:28:16,225][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:28:16,729][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:28:17,232][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:28:17,732][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:28:18,232][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:28:18,733][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10491 tokens. +[2026-03-26 00:28:19,405][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 60.60%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:32 +[2026-03-26 00:28:20,167][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:28:20,169][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:28:20,171][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:28:21,457][__main__][INFO] - Iteration 231 took 56s (36.25% Gen, 61.48% Train). Generation: 20s, Training: 34s. Estimated remaining time: 43h 47m 49s. Estimated total time: 47h 10m 30s. Time estimates for 10 more iterations: 9m 26s, 100 more iterations: 1h 34m 21s, 500 more iterations: 7h 51m 45s. +[2026-03-26 00:28:21,459][__main__][INFO] - Starting iteration 231. +[2026-03-26 00:28:21,860][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 23 and human policies 1. +[2026-03-26 00:28:21,860][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:28:26,368][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 00:28:39,853][__main__][INFO] - Number of regex retries in iteration 231: 1 +[2026-03-26 00:28:39,854][__main__][INFO] - agents played in iteration 231 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:28:40,619][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:28:40,642][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:28:40,665][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:28:40,688][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:28:40,688][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:28:40,689][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:28:41,329][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:28:41,786][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:28:42,289][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:28:42,790][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:28:43,290][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:28:43,786][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:28:44,286][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:28:44,785][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:28:45,284][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:28:45,785][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:28:46,283][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:28:46,782][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:28:47,280][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:28:47,779][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:28:48,277][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:28:48,776][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:28:49,275][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:28:49,774][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:28:50,276][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:28:50,773][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:28:51,272][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:28:51,771][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:28:52,269][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:28:52,767][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:28:53,267][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:28:53,764][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:28:54,263][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:28:54,761][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:28:55,261][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:28:55,756][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:28:56,252][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:28:56,751][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:28:57,248][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:28:57,746][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:28:58,243][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:28:58,743][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:28:59,240][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:28:59,738][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:29:00,237][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:29:00,731][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:29:01,228][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:29:01,724][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:29:02,226][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:29:02,725][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:29:03,226][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:29:03,723][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:29:04,222][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:29:04,721][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:29:05,221][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:29:05,719][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:29:06,217][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:29:06,715][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:29:07,213][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:29:07,711][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:29:08,206][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:29:08,704][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:29:09,198][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:29:09,697][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:29:10,197][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:29:10,697][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:29:11,196][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:29:11,696][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:29:12,193][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:29:12,689][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:29:13,187][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10532 tokens. +[2026-03-26 00:29:13,862][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 60.56%, Block Peak % of device VRAM: 62.20%, ΔTime: 00:00:32 +[2026-03-26 00:29:14,600][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:29:14,604][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:29:14,606][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:29:15,312][__main__][INFO] - Iteration 232 took 53s (33.66% Gen, 65.01% Train). Generation: 17s, Training: 34s. Estimated remaining time: 41h 9m 3s. Estimated total time: 44h 32m 38s. Time estimates for 10 more iterations: 8m 54s, 100 more iterations: 1h 29m 5s, 500 more iterations: 7h 25m 26s. +[2026-03-26 00:29:15,314][__main__][INFO] - Starting iteration 232. +[2026-03-26 00:29:15,714][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 23 and human policies 1. +[2026-03-26 00:29:15,715][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:29:21,557][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 00:29:32,838][__main__][INFO] - Number of regex retries in iteration 232: 1 +[2026-03-26 00:29:32,839][__main__][INFO] - agents played in iteration 232 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:29:33,606][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:29:33,630][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:29:33,653][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:29:33,675][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:29:33,676][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:29:33,676][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:29:34,322][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:29:34,781][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:29:35,286][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:29:35,786][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:29:36,286][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:29:36,787][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:29:37,287][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:29:37,788][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:29:38,287][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:29:38,788][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:29:39,288][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:29:39,787][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:29:40,288][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:29:40,791][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:29:41,290][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:29:41,791][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:29:42,291][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:29:42,788][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:29:43,286][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:29:43,787][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:29:44,284][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:29:44,785][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:29:45,282][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:29:45,784][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:29:46,280][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:29:46,781][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:29:47,279][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:29:47,779][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:29:48,281][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:29:48,783][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:29:49,283][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:29:49,782][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:29:50,281][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:29:50,777][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:29:51,274][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:29:51,770][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:29:52,268][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:29:52,768][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:29:53,265][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:29:53,765][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:29:54,261][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:29:54,758][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:29:55,256][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:29:55,756][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:29:56,257][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:29:56,756][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:29:57,252][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:29:57,750][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:29:58,248][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:29:58,749][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:29:59,247][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:29:59,746][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:30:00,244][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:30:00,744][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:30:01,245][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:30:01,745][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:30:02,245][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:30:02,741][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:30:03,239][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:30:03,738][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:30:04,239][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:30:04,741][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:30:05,237][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:30:05,738][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:30:06,237][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10483 tokens. +[2026-03-26 00:30:06,916][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 60.62%, Block Peak % of device VRAM: 62.28%, ΔTime: 00:00:32 +[2026-03-26 00:30:07,661][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:30:07,664][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:30:07,665][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:30:08,377][__main__][INFO] - Iteration 233 took 52s (32.52% Gen, 66.13% Train). Generation: 17s, Training: 34s. Estimated remaining time: 40h 28m 41s. Estimated total time: 43h 53m 9s. Time estimates for 10 more iterations: 8m 46s, 100 more iterations: 1h 27m 46s, 500 more iterations: 7h 18m 51s. +[2026-03-26 00:30:08,379][__main__][INFO] - Starting iteration 233. +[2026-03-26 00:30:08,777][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 23 and human policies 1. +[2026-03-26 00:30:08,778][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:30:27,402][__main__][INFO] - Number of regex retries in iteration 233: 0 +[2026-03-26 00:30:27,403][__main__][INFO] - agents played in iteration 233 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:30:28,170][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:30:28,193][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:30:28,216][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:30:28,239][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:30:28,239][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:30:28,240][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:30:28,891][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:30:29,344][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:30:29,850][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:30:30,349][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:30:30,850][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:30:31,354][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:30:31,852][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:30:32,351][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:30:32,849][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:30:33,350][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:30:33,869][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:30:34,370][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:30:34,870][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:30:35,370][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:30:35,871][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:30:36,371][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:30:36,871][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:30:37,371][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:30:37,871][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:30:38,368][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:30:38,869][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:30:39,370][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:30:39,871][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:30:40,371][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:30:40,870][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:30:41,371][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:30:41,873][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:30:42,373][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:30:42,873][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:30:43,374][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:30:43,873][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:30:44,374][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:30:44,876][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:30:45,372][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:30:45,869][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:30:46,367][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:30:46,865][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:30:47,365][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:30:47,864][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:30:48,363][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:30:48,861][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:30:49,357][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:30:49,855][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:30:50,354][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:30:50,852][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:30:51,349][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:30:51,849][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:30:52,348][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:30:52,847][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:30:53,351][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:30:53,856][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:30:54,358][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:30:54,858][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:30:55,357][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:30:55,856][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:30:56,355][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:30:56,850][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:30:57,348][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:30:57,848][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:30:58,345][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:30:58,846][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:30:59,345][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:30:59,842][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:31:00,339][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:31:00,838][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10484 tokens. +[2026-03-26 00:31:01,492][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.25%, Current % of VRAM taken: 60.68%, Block Peak % of device VRAM: 62.28%, ΔTime: 00:00:32 +[2026-03-26 00:31:02,242][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:31:02,245][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:31:02,246][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:31:02,974][__main__][INFO] - Iteration 234 took 54s (34.37% Gen, 64.29% Train). Generation: 18s, Training: 34s. Estimated remaining time: 41h 44m 30s. Estimated total time: 45h 9m 53s. Time estimates for 10 more iterations: 9m 1s, 100 more iterations: 1h 30m 19s, 500 more iterations: 7h 31m 38s. +[2026-03-26 00:31:02,976][__main__][INFO] - Starting iteration 234. +[2026-03-26 00:31:03,375][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 23 and human policies 1. +[2026-03-26 00:31:03,376][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:31:24,423][__main__][INFO] - Number of regex retries in iteration 234: 0 +[2026-03-26 00:31:24,424][__main__][INFO] - agents played in iteration 234 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:31:25,202][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:31:25,226][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:31:25,249][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:31:25,271][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:31:25,272][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:31:25,272][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:31:25,918][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:31:26,373][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:31:26,879][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:31:27,378][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:31:27,876][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:31:28,375][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:31:28,876][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:31:29,376][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:31:29,875][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:31:30,373][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:31:30,868][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:31:31,366][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:31:31,866][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:31:32,364][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:31:32,863][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:31:33,362][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:31:33,857][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:31:34,357][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:31:34,852][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:31:35,350][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:31:35,850][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:31:36,351][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:31:36,852][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:31:37,350][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:31:37,848][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:31:38,348][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:31:38,848][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:31:39,348][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:31:39,847][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:31:40,345][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:31:40,845][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:31:41,348][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:31:41,847][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:31:42,347][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:31:42,845][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:31:43,344][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:31:43,844][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:31:44,344][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:31:44,843][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:31:45,343][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:31:45,843][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:31:46,339][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:31:46,835][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:31:47,331][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:31:47,826][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:31:48,324][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:31:48,823][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:31:49,321][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:31:49,818][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:31:50,315][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:31:50,813][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:31:51,309][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:31:51,807][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:31:52,302][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:31:52,801][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:31:53,297][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:31:53,796][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:31:54,292][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:31:54,789][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:31:55,289][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:31:55,789][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:31:56,288][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:31:56,787][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:31:57,284][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:31:57,781][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10514 tokens. +[2026-03-26 00:31:58,442][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 60.54%, Block Peak % of device VRAM: 62.34%, ΔTime: 00:00:32 +[2026-03-26 00:31:59,185][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:31:59,187][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:31:59,189][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:31:59,918][__main__][INFO] - Iteration 235 took 56s (37.23% Gen, 61.48% Train). Generation: 21s, Training: 34s. Estimated remaining time: 43h 40m 50s. Estimated total time: 47h 7m 10s. Time estimates for 10 more iterations: 9m 25s, 100 more iterations: 1h 34m 14s, 500 more iterations: 7h 51m 11s. +[2026-03-26 00:31:59,920][__main__][INFO] - Starting iteration 235. +[2026-03-26 00:32:00,321][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 23 and human policies 1. +[2026-03-26 00:32:00,321][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:32:18,305][__main__][INFO] - Number of regex retries in iteration 235: 0 +[2026-03-26 00:32:18,306][__main__][INFO] - agents played in iteration 235 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:32:19,085][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:32:19,108][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:32:19,131][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:32:19,154][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:32:19,154][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:32:19,155][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:32:19,797][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:32:20,249][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:32:20,752][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:32:21,253][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:32:21,749][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:32:22,247][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:32:22,744][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:32:23,242][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:32:23,742][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:32:24,243][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:32:24,743][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:32:25,249][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:32:25,752][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:32:26,251][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:32:26,751][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:32:27,251][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:32:27,772][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:32:28,271][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:32:28,769][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:32:29,268][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:32:29,767][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:32:30,266][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:32:30,766][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:32:31,265][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:32:31,764][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:32:32,266][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:32:32,765][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:32:33,265][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:32:33,764][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:32:34,263][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:32:34,762][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:32:35,262][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:32:35,760][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:32:36,257][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:32:36,757][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:32:37,255][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:32:37,754][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:32:38,253][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:32:38,752][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:32:39,250][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:32:39,750][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:32:40,246][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:32:40,745][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:32:41,245][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:32:41,745][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:32:42,245][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:32:42,744][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:32:43,241][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:32:43,737][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:32:44,234][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:32:44,730][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:32:45,229][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:32:45,728][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:32:46,227][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:32:46,726][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:32:47,224][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:32:47,720][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:32:48,218][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:32:48,714][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:32:49,213][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:32:49,710][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:32:50,209][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:32:50,708][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:32:51,207][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:32:51,707][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10480 tokens. +[2026-03-26 00:32:52,381][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 60.64%, Block Peak % of device VRAM: 62.47%, ΔTime: 00:00:32 +[2026-03-26 00:32:53,151][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:32:53,153][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:32:53,156][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:32:53,886][__main__][INFO] - Iteration 236 took 53s (33.57% Gen, 65.06% Train). Generation: 17s, Training: 34s. Estimated remaining time: 41h 11m 2s. Estimated total time: 44h 38m 16s. Time estimates for 10 more iterations: 8m 55s, 100 more iterations: 1h 29m 16s, 500 more iterations: 7h 26m 22s. +[2026-03-26 00:32:53,888][__main__][INFO] - Starting iteration 236. +[2026-03-26 00:32:54,288][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 23 and human policies 1. +[2026-03-26 00:32:54,289][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:32:58,128][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 00:33:06,772][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 1 y book, 10 z balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 00:33:11,087][__main__][INFO] - Number of regex retries in iteration 236: 2 +[2026-03-26 00:33:11,087][__main__][INFO] - agents played in iteration 236 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:33:11,866][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:33:11,889][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:33:11,912][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:33:11,935][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:33:11,936][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:33:11,936][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:33:12,580][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:33:13,038][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:33:13,543][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:33:14,042][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:33:14,541][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:33:15,040][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:33:15,538][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:33:16,038][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:33:16,538][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:33:17,034][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:33:17,534][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:33:18,030][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:33:18,527][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:33:19,027][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:33:19,521][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:33:20,020][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:33:20,518][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:33:21,015][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:33:21,513][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:33:22,012][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:33:22,509][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:33:23,006][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:33:23,504][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:33:24,000][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:33:24,500][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:33:25,001][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:33:25,501][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:33:25,999][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:33:26,500][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:33:26,999][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:33:27,497][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:33:27,995][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:33:28,493][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:33:28,991][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:33:29,490][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:33:29,985][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:33:30,485][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:33:30,983][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:33:31,482][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:33:31,983][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:33:32,483][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:33:32,982][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:33:33,482][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:33:33,982][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:33:34,483][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:33:34,982][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:33:35,483][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:33:35,984][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:33:36,482][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:33:36,978][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:33:37,477][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:33:37,976][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:33:38,474][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:33:38,969][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:33:39,466][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:33:39,964][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:33:40,463][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:33:40,960][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:33:41,461][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:33:41,957][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:33:42,455][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:33:42,951][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:33:43,448][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:33:43,947][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:33:44,445][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10403 tokens. +[2026-03-26 00:33:45,116][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 60.57%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:32 +[2026-03-26 00:33:45,860][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:33:45,863][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:33:45,864][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:33:46,615][__main__][INFO] - Iteration 237 took 52s (32.10% Gen, 66.46% Train). Generation: 16s, Training: 34s. Estimated remaining time: 40h 8m 16s. Estimated total time: 43h 36m 22s. Time estimates for 10 more iterations: 8m 43s, 100 more iterations: 1h 27m 12s, 500 more iterations: 7h 16m 3s. +[2026-03-26 00:33:46,617][__main__][INFO] - Starting iteration 237. +[2026-03-26 00:33:47,016][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 23 and human policies 1. +[2026-03-26 00:33:47,017][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:34:04,655][__main__][INFO] - Number of regex retries in iteration 237: 0 +[2026-03-26 00:34:04,656][__main__][INFO] - agents played in iteration 237 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:34:05,429][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:34:05,453][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:34:05,475][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:34:05,499][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:34:05,499][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:34:05,500][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:34:06,146][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:34:06,601][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:34:07,102][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:34:07,603][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:34:08,103][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:34:08,598][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:34:09,095][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:34:09,594][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:34:10,093][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:34:10,593][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:34:11,090][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:34:11,591][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:34:12,090][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:34:12,588][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:34:13,088][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:34:13,590][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:34:14,086][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:34:14,586][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:34:15,082][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:34:15,578][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:34:16,075][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:34:16,571][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:34:17,067][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:34:17,564][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:34:18,062][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:34:18,559][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:34:19,060][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:34:19,559][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:34:20,059][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:34:20,557][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:34:21,054][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:34:21,553][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:34:22,055][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:34:22,555][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:34:23,058][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:34:23,557][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:34:24,056][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:34:24,555][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:34:25,055][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:34:25,559][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:34:26,058][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:34:26,556][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:34:27,056][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:34:27,554][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:34:28,052][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:34:28,550][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:34:29,049][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:34:29,549][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:34:30,048][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:34:30,544][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:34:31,040][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:34:31,538][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:34:32,036][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:34:32,535][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:34:33,030][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:34:33,530][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:34:34,028][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:34:34,527][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:34:35,026][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:34:35,522][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:34:36,021][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:34:36,523][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:34:37,020][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:34:37,515][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:34:38,014][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10465 tokens. +[2026-03-26 00:34:38,689][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 60.61%, Block Peak % of device VRAM: 62.20%, ΔTime: 00:00:32 +[2026-03-26 00:34:39,432][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:34:39,434][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:34:39,436][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:34:40,159][__main__][INFO] - Iteration 238 took 53s (33.19% Gen, 65.44% Train). Generation: 17s, Training: 34s. Estimated remaining time: 40h 48m 11s. Estimated total time: 44h 17m 11s. Time estimates for 10 more iterations: 8m 51s, 100 more iterations: 1h 28m 34s, 500 more iterations: 7h 22m 51s. +[2026-03-26 00:34:40,162][__main__][INFO] - Starting iteration 238. +[2026-03-26 00:34:40,570][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 23 and human policies 1. +[2026-03-26 00:34:40,571][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:34:44,915][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 00:34:45,147][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 00:34:45,149][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 00:34:58,824][__main__][INFO] - Number of regex retries in iteration 238: 3 +[2026-03-26 00:34:58,825][__main__][INFO] - agents played in iteration 238 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:34:59,598][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:34:59,623][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:34:59,647][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:34:59,670][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:34:59,671][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:34:59,671][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:35:00,326][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:35:00,781][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:35:01,285][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:35:01,787][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:35:02,287][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:35:02,786][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:35:03,285][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:35:03,782][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:35:04,284][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:35:04,783][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:35:05,279][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:35:05,778][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:35:06,280][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:35:06,776][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:35:07,273][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:35:07,772][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:35:08,268][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:35:08,764][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:35:09,261][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:35:09,757][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:35:10,256][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:35:10,753][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:35:11,250][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:35:11,746][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:35:12,246][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:35:12,744][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:35:13,241][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:35:13,739][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:35:14,236][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:35:14,732][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:35:15,229][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:35:15,728][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:35:16,225][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:35:16,722][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:35:17,222][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:35:17,721][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:35:18,220][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:35:18,719][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:35:19,219][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:35:19,715][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:35:20,213][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:35:20,709][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:35:21,207][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:35:21,704][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:35:22,203][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:35:22,701][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:35:23,198][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:35:23,696][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:35:24,192][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:35:24,693][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:35:25,192][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:35:25,690][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:35:26,187][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:35:26,687][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:35:27,186][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:35:27,685][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:35:28,180][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:35:28,678][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:35:29,177][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:35:29,673][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:35:30,170][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:35:30,668][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:35:31,169][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:35:31,671][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:35:32,171][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10391 tokens. +[2026-03-26 00:35:32,847][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 60.61%, Block Peak % of device VRAM: 62.27%, ΔTime: 00:00:32 +[2026-03-26 00:35:33,595][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:35:33,597][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:35:33,599][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:35:34,326][__main__][INFO] - Iteration 239 took 53s (33.96% Gen, 64.69% Train). Generation: 18s, Training: 34s. Estimated remaining time: 41h 17m 55s. Estimated total time: 44h 47m 49s. Time estimates for 10 more iterations: 8m 57s, 100 more iterations: 1h 29m 35s, 500 more iterations: 7h 27m 58s. +[2026-03-26 00:35:34,328][__main__][INFO] - Starting iteration 239. +[2026-03-26 00:35:34,729][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 23 and human policies 1. +[2026-03-26 00:35:34,730][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:35:52,444][__main__][INFO] - Number of regex retries in iteration 239: 0 +[2026-03-26 00:35:52,445][__main__][INFO] - agents played in iteration 239 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:35:53,216][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:35:53,239][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:35:53,262][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:35:53,284][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:35:53,285][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:35:53,286][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:35:53,919][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:35:54,376][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:35:54,879][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:35:55,386][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:35:55,887][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:35:56,386][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:35:56,884][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:35:57,383][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:35:57,884][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:35:58,379][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:35:58,878][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:35:59,373][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:35:59,870][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:36:00,368][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:36:00,866][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:36:01,361][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:36:01,861][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:36:02,357][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:36:02,856][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:36:03,351][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:36:03,847][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:36:04,344][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:36:04,842][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:36:05,337][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:36:05,836][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:36:06,334][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:36:06,831][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:36:07,328][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:36:07,825][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:36:08,324][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:36:08,821][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:36:09,318][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:36:09,817][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:36:10,313][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:36:10,809][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:36:11,306][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:36:11,802][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:36:12,298][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:36:12,793][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:36:13,289][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:36:13,786][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:36:14,281][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:36:14,778][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:36:15,278][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:36:15,774][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:36:16,270][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:36:16,769][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:36:17,269][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:36:17,770][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:36:18,264][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:36:18,760][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:36:19,259][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:36:19,753][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:36:20,248][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:36:20,744][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:36:21,239][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:36:21,736][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:36:22,230][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:36:22,726][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:36:23,224][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:36:23,725][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:36:24,218][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:36:24,714][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:36:25,210][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:36:25,707][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10365 tokens. +[2026-03-26 00:36:26,382][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.46%, ΔTime: 00:00:32 +[2026-03-26 00:36:27,097][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:36:27,099][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:36:27,101][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:36:27,829][__main__][INFO] - Iteration 240 took 53s (33.36% Gen, 65.27% Train). Generation: 17s, Training: 34s. Estimated remaining time: 40h 44m 15s. Estimated total time: 44h 15m 2s. Time estimates for 10 more iterations: 8m 51s, 100 more iterations: 1h 28m 30s, 500 more iterations: 7h 22m 30s. +[2026-03-26 00:36:27,832][__main__][INFO] - Starting iteration 240. +[2026-03-26 00:36:28,231][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 23 and human policies 1. +[2026-03-26 00:36:28,231][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:36:47,006][__main__][INFO] - Number of regex retries in iteration 240: 0 +[2026-03-26 00:36:47,007][__main__][INFO] - agents played in iteration 240 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:36:47,777][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:36:47,801][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:36:47,824][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:36:47,847][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:36:47,847][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:36:47,848][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:36:48,493][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:36:48,944][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:36:49,448][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:36:49,948][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:36:50,451][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:36:50,953][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:36:51,454][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:36:51,952][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:36:52,453][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:36:52,951][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:36:53,448][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:36:53,945][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:36:54,442][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:36:54,942][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:36:55,443][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:36:55,943][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:36:56,441][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:36:56,939][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:36:57,437][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:36:57,933][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:36:58,432][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:36:58,932][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:36:59,429][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:36:59,928][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:37:00,428][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:37:00,922][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:37:01,419][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:37:01,918][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:37:02,414][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:37:02,913][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:37:03,409][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:37:03,909][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:37:04,406][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:37:04,906][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:37:05,405][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:37:05,905][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:37:06,404][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:37:06,904][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:37:07,404][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:37:07,903][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:37:08,403][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:37:08,901][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:37:09,395][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:37:09,895][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:37:10,392][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:37:10,891][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:37:11,391][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:37:11,889][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:37:12,385][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:37:12,882][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:37:13,378][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:37:13,874][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:37:14,372][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:37:14,867][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:37:15,366][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:37:15,865][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:37:16,364][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:37:16,864][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:37:17,364][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:37:17,861][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:37:18,361][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:37:18,861][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:37:19,362][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:37:19,863][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:37:20,362][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10445 tokens. +[2026-03-26 00:37:21,040][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 60.66%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 +[2026-03-26 00:37:21,782][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:37:21,784][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:37:21,786][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:37:23,375][__main__][INFO] - Iteration 241 took 55s (34.05% Gen, 63.07% Train). Generation: 18s, Training: 34s. Estimated remaining time: 42h 25m 31s. Estimated total time: 45h 57m 14s. Time estimates for 10 more iterations: 9m 11s, 100 more iterations: 1h 31m 54s, 500 more iterations: 7h 39m 32s. +[2026-03-26 00:37:23,377][__main__][INFO] - Starting iteration 241. +[2026-03-26 00:37:23,777][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 24 and human policies 1. +[2026-03-26 00:37:23,777][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:37:42,530][__main__][INFO] - Number of regex retries in iteration 241: 0 +[2026-03-26 00:37:42,531][__main__][INFO] - agents played in iteration 241 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:37:43,303][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:37:43,326][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:37:43,349][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:37:43,372][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:37:43,372][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:37:43,373][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:37:44,028][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:37:44,486][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:37:44,988][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:37:45,488][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:37:45,987][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:37:46,486][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:37:46,985][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:37:47,482][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:37:47,982][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:37:48,479][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:37:48,978][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:37:49,474][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:37:49,974][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:37:50,475][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:37:50,973][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:37:51,471][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:37:51,967][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:37:52,462][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:37:52,964][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:37:53,464][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:37:53,960][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:37:54,456][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:37:54,953][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:37:55,451][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:37:55,951][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:37:56,450][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:37:56,952][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:37:57,451][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:37:57,951][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:37:58,450][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:37:58,949][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:37:59,448][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:37:59,947][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:38:00,446][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:38:00,943][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:38:01,444][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:38:01,943][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:38:02,440][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:38:02,935][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:38:03,429][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:38:03,928][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:38:04,423][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:38:04,920][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:38:05,414][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:38:05,910][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:38:06,405][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:38:06,901][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:38:07,396][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:38:07,892][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:38:08,388][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:38:08,884][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:38:09,382][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:38:09,878][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:38:10,377][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:38:10,874][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:38:11,369][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:38:11,867][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:38:12,366][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:38:12,864][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:38:13,362][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:38:13,883][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:38:14,383][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:38:14,883][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:38:15,382][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:38:15,880][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10304 tokens. +[2026-03-26 00:38:16,574][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 60.60%, Block Peak % of device VRAM: 62.18%, ΔTime: 00:00:32 +[2026-03-26 00:38:17,336][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:38:17,338][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:38:17,340][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:38:18,064][__main__][INFO] - Iteration 242 took 54s (34.54% Gen, 64.12% Train). Generation: 18s, Training: 34s. Estimated remaining time: 41h 41m 46s. Estimated total time: 45h 14m 24s. Time estimates for 10 more iterations: 9m 2s, 100 more iterations: 1h 30m 28s, 500 more iterations: 7h 32m 24s. +[2026-03-26 00:38:18,066][__main__][INFO] - Starting iteration 242. +[2026-03-26 00:38:18,464][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 24 and human policies 1. +[2026-03-26 00:38:18,465][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:38:35,473][__main__][INFO] - Number of regex retries in iteration 242: 0 +[2026-03-26 00:38:35,474][__main__][INFO] - agents played in iteration 242 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:38:36,260][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:38:36,284][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:38:36,308][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:38:36,331][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:38:36,332][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:38:36,332][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:38:36,978][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:38:37,433][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:38:37,934][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:38:38,436][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:38:38,932][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:38:39,427][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:38:39,925][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:38:40,422][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:38:40,918][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:38:41,417][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:38:41,913][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:38:42,412][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:38:42,911][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:38:43,411][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:38:43,908][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:38:44,414][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:38:44,914][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:38:45,414][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:38:45,914][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:38:46,414][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:38:46,913][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:38:47,413][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:38:47,913][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:38:48,411][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:38:48,908][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:38:49,411][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:38:49,912][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:38:50,411][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:38:50,909][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:38:51,409][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:38:51,910][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:38:52,407][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:38:52,909][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:38:53,407][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:38:53,906][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:38:54,405][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:38:54,905][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:38:55,406][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:38:55,905][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:38:56,401][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:38:56,899][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:38:57,395][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:38:57,895][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:38:58,390][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:38:58,885][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:38:59,380][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:38:59,874][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:39:00,371][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:39:00,869][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:39:01,369][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:39:01,869][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:39:02,368][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:39:02,867][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:39:03,367][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:39:03,868][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:39:04,367][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:39:04,866][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:39:05,365][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:39:05,862][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:39:06,361][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:39:06,859][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:39:07,359][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:39:07,858][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:39:08,352][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:39:08,851][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10453 tokens. +[2026-03-26 00:39:09,553][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.40%, ΔTime: 00:00:32 +[2026-03-26 00:39:10,314][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:39:10,317][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:39:10,319][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:39:11,051][__main__][INFO] - Iteration 243 took 52s (32.34% Gen, 66.26% Train). Generation: 17s, Training: 34s. Estimated remaining time: 40h 15m 53s. Estimated total time: 43h 49m 24s. Time estimates for 10 more iterations: 8m 45s, 100 more iterations: 1h 27m 38s, 500 more iterations: 7h 18m 14s. +[2026-03-26 00:39:11,054][__main__][INFO] - Starting iteration 243. +[2026-03-26 00:39:11,455][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 24 and human policies 1. +[2026-03-26 00:39:11,456][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:39:23,669][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 00:39:29,351][__main__][INFO] - Number of regex retries in iteration 243: 1 +[2026-03-26 00:39:29,352][__main__][INFO] - agents played in iteration 243 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:39:30,141][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:39:30,164][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:39:30,187][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:39:30,210][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:39:30,211][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:39:30,211][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:39:30,856][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:39:31,307][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:39:31,811][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:39:32,308][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:39:32,806][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:39:33,302][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:39:33,799][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:39:34,294][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:39:34,789][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:39:35,287][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:39:35,783][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:39:36,280][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:39:36,777][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:39:37,278][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:39:37,777][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:39:38,276][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:39:38,772][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:39:39,271][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:39:39,769][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:39:40,268][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:39:40,771][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:39:41,270][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:39:41,770][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:39:42,271][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:39:42,769][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:39:43,266][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:39:43,765][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:39:44,261][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:39:44,762][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:39:45,259][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:39:45,757][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:39:46,253][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:39:46,749][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:39:47,247][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:39:47,746][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:39:48,243][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:39:48,739][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:39:49,239][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:39:49,739][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:39:50,259][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:39:50,757][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:39:51,254][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:39:51,750][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:39:52,249][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:39:52,746][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:39:53,247][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:39:53,744][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:39:54,242][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:39:54,737][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:39:55,235][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:39:55,730][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:39:56,229][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:39:56,726][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:39:57,224][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:39:57,723][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:39:58,219][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:39:58,716][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:39:59,216][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:39:59,711][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:40:00,208][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:40:00,707][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:40:01,205][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:40:01,702][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:40:02,200][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:40:02,701][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10252 tokens. +[2026-03-26 00:40:03,375][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 60.65%, Block Peak % of device VRAM: 62.23%, ΔTime: 00:00:32 +[2026-03-26 00:40:04,121][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:40:04,123][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:40:04,125][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:40:04,848][__main__][INFO] - Iteration 244 took 53s (33.52% Gen, 65.13% Train). Generation: 17s, Training: 34s. Estimated remaining time: 40h 55m 14s. Estimated total time: 44h 29m 38s. Time estimates for 10 more iterations: 8m 53s, 100 more iterations: 1h 28m 59s, 500 more iterations: 7h 24m 56s. +[2026-03-26 00:40:04,850][__main__][INFO] - Starting iteration 244. +[2026-03-26 00:40:05,249][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 24 and human policies 1. +[2026-03-26 00:40:05,249][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:40:12,942][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 00:40:23,557][__main__][INFO] - Number of regex retries in iteration 244: 1 +[2026-03-26 00:40:23,558][__main__][INFO] - agents played in iteration 244 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:40:24,328][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:40:24,351][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:40:24,376][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:40:24,399][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:40:24,400][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:40:24,400][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:40:25,059][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:40:25,511][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:40:26,014][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:40:26,514][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:40:27,012][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:40:27,510][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:40:28,013][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:40:28,513][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:40:29,012][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:40:29,512][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:40:30,009][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:40:30,505][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:40:31,005][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:40:31,509][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:40:32,010][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:40:32,510][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:40:33,011][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:40:33,513][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:40:34,017][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:40:34,514][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:40:35,011][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:40:35,508][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:40:36,005][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:40:36,527][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:40:37,023][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:40:37,527][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:40:38,027][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:40:38,525][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:40:39,026][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:40:39,527][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:40:40,027][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:40:40,524][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:40:41,021][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:40:41,528][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:40:42,045][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:40:42,545][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:40:43,047][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:40:43,549][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:40:44,049][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:40:44,548][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:40:45,048][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:40:45,547][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:40:46,044][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:40:46,541][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:40:47,038][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:40:47,542][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:40:48,042][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:40:48,542][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:40:49,040][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:40:49,537][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:40:50,034][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:40:50,532][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:40:51,031][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:40:51,530][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:40:52,031][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:40:52,531][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:40:53,027][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:40:53,525][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:40:54,025][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:40:54,523][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:40:55,025][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:40:55,525][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:40:56,025][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:40:56,525][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:40:57,026][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10341 tokens. +[2026-03-26 00:40:57,702][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 60.64%, Block Peak % of device VRAM: 62.22%, ΔTime: 00:00:32 +[2026-03-26 00:40:58,455][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:40:58,457][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:40:58,459][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:40:59,259][__main__][INFO] - Iteration 245 took 54s (33.90% Gen, 64.62% Train). Generation: 18s, Training: 34s. Estimated remaining time: 41h 25m 14s. Estimated total time: 45h 0m 33s. Time estimates for 10 more iterations: 9m 0s, 100 more iterations: 1h 30m 1s, 500 more iterations: 7h 30m 5s. +[2026-03-26 00:40:59,261][__main__][INFO] - Starting iteration 245. +[2026-03-26 00:40:59,658][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 24 and human policies 1. +[2026-03-26 00:40:59,659][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:41:18,166][__main__][INFO] - Number of regex retries in iteration 245: 0 +[2026-03-26 00:41:18,167][__main__][INFO] - agents played in iteration 245 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:41:18,940][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:41:18,963][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:41:18,986][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:41:19,009][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:41:19,010][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:41:19,010][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:41:19,641][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:41:20,093][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:41:20,595][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:41:21,093][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:41:21,591][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:41:22,088][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:41:22,589][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:41:23,092][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:41:23,592][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:41:24,113][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:41:24,618][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:41:25,117][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:41:25,615][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:41:26,112][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:41:26,612][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:41:27,110][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:41:27,610][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:41:28,107][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:41:28,603][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:41:29,099][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:41:29,594][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:41:30,092][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:41:30,590][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:41:31,087][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:41:31,584][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:41:32,082][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:41:32,582][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:41:33,084][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:41:33,580][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:41:34,080][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:41:34,579][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:41:35,075][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:41:35,573][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:41:36,070][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:41:36,570][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:41:37,070][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:41:37,567][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:41:38,064][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:41:38,560][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:41:39,056][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:41:39,552][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:41:40,048][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:41:40,549][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:41:41,047][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:41:41,549][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:41:42,049][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:41:42,548][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:41:43,048][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:41:43,550][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:41:44,046][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:41:44,545][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:41:45,042][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:41:45,537][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:41:46,033][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:41:46,532][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:41:47,031][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:41:47,528][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:41:48,028][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:41:48,526][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:41:49,023][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:41:49,519][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:41:50,015][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:41:50,512][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:41:51,011][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:41:51,510][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10311 tokens. +[2026-03-26 00:41:52,173][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 60.60%, Block Peak % of device VRAM: 62.33%, ΔTime: 00:00:32 +[2026-03-26 00:41:52,918][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:41:52,920][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:41:52,921][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:41:53,635][__main__][INFO] - Iteration 246 took 53s (34.29% Gen, 64.39% Train). Generation: 18s, Training: 34s. Estimated remaining time: 41h 22m 37s. Estimated total time: 44h 58m 50s. Time estimates for 10 more iterations: 8m 59s, 100 more iterations: 1h 29m 57s, 500 more iterations: 7h 29m 48s. +[2026-03-26 00:41:53,637][__main__][INFO] - Starting iteration 246. +[2026-03-26 00:41:54,037][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 24 and human policies 1. +[2026-03-26 00:41:54,038][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:42:07,060][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 1 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 00:42:11,005][__main__][INFO] - Number of regex retries in iteration 246: 1 +[2026-03-26 00:42:11,006][__main__][INFO] - agents played in iteration 246 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:42:11,784][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:42:11,807][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:42:11,830][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:42:11,853][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:42:11,853][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:42:11,854][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:42:12,493][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:42:12,947][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:42:13,455][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:42:13,955][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:42:14,456][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:42:14,952][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:42:15,452][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:42:15,950][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:42:16,452][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:42:16,953][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:42:17,454][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:42:17,953][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:42:18,450][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:42:18,955][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:42:19,460][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:42:19,967][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:42:20,471][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:42:20,971][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:42:21,470][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:42:21,969][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:42:22,467][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:42:22,964][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:42:23,463][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:42:23,964][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:42:24,463][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:42:24,959][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:42:25,455][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:42:25,953][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:42:26,451][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:42:26,947][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:42:27,444][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:42:27,944][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:42:28,440][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:42:28,939][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:42:29,437][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:42:29,940][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:42:30,439][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:42:30,938][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:42:31,455][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:42:31,954][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:42:32,455][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:42:32,951][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:42:33,447][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:42:33,945][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:42:34,445][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:42:34,941][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:42:35,437][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:42:35,933][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:42:36,431][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:42:36,928][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:42:37,424][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:42:37,923][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:42:38,418][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:42:38,914][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:42:39,412][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:42:39,910][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:42:40,409][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:42:40,904][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:42:41,403][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:42:41,898][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:42:42,396][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:42:42,893][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:42:43,390][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:42:43,887][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:42:44,384][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10343 tokens. +[2026-03-26 00:42:45,045][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.57%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 +[2026-03-26 00:42:45,790][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:42:45,792][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:42:45,794][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:42:46,520][__main__][INFO] - Iteration 247 took 52s (32.33% Gen, 66.28% Train). Generation: 16s, Training: 34s. Estimated remaining time: 40h 7m 4s. Estimated total time: 43h 44m 10s. Time estimates for 10 more iterations: 8m 44s, 100 more iterations: 1h 27m 28s, 500 more iterations: 7h 17m 21s. +[2026-03-26 00:42:46,523][__main__][INFO] - Starting iteration 247. +[2026-03-26 00:42:46,920][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 24 and human policies 1. +[2026-03-26 00:42:46,921][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:42:51,235][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 00:43:04,291][__main__][INFO] - Number of regex retries in iteration 247: 1 +[2026-03-26 00:43:04,291][__main__][INFO] - agents played in iteration 247 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:43:05,064][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:43:05,088][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:43:05,112][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:43:05,135][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:43:05,136][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:43:05,136][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:43:05,773][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:43:06,228][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:43:06,733][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:43:07,229][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:43:07,729][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:43:08,227][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:43:08,722][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:43:09,219][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:43:09,716][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:43:10,216][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:43:10,715][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:43:11,214][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:43:11,709][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:43:12,205][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:43:12,708][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:43:13,211][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:43:13,710][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:43:14,208][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:43:14,707][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:43:15,204][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:43:15,700][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:43:16,195][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:43:16,694][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:43:17,189][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:43:17,689][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:43:18,186][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:43:18,683][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:43:19,184][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:43:19,685][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:43:20,183][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:43:20,680][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:43:21,177][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:43:21,675][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:43:22,171][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:43:22,668][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:43:23,165][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:43:23,665][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:43:24,162][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:43:24,657][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:43:25,156][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:43:25,651][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:43:26,151][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:43:26,650][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:43:27,147][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:43:27,645][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:43:28,145][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:43:28,639][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:43:29,137][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:43:29,637][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:43:30,135][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:43:30,631][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:43:31,130][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:43:31,631][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:43:32,131][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:43:32,631][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:43:33,130][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:43:33,628][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:43:34,123][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:43:34,621][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:43:35,121][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:43:35,626][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:43:36,130][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:43:36,629][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:43:37,130][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:43:37,629][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10322 tokens. +[2026-03-26 00:43:38,316][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 60.50%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:32 +[2026-03-26 00:43:39,084][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:43:39,087][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:43:39,088][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:43:39,814][__main__][INFO] - Iteration 248 took 52s (32.84% Gen, 65.79% Train). Generation: 17s, Training: 34s. Estimated remaining time: 40h 26m 44s. Estimated total time: 44h 4m 43s. Time estimates for 10 more iterations: 8m 48s, 100 more iterations: 1h 28m 9s, 500 more iterations: 7h 20m 47s. +[2026-03-26 00:43:39,817][__main__][INFO] - Starting iteration 248. +[2026-03-26 00:43:40,220][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 24 and human policies 1. +[2026-03-26 00:43:40,221][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:43:58,630][__main__][INFO] - Number of regex retries in iteration 248: 0 +[2026-03-26 00:43:58,631][__main__][INFO] - agents played in iteration 248 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:43:59,416][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:43:59,439][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:43:59,462][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:43:59,485][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:43:59,486][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:43:59,486][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:44:00,130][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:44:00,588][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:44:01,092][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:44:01,590][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:44:02,092][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:44:02,593][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:44:03,092][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:44:03,589][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:44:04,084][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:44:04,583][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:44:05,085][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:44:05,581][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:44:06,078][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:44:06,575][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:44:07,074][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:44:07,575][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:44:08,075][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:44:08,573][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:44:09,074][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:44:09,571][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:44:10,071][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:44:10,571][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:44:11,070][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:44:11,570][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:44:12,069][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:44:12,568][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:44:13,066][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:44:13,565][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:44:14,062][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:44:14,557][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:44:15,053][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:44:15,553][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:44:16,049][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:44:16,549][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:44:17,047][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:44:17,545][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:44:18,041][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:44:18,537][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:44:19,036][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:44:19,533][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:44:20,031][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:44:20,528][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:44:21,026][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:44:21,525][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:44:22,021][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:44:22,522][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:44:23,020][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:44:23,519][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:44:24,014][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:44:24,509][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:44:25,005][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:44:25,502][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:44:25,999][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:44:26,495][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:44:26,994][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:44:27,490][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:44:27,987][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:44:28,484][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:44:28,985][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:44:29,480][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:44:29,979][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:44:30,474][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:44:30,974][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:44:31,473][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:44:31,976][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10228 tokens. +[2026-03-26 00:44:32,646][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 60.63%, Block Peak % of device VRAM: 62.20%, ΔTime: 00:00:32 +[2026-03-26 00:44:33,395][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:44:33,397][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:44:33,403][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:44:34,254][__main__][INFO] - Iteration 249 took 54s (34.07% Gen, 64.35% Train). Generation: 18s, Training: 34s. Estimated remaining time: 41h 22m 49s. Estimated total time: 45h 1m 43s. Time estimates for 10 more iterations: 9m 0s, 100 more iterations: 1h 30m 3s, 500 more iterations: 7h 30m 17s. +[2026-03-26 00:44:34,256][__main__][INFO] - Starting iteration 249. +[2026-03-26 00:44:34,654][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 24 and human policies 1. +[2026-03-26 00:44:34,655][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:44:52,181][__main__][INFO] - Number of regex retries in iteration 249: 0 +[2026-03-26 00:44:52,182][__main__][INFO] - agents played in iteration 249 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:44:52,948][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:44:52,971][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:44:52,994][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:44:53,017][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:44:53,018][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:44:53,018][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:44:53,654][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:44:54,110][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:44:54,614][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:44:55,113][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:44:55,612][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:44:56,109][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:44:56,609][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:44:57,108][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:44:57,608][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:44:58,106][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:44:58,604][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:44:59,106][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:44:59,605][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:45:00,106][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:45:00,604][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:45:01,103][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:45:01,603][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:45:02,100][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:45:02,596][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:45:03,096][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:45:03,592][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:45:04,087][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:45:04,585][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:45:05,080][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:45:05,579][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:45:06,084][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:45:06,587][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:45:07,088][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:45:07,609][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:45:08,111][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:45:08,611][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:45:09,113][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:45:09,612][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:45:10,108][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:45:10,605][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:45:11,100][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:45:11,595][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:45:12,092][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:45:12,587][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:45:13,084][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:45:13,580][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:45:14,075][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:45:14,571][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:45:15,069][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:45:15,564][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:45:16,060][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:45:16,556][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:45:17,052][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:45:17,551][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:45:18,051][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:45:18,547][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:45:19,045][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:45:19,544][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:45:20,037][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:45:20,535][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:45:21,034][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:45:21,530][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:45:22,026][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:45:22,524][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:45:23,020][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:45:23,517][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:45:24,017][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:45:24,515][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:45:25,012][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:45:25,508][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10369 tokens. +[2026-03-26 00:45:26,183][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.47%, ΔTime: 00:00:32 +[2026-03-26 00:45:26,926][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:45:26,929][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:45:26,930][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:45:27,650][__main__][INFO] - Iteration 250 took 52s (33.07% Gen, 65.57% Train). Generation: 17s, Training: 34s. Estimated remaining time: 40h 30m 1s. Estimated total time: 44h 9m 48s. Time estimates for 10 more iterations: 8m 49s, 100 more iterations: 1h 28m 19s, 500 more iterations: 7h 21m 38s. +[2026-03-26 00:45:27,652][__main__][INFO] - Starting iteration 250. +[2026-03-26 00:45:28,050][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 24 and human policies 1. +[2026-03-26 00:45:28,051][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:45:33,909][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 00:45:46,595][__main__][INFO] - Number of regex retries in iteration 250: 1 +[2026-03-26 00:45:46,596][__main__][INFO] - agents played in iteration 250 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:45:47,366][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:45:47,390][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:45:47,415][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:45:47,438][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:45:47,439][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:45:47,439][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:45:48,081][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:45:48,539][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:45:49,041][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:45:49,542][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:45:50,042][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:45:50,540][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:45:51,040][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:45:51,538][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:45:52,034][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:45:52,535][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:45:53,034][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:45:53,533][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:45:54,037][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:45:54,532][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:45:55,028][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:45:55,526][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:45:56,026][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:45:56,527][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:45:57,025][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:45:57,525][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:45:58,025][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:45:58,525][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:45:59,021][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:45:59,519][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:46:00,017][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:46:00,512][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:46:01,009][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:46:01,507][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:46:02,007][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:46:02,505][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:46:03,002][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:46:03,499][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:46:03,995][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:46:04,492][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:46:04,992][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:46:05,487][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:46:05,987][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:46:06,486][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:46:06,986][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:46:07,483][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:46:07,980][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:46:08,474][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:46:08,969][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:46:09,465][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:46:09,959][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:46:10,459][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:46:10,954][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:46:11,452][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:46:11,947][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:46:12,443][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:46:12,938][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:46:13,434][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:46:13,932][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:46:14,427][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:46:14,922][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:46:15,421][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:46:15,918][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:46:16,417][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:46:16,913][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:46:17,409][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:46:17,905][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:46:18,404][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:46:18,903][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:46:19,398][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:46:19,894][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10287 tokens. +[2026-03-26 00:46:20,563][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 60.54%, Block Peak % of device VRAM: 62.23%, ΔTime: 00:00:32 +[2026-03-26 00:46:21,309][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:46:21,311][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:46:21,313][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:46:22,670][__main__][INFO] - Iteration 251 took 54s (33.95% Gen, 63.56% Train). Generation: 18s, Training: 34s. Estimated remaining time: 41h 50m 19s. Estimated total time: 45h 31m 1s. Time estimates for 10 more iterations: 9m 6s, 100 more iterations: 1h 31m 2s, 500 more iterations: 7h 35m 10s. +[2026-03-26 00:46:22,672][__main__][INFO] - Starting iteration 251. +[2026-03-26 00:46:23,070][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 25 and human policies 1. +[2026-03-26 00:46:23,071][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:46:41,270][__main__][INFO] - Number of regex retries in iteration 251: 0 +[2026-03-26 00:46:41,270][__main__][INFO] - agents played in iteration 251 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:46:42,035][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:46:42,058][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:46:42,082][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:46:42,105][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:46:42,105][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:46:42,106][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:46:42,749][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:46:43,205][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:46:43,710][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:46:44,210][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:46:44,708][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:46:45,209][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:46:45,708][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:46:46,208][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:46:46,708][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:46:47,210][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:46:47,711][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:46:48,210][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:46:48,715][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:46:49,216][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:46:49,717][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:46:50,220][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:46:50,722][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:46:51,221][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:46:51,716][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:46:52,212][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:46:52,708][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:46:53,203][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:46:53,701][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:46:54,195][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:46:54,692][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:46:55,192][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:46:55,693][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:46:56,193][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:46:56,693][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:46:57,194][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:46:57,694][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:46:58,191][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:46:58,693][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:46:59,195][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:46:59,694][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:47:00,192][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:47:00,692][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:47:01,191][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:47:01,691][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:47:02,193][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:47:02,695][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:47:03,194][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:47:03,690][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:47:04,188][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:47:04,686][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:47:05,185][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:47:05,683][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:47:06,185][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:47:06,683][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:47:07,179][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:47:07,678][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:47:08,172][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:47:08,669][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:47:09,168][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:47:09,665][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:47:10,162][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:47:10,657][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:47:11,157][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:47:11,658][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:47:12,162][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:47:12,662][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:47:13,161][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:47:13,661][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:47:14,159][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:47:14,657][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10381 tokens. +[2026-03-26 00:47:15,339][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 60.56%, Block Peak % of device VRAM: 62.46%, ΔTime: 00:00:32 +[2026-03-26 00:47:16,082][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:47:16,085][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:47:16,086][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:47:16,836][__main__][INFO] - Iteration 252 took 53s (33.85% Gen, 64.75% Train). Generation: 18s, Training: 34s. Estimated remaining time: 41h 6m 41s. Estimated total time: 44h 48m 18s. Time estimates for 10 more iterations: 8m 57s, 100 more iterations: 1h 29m 36s, 500 more iterations: 7h 28m 3s. +[2026-03-26 00:47:16,838][__main__][INFO] - Starting iteration 252. +[2026-03-26 00:47:17,239][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 25 and human policies 1. +[2026-03-26 00:47:17,239][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:47:21,703][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 00:47:28,579][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 1 books, 1 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 00:47:35,227][__main__][INFO] - Number of regex retries in iteration 252: 2 +[2026-03-26 00:47:35,228][__main__][INFO] - agents played in iteration 252 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:47:35,996][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:47:36,019][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:47:36,042][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:47:36,065][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:47:36,065][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:47:36,066][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:47:36,711][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:47:37,169][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:47:37,674][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:47:38,177][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:47:38,681][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:47:39,182][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:47:39,682][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:47:40,182][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:47:40,679][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:47:41,183][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:47:41,682][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:47:42,180][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:47:42,680][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:47:43,176][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:47:43,673][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:47:44,177][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:47:44,678][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:47:45,174][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:47:45,672][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:47:46,170][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:47:46,668][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:47:47,164][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:47:47,660][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:47:48,160][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:47:48,656][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:47:49,152][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:47:49,649][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:47:50,149][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:47:50,646][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:47:51,143][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:47:51,641][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:47:52,138][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:47:52,634][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:47:53,136][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:47:53,633][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:47:54,131][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:47:54,629][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:47:55,127][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:47:55,625][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:47:56,121][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:47:56,619][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:47:57,119][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:47:57,615][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:47:58,112][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:47:58,610][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:47:59,109][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:47:59,606][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:48:00,104][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:48:00,603][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:48:01,106][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:48:01,604][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:48:02,104][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:48:02,606][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:48:03,103][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:48:03,602][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:48:04,101][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:48:04,601][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:48:05,097][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:48:05,594][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:48:06,091][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:48:06,594][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:48:07,093][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:48:07,591][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:48:08,089][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:48:08,584][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10264 tokens. +[2026-03-26 00:48:09,255][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 60.56%, Block Peak % of device VRAM: 62.28%, ΔTime: 00:00:32 +[2026-03-26 00:48:09,999][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:48:10,001][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:48:10,003][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:48:10,776][__main__][INFO] - Iteration 253 took 53s (33.60% Gen, 64.95% Train). Generation: 17s, Training: 34s. Estimated remaining time: 40h 54m 23s. Estimated total time: 44h 36m 53s. Time estimates for 10 more iterations: 8m 55s, 100 more iterations: 1h 29m 13s, 500 more iterations: 7h 26m 8s. +[2026-03-26 00:48:10,778][__main__][INFO] - Starting iteration 253. +[2026-03-26 00:48:11,178][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 25 and human policies 1. +[2026-03-26 00:48:11,178][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:48:36,924][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 00:48:37,751][__main__][INFO] - Number of regex retries in iteration 253: 1 +[2026-03-26 00:48:37,752][__main__][INFO] - agents played in iteration 253 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:48:38,529][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:48:38,552][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:48:38,575][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:48:38,598][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:48:38,599][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:48:38,599][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:48:39,251][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:48:39,708][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:48:40,210][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:48:40,709][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:48:41,206][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:48:41,704][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:48:42,258][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:48:42,762][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:48:43,264][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:48:43,766][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:48:44,263][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:48:44,762][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:48:45,259][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:48:45,756][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:48:46,253][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:48:46,751][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:48:47,248][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:48:47,749][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:48:48,247][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:48:48,745][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:48:49,241][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:48:49,740][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:48:50,237][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:48:50,733][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:48:51,231][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:48:51,730][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:48:52,229][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:48:52,726][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:48:53,768][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:48:54,264][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:48:54,761][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:48:55,259][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:48:55,757][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:48:56,257][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:48:56,757][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:48:57,256][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:48:57,756][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:48:58,254][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:48:58,751][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:48:59,249][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:48:59,745][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:49:00,242][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:49:00,739][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:49:01,236][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:49:01,734][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:49:02,230][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:49:02,729][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:49:03,227][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:49:03,725][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:49:04,223][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:49:04,719][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:49:05,216][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:49:05,712][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:49:06,209][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:49:06,707][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:49:07,204][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:49:07,704][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:49:08,204][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:49:08,703][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:49:09,202][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:49:09,699][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:49:10,195][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:49:10,693][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:49:11,197][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:49:11,724][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10308 tokens. +[2026-03-26 00:49:12,423][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.26%, Current % of VRAM taken: 60.69%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:33 +[2026-03-26 00:49:13,230][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:49:13,233][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:49:13,234][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:49:13,927][__main__][INFO] - Iteration 254 took 1m 2s (42.35% Gen, 56.55% Train). Generation: 26s, Training: 35s. Estimated remaining time: 48h 33m 55s. Estimated total time: 52h 17m 28s. Time estimates for 10 more iterations: 10m 27s, 100 more iterations: 1h 44m 34s, 500 more iterations: 8h 42m 54s. +[2026-03-26 00:49:13,933][__main__][INFO] - Starting iteration 254. +[2026-03-26 00:49:14,331][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 25 and human policies 1. +[2026-03-26 00:49:14,332][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:49:35,279][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 00:49:39,858][__main__][INFO] - Number of regex retries in iteration 254: 1 +[2026-03-26 00:49:39,858][__main__][INFO] - agents played in iteration 254 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:49:40,864][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:49:40,887][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:49:40,911][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:49:40,934][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:49:40,934][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:49:40,935][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:49:41,634][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:49:42,091][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:49:42,595][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:49:43,091][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:49:43,588][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:49:44,085][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:49:44,582][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:49:45,080][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:49:45,578][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:49:46,078][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:49:46,577][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:49:47,074][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:49:47,572][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:49:48,071][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:49:48,572][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:49:49,071][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:49:49,569][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:49:50,073][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:49:50,570][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:49:51,072][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:49:51,596][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:49:52,096][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:49:52,594][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:49:53,095][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:49:53,592][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:49:54,090][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:49:54,590][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:49:55,089][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:49:55,588][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:49:56,091][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:49:56,588][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:49:57,085][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:49:57,585][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:49:58,082][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:49:58,579][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:49:59,077][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:49:59,576][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:50:00,074][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:50:00,571][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:50:01,068][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:50:01,566][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:50:02,072][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:50:02,572][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:50:03,073][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:50:03,574][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:50:04,074][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:50:04,575][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:50:05,074][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:50:05,574][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:50:06,070][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:50:06,570][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:50:07,069][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:50:07,565][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:50:08,061][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:50:08,561][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:50:09,058][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:50:09,554][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:50:10,054][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:50:10,552][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:50:11,051][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:50:11,550][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:50:12,048][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:50:12,546][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:50:13,042][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:50:13,544][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10168 tokens. +[2026-03-26 00:50:14,232][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 60.65%, Block Peak % of device VRAM: 62.10%, ΔTime: 00:00:32 +[2026-03-26 00:50:15,053][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:50:15,056][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:50:15,057][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:50:15,795][__main__][INFO] - Iteration 255 took 1m 1s (41.53% Gen, 57.27% Train). Generation: 25s, Training: 35s. Estimated remaining time: 47h 28m 36s. Estimated total time: 51h 13m 11s. Time estimates for 10 more iterations: 10m 14s, 100 more iterations: 1h 42m 26s, 500 more iterations: 8h 32m 11s. +[2026-03-26 00:50:15,797][__main__][INFO] - Starting iteration 255. +[2026-03-26 00:50:16,198][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 25 and human policies 1. +[2026-03-26 00:50:16,199][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:50:24,326][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 00:50:35,164][__main__][INFO] - Number of regex retries in iteration 255: 1 +[2026-03-26 00:50:35,165][__main__][INFO] - agents played in iteration 255 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:50:35,924][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:50:35,947][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:50:35,970][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:50:35,993][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:50:35,994][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:50:35,994][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:50:36,634][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:50:37,088][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:50:37,592][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:50:38,091][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:50:38,592][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:50:39,092][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:50:39,588][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:50:40,091][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:50:40,590][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:50:41,089][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:50:41,589][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:50:42,088][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:50:42,586][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:50:43,084][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:50:43,582][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:50:44,080][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:50:44,584][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:50:45,081][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:50:45,576][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:50:46,073][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:50:46,569][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:50:47,063][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:50:47,558][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:50:48,056][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:50:48,552][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:50:49,049][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:50:49,548][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:50:50,047][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:50:50,546][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:50:51,043][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:50:51,562][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:50:52,061][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:50:52,560][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:50:53,056][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:50:53,555][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:50:54,052][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:50:54,549][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:50:55,047][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:50:55,543][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:50:56,044][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:50:56,539][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:50:57,035][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:50:57,531][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:50:58,028][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:50:58,524][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:50:59,018][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:50:59,513][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:51:00,009][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:51:00,505][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:51:01,004][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:51:01,505][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:51:02,002][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:51:02,502][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:51:03,003][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:51:03,502][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:51:03,998][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:51:04,497][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:51:04,993][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:51:05,491][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:51:05,989][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:51:06,490][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:51:06,989][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:51:07,485][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:51:07,981][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:51:08,478][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10242 tokens. +[2026-03-26 00:51:09,149][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.58%, Block Peak % of device VRAM: 62.24%, ΔTime: 00:00:32 +[2026-03-26 00:51:09,895][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:51:09,897][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:51:09,898][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:51:10,665][__main__][INFO] - Iteration 256 took 54s (34.82% Gen, 63.77% Train). Generation: 18s, Training: 34s. Estimated remaining time: 41h 37m 52s. Estimated total time: 45h 23m 23s. Time estimates for 10 more iterations: 9m 4s, 100 more iterations: 1h 30m 46s, 500 more iterations: 7h 33m 53s. +[2026-03-26 00:51:10,667][__main__][INFO] - Starting iteration 256. +[2026-03-26 00:51:11,066][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 25 and human policies 1. +[2026-03-26 00:51:11,066][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:51:29,755][__main__][INFO] - Number of regex retries in iteration 256: 0 +[2026-03-26 00:51:29,756][__main__][INFO] - agents played in iteration 256 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:51:30,513][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:51:30,536][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:51:30,559][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:51:30,582][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:51:30,583][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:51:30,583][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:51:31,243][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:51:31,697][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:51:32,199][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:51:32,697][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:51:33,200][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:51:33,699][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:51:34,194][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:51:34,692][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:51:35,189][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:51:35,687][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:51:36,184][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:51:36,682][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:51:37,179][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:51:37,676][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:51:38,173][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:51:38,672][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:51:39,169][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:51:39,690][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:51:40,193][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:51:40,690][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:51:41,192][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:51:41,693][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:51:42,190][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:51:42,691][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:51:43,190][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:51:43,686][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:51:44,186][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:51:44,684][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:51:45,183][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:51:45,678][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:51:46,176][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:51:46,675][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:51:47,173][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:51:47,676][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:51:48,175][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:51:48,674][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:51:49,174][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:51:49,674][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:51:50,176][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:51:50,676][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:51:51,176][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:51:51,675][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:51:52,175][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:51:52,674][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:51:53,171][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:51:53,669][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:51:54,167][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:51:54,669][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:51:55,168][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:51:55,668][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:51:56,165][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:51:56,664][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:51:57,160][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:51:57,660][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:51:58,158][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:51:58,655][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:51:59,151][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:51:59,648][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:52:00,143][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:52:00,639][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:52:01,136][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:52:01,632][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:52:02,130][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:52:02,629][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:52:03,128][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10276 tokens. +[2026-03-26 00:52:03,819][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.57%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:32 +[2026-03-26 00:52:04,562][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:52:04,564][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:52:04,566][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:52:05,403][__main__][INFO] - Iteration 257 took 54s (34.40% Gen, 64.06% Train). Generation: 18s, Training: 34s. Estimated remaining time: 41h 30m 29s. Estimated total time: 45h 16m 54s. Time estimates for 10 more iterations: 9m 3s, 100 more iterations: 1h 30m 33s, 500 more iterations: 7h 32m 49s. +[2026-03-26 00:52:05,407][__main__][INFO] - Starting iteration 257. +[2026-03-26 00:52:05,806][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 25 and human policies 1. +[2026-03-26 00:52:05,807][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:52:25,295][__main__][INFO] - Number of regex retries in iteration 257: 0 +[2026-03-26 00:52:25,296][__main__][INFO] - agents played in iteration 257 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:52:26,065][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:52:26,088][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:52:26,111][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:52:26,134][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:52:26,134][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:52:26,135][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:52:26,777][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:52:27,228][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:52:27,726][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:52:28,223][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:52:28,722][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:52:29,218][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:52:29,713][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:52:30,211][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:52:30,707][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:52:31,202][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:52:31,699][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:52:32,195][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:52:32,691][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:52:33,189][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:52:33,692][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:52:34,187][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:52:34,692][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:52:35,213][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:52:35,712][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:52:36,213][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:52:36,714][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:52:37,214][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:52:37,712][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:52:38,213][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:52:38,714][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:52:39,210][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:52:39,707][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:52:40,203][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:52:40,699][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:52:41,196][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:52:41,692][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:52:42,191][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:52:42,688][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:52:43,187][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:52:43,685][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:52:44,184][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:52:44,683][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:52:45,182][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:52:45,681][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:52:46,179][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:52:46,678][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:52:47,178][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:52:47,681][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:52:48,179][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:52:48,678][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:52:49,178][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:52:49,679][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:52:50,179][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:52:50,678][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:52:51,177][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:52:51,675][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:52:52,174][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:52:52,671][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:52:53,168][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:52:53,665][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:52:54,166][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:52:54,666][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:52:55,161][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:52:55,660][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:52:56,158][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:52:56,657][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:52:57,153][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:52:57,651][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:52:58,150][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:52:58,647][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10353 tokens. +[2026-03-26 00:52:59,322][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.36%, ΔTime: 00:00:32 +[2026-03-26 00:53:00,079][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:53:00,081][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:53:00,082][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:53:00,922][__main__][INFO] - Iteration 258 took 55s (35.36% Gen, 63.12% Train). Generation: 19s, Training: 34s. Estimated remaining time: 42h 8m 26s. Estimated total time: 45h 55m 47s. Time estimates for 10 more iterations: 9m 11s, 100 more iterations: 1h 31m 51s, 500 more iterations: 7h 39m 17s. +[2026-03-26 00:53:00,924][__main__][INFO] - Starting iteration 258. +[2026-03-26 00:53:01,324][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 25 and human policies 1. +[2026-03-26 00:53:01,324][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:53:18,479][__main__][INFO] - Number of regex retries in iteration 258: 0 +[2026-03-26 00:53:18,480][__main__][INFO] - agents played in iteration 258 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:53:19,243][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:53:19,266][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:53:19,290][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:53:19,312][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:53:19,313][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:53:19,314][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:53:19,976][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:53:20,432][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:53:20,936][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:53:21,434][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:53:21,934][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:53:22,434][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:53:22,934][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:53:23,437][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:53:23,936][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:53:24,433][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:53:24,930][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:53:25,426][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:53:25,923][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:53:26,420][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:53:26,919][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:53:27,417][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:53:27,913][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:53:28,409][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:53:28,909][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:53:29,409][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:53:29,909][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:53:30,409][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:53:30,909][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:53:31,407][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:53:31,908][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:53:32,406][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:53:32,903][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:53:33,398][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:53:33,897][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:53:34,393][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:53:34,890][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:53:35,387][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:53:35,883][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:53:36,379][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:53:36,876][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:53:37,373][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:53:37,877][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:53:38,376][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:53:38,874][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:53:39,375][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:53:39,875][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:53:40,374][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:53:40,871][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:53:41,370][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:53:41,870][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:53:42,368][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:53:42,866][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:53:43,363][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:53:43,863][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:53:44,360][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:53:44,857][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:53:45,356][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:53:45,853][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:53:46,349][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:53:46,845][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:53:47,341][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:53:47,841][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:53:48,337][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:53:48,836][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:53:49,337][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:53:49,833][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:53:50,331][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:53:50,829][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:53:51,327][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:53:51,825][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10319 tokens. +[2026-03-26 00:53:52,513][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.58%, Block Peak % of device VRAM: 62.24%, ΔTime: 00:00:32 +[2026-03-26 00:53:53,269][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:53:53,271][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:53:53,273][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:53:54,007][__main__][INFO] - Iteration 259 took 52s (32.56% Gen, 66.04% Train). Generation: 17s, Training: 34s. Estimated remaining time: 40h 5m 58s. Estimated total time: 43h 54m 12s. Time estimates for 10 more iterations: 8m 46s, 100 more iterations: 1h 27m 48s, 500 more iterations: 7h 19m 2s. +[2026-03-26 00:53:54,009][__main__][INFO] - Starting iteration 259. +[2026-03-26 00:53:54,408][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 25 and human policies 1. +[2026-03-26 00:53:54,408][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:54:11,254][__main__][INFO] - Number of regex retries in iteration 259: 0 +[2026-03-26 00:54:11,255][__main__][INFO] - agents played in iteration 259 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:54:12,033][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:54:12,057][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:54:12,080][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:54:12,103][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:54:12,104][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:54:12,104][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:54:12,756][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:54:13,212][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:54:13,716][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:54:14,216][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:54:14,718][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:54:15,215][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:54:15,713][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:54:16,215][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:54:16,713][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:54:17,214][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:54:17,711][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:54:18,207][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:54:18,705][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:54:19,201][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:54:19,701][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:54:20,198][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:54:20,694][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:54:21,191][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:54:21,688][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:54:22,190][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:54:22,692][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:54:23,191][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:54:23,689][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:54:24,187][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:54:24,683][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:54:25,180][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:54:25,680][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:54:26,176][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:54:26,672][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:54:27,171][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:54:27,671][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:54:28,169][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:54:28,667][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:54:29,188][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:54:29,685][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:54:30,182][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:54:30,684][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:54:31,187][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:54:31,686][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:54:32,185][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:54:32,686][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:54:33,185][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:54:33,686][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:54:34,183][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:54:34,684][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:54:35,182][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:54:35,679][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:54:36,180][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:54:36,678][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:54:37,176][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:54:37,675][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:54:38,172][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:54:38,670][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:54:39,171][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:54:39,667][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:54:40,166][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:54:40,662][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:54:41,159][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:54:41,658][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:54:42,158][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:54:42,655][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:54:43,156][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:54:43,654][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:54:44,154][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:54:44,651][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10279 tokens. +[2026-03-26 00:54:45,330][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.15%, ΔTime: 00:00:32 +[2026-03-26 00:54:46,083][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:54:46,085][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:54:46,086][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:54:46,814][__main__][INFO] - Iteration 260 took 52s (32.15% Gen, 66.46% Train). Generation: 16s, Training: 34s. Estimated remaining time: 39h 51m 13s. Estimated total time: 43h 40m 19s. Time estimates for 10 more iterations: 8m 44s, 100 more iterations: 1h 27m 20s, 500 more iterations: 7h 16m 43s. +[2026-03-26 00:54:46,817][__main__][INFO] - Starting iteration 260. +[2026-03-26 00:54:47,216][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 25 and human policies 1. +[2026-03-26 00:54:47,217][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:54:51,511][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 00:54:51,770][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 00:55:05,497][__main__][INFO] - Number of regex retries in iteration 260: 2 +[2026-03-26 00:55:05,498][__main__][INFO] - agents played in iteration 260 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:55:06,283][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:55:06,307][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:55:06,332][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:55:06,357][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:55:06,357][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:55:06,358][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:55:07,017][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:55:07,469][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:55:07,977][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:55:08,478][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:55:08,974][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:55:09,472][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:55:09,981][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:55:10,482][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:55:10,979][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:55:11,479][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:55:11,980][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:55:12,478][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:55:12,977][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:55:13,475][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:55:13,977][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:55:14,479][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:55:14,976][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:55:15,476][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:55:15,971][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:55:16,469][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:55:16,971][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:55:17,468][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:55:17,965][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:55:18,462][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:55:18,958][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:55:19,454][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:55:19,951][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:55:20,454][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:55:20,952][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:55:21,449][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:55:21,946][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:55:22,441][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:55:22,940][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:55:23,438][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:55:23,937][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:55:24,436][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:55:24,936][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:55:25,436][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:55:25,935][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:55:26,431][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:55:26,928][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:55:27,424][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:55:27,919][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:55:28,416][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:55:28,912][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:55:29,408][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:55:29,906][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:55:30,402][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:55:30,900][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:55:31,397][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:55:31,896][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:55:32,396][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:55:32,894][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:55:33,395][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:55:33,896][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:55:34,394][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:55:34,892][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:55:35,393][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:55:35,890][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:55:36,389][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:55:36,887][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:55:37,386][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:55:37,887][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:55:38,383][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:55:38,883][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10272 tokens. +[2026-03-26 00:55:39,563][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 60.63%, Block Peak % of device VRAM: 62.10%, ΔTime: 00:00:32 +[2026-03-26 00:55:40,320][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:55:40,323][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:55:40,324][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:55:41,641][__main__][INFO] - Iteration 261 took 54s (33.59% Gen, 63.99% Train). Generation: 18s, Training: 34s. Estimated remaining time: 41h 31m 15s. Estimated total time: 45h 21m 17s. Time estimates for 10 more iterations: 9m 4s, 100 more iterations: 1h 30m 42s, 500 more iterations: 7h 33m 32s. +[2026-03-26 00:55:41,644][__main__][INFO] - Starting iteration 261. +[2026-03-26 00:55:42,045][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 26 and human policies 1. +[2026-03-26 00:55:42,046][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:55:52,595][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls -By proposing to take all 10 hats, I secure 10 hats, maximizing the points given my high valuation of hats and Bob's low valuation of them. did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:11:26,662][__main__][INFO] - Number of regex retries in iteration 153: 2 -[2025-11-09 20:11:26,663][__main__][INFO] - agents played in iteration 153 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:11:27,439][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:11:27,463][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:11:27,488][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:11:27,509][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:11:27,510][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:11:27,510][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:11:28,167][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:11:28,626][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:11:29,132][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:11:29,840][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:11:30,342][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:11:30,842][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:11:31,345][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:11:31,842][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:11:32,344][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:11:32,842][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:11:33,342][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:11:33,842][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:11:34,339][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:11:34,838][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:11:35,338][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:11:35,838][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:11:36,336][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:11:36,835][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:11:37,333][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:11:37,832][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:11:38,330][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:11:38,826][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:11:39,324][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:11:39,821][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:11:40,320][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:11:40,821][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:11:41,319][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:11:41,818][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:11:42,317][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:11:42,819][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:11:43,319][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:11:43,818][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:11:44,315][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:11:44,813][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:11:45,315][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:11:45,815][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:11:46,314][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:11:46,816][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:11:47,314][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:11:47,814][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:11:48,312][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:11:48,814][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:11:49,316][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:11:49,816][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:11:50,315][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:11:50,819][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:11:51,318][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:11:51,817][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:11:52,317][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:11:52,817][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:11:53,320][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:11:53,820][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:11:54,320][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:11:54,820][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:11:55,321][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:11:55,824][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:11:56,327][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:11:56,827][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:11:57,330][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:11:57,831][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:11:58,332][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:11:58,832][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:11:59,331][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:11:59,832][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:12:00,333][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10851 tokens. -[2025-11-09 20:12:01,044][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 55.71%, Block Peak % of device VRAM: 62.35%, ΔTime: 00:00:32 -[2025-11-09 20:12:01,769][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:12:01,771][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:12:01,772][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:12:02,703][__main__][INFO] - Iteration 154 took 53s (32.06% Gen, 66.18% Train). Generation: 17s, Training: 35s. Estimated remaining time: 41h 59m 35s. Estimated total time: 44h 12m 41s. Time estimates for 10 more iterations: 8m 50s, 100 more iterations: 1h 28m 25s, 500 more iterations: 7h 22m 6s. -[2025-11-09 20:12:02,705][__main__][INFO] - Starting iteration 154. -[2025-11-09 20:12:03,169][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 15 and human policies 1. -[2025-11-09 20:12:03,169][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:12:15,350][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 1, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:12:17,854][__main__][INFO] - Number of regex retries in iteration 154: 1 -[2025-11-09 20:12:17,855][__main__][INFO] - agents played in iteration 154 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:12:18,660][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:12:18,682][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:12:18,705][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:12:18,727][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:12:18,728][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:12:18,729][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:12:19,386][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:12:19,844][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:12:20,346][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:12:20,848][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:12:21,345][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:12:21,840][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:12:22,340][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:12:22,838][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:12:23,334][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:12:23,839][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:12:24,338][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:12:24,841][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:12:25,344][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:12:25,845][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:12:26,343][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:12:26,845][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:12:27,345][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:12:27,840][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:12:28,339][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:12:28,841][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:12:29,339][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:12:29,837][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:12:30,337][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:12:30,835][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:12:31,333][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:12:31,832][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:12:32,334][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:12:32,837][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:12:33,341][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:12:33,842][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:12:34,340][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:12:34,837][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:12:35,336][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:12:35,833][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:12:36,326][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:12:36,823][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:12:37,319][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:12:37,813][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:12:38,309][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:12:38,803][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:12:39,309][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:12:39,805][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:12:40,302][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:12:40,798][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:12:41,295][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:12:41,792][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:12:42,291][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:12:42,788][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:12:43,294][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:12:43,792][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:12:44,291][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:12:44,793][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:12:45,292][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:12:45,791][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:12:46,290][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:12:46,791][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:12:47,286][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:12:47,791][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:12:48,292][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:12:48,803][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:12:49,304][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:12:49,807][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:12:50,308][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:12:50,809][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:12:51,310][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10863 tokens. -[2025-11-09 20:12:51,980][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.26%, Current % of VRAM taken: 55.76%, Block Peak % of device VRAM: 62.38%, ΔTime: 00:00:32 -[2025-11-09 20:12:52,717][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:12:52,718][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:12:52,720][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:12:53,667][__main__][INFO] - Iteration 155 took 50s (29.08% Gen, 69.04% Train). Generation: 14s, Training: 34s. Estimated remaining time: 39h 50m 58s. Estimated total time: 42h 4m 55s. Time estimates for 10 more iterations: 8m 24s, 100 more iterations: 1h 24m 9s, 500 more iterations: 7h 0m 49s. -[2025-11-09 20:12:53,669][__main__][INFO] - Starting iteration 155. -[2025-11-09 20:12:54,143][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 15 and human policies 1. -[2025-11-09 20:12:54,144][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:13:07,167][__main__][INFO] - Number of regex retries in iteration 155: 0 -[2025-11-09 20:13:07,168][__main__][INFO] - agents played in iteration 155 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:13:08,013][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:13:08,041][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:13:08,067][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:13:08,091][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:13:08,091][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:13:08,092][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:13:08,731][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:13:09,191][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:13:09,696][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:13:10,195][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:13:10,694][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:13:11,212][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:13:11,711][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:13:12,209][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:13:12,707][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:13:13,208][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:13:13,718][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:13:14,218][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:13:14,718][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:13:15,217][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:13:15,718][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:13:16,218][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:13:16,719][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:13:17,218][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:13:17,717][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:13:18,219][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:13:18,717][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:13:19,215][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:13:19,713][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:13:20,212][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:13:20,711][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:13:21,208][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:13:21,708][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:13:22,210][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:13:22,707][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:13:23,204][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:13:23,701][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:13:24,202][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:13:24,700][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:13:25,205][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:13:25,701][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:13:26,198][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:13:26,697][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:13:27,194][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:13:27,689][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:13:28,190][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:13:28,687][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:13:29,197][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:13:29,696][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:13:30,192][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:13:30,690][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:13:31,186][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:13:31,683][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:13:32,181][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:13:32,679][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:13:33,221][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:13:33,720][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:13:34,220][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:13:34,715][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:13:35,217][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:13:35,719][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:13:36,222][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:13:36,725][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:13:37,226][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:13:37,725][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:13:38,225][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:13:38,722][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:13:39,225][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:13:39,726][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:13:40,227][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:13:40,728][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10868 tokens. -[2025-11-09 20:13:41,436][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.29%, Current % of VRAM taken: 55.78%, Block Peak % of device VRAM: 62.61%, ΔTime: 00:00:32 -[2025-11-09 20:13:42,207][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:13:42,209][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:13:42,210][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:13:43,146][__main__][INFO] - Iteration 156 took 49s (26.58% Gen, 71.51% Train). Generation: 13s, Training: 35s. Estimated remaining time: 38h 35m 26s. Estimated total time: 40h 50m 12s. Time estimates for 10 more iterations: 8m 10s, 100 more iterations: 1h 21m 40s, 500 more iterations: 6h 48m 22s. -[2025-11-09 20:13:43,148][__main__][INFO] - Starting iteration 156. -[2025-11-09 20:13:43,613][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 15 and human policies 1. -[2025-11-09 20:13:43,614][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:13:57,541][__main__][INFO] - Number of regex retries in iteration 156: 0 -[2025-11-09 20:13:57,542][__main__][INFO] - agents played in iteration 156 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:13:58,362][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.67%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:13:58,389][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.67%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:13:58,415][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.67%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:13:58,438][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.67%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:13:58,439][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:13:58,440][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:13:59,102][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:13:59,557][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:14:00,062][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:14:00,562][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:14:01,069][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:14:01,569][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:14:02,068][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:14:02,565][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:14:03,061][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:14:03,560][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:14:04,063][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:14:04,561][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:14:05,063][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:14:05,561][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:14:06,060][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:14:06,559][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:14:07,058][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:14:07,560][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:14:08,058][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:14:08,558][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:14:09,054][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:14:09,553][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:14:10,052][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:14:10,550][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:14:11,049][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:14:11,547][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:14:12,045][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:14:12,545][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:14:13,043][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:14:13,541][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:14:14,043][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:14:14,542][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:14:15,041][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:14:15,542][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:14:16,040][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:14:16,537][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:14:17,035][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:14:17,533][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:14:18,033][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:14:18,531][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:14:19,030][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:14:19,527][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:14:20,023][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:14:20,525][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:14:21,021][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:14:21,518][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:14:22,020][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:14:22,517][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:14:23,015][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:14:23,513][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:14:24,010][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:14:24,516][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:14:25,014][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:14:25,512][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:14:26,013][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:14:26,509][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:14:27,007][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:14:27,505][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:14:28,007][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:14:28,525][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:14:29,024][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:14:29,523][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:14:30,022][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:14:30,520][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:14:31,021][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10865 tokens. -[2025-11-09 20:14:31,736][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 55.65%, Block Peak % of device VRAM: 62.20%, ΔTime: 00:00:32 -[2025-11-09 20:14:32,478][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:14:32,480][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:14:32,482][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:14:33,385][__main__][INFO] - Iteration 157 took 49s (27.98% Gen, 70.20% Train). Generation: 13s, Training: 34s. Estimated remaining time: 39h 13m 0s. Estimated total time: 41h 28m 36s. Time estimates for 10 more iterations: 8m 17s, 100 more iterations: 1h 22m 57s, 500 more iterations: 6h 54m 46s. -[2025-11-09 20:14:33,387][__main__][INFO] - Starting iteration 157. -[2025-11-09 20:14:33,864][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 15 and human policies 1. -[2025-11-09 20:14:33,865][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:14:49,168][__main__][INFO] - Number of regex retries in iteration 157: 0 -[2025-11-09 20:14:49,169][__main__][INFO] - agents played in iteration 157 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:14:49,996][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:14:50,020][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:14:50,043][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:14:50,065][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:14:50,066][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:14:50,067][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:14:50,734][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:14:51,216][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:14:51,721][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:14:52,222][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:14:52,731][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:14:53,230][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:14:53,736][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:14:54,234][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:14:54,733][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:14:55,231][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:14:55,729][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:14:56,226][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:14:56,724][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:14:57,221][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:14:57,738][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:14:58,235][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:14:58,736][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:14:59,237][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:14:59,739][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:15:00,242][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:15:00,741][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:15:01,240][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:15:01,747][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:15:02,246][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:15:02,749][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:15:03,247][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:15:03,749][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:15:04,255][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:15:04,757][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:15:05,260][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:15:05,760][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:15:06,259][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:15:06,762][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:15:07,260][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:15:07,776][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:15:08,280][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:15:08,779][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:15:09,280][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:15:09,778][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:15:10,275][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:15:10,774][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:15:11,269][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:15:11,767][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:15:12,265][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:15:12,761][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:15:13,256][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:15:13,758][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:15:14,256][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:15:14,755][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:15:15,250][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:15:15,747][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:15:16,241][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:15:16,740][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:15:17,240][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:15:17,736][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:15:18,234][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:15:18,730][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:15:19,231][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:15:19,732][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:15:20,231][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:15:20,732][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:15:21,253][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:15:21,757][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:15:22,258][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:15:22,761][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10862 tokens. -[2025-11-09 20:15:23,486][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.25%, Current % of VRAM taken: 55.75%, Block Peak % of device VRAM: 62.42%, ΔTime: 00:00:32 -[2025-11-09 20:15:24,219][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:15:24,221][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:15:24,223][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:15:25,117][__main__][INFO] - Iteration 158 took 51s (29.86% Gen, 68.39% Train). Generation: 15s, Training: 35s. Estimated remaining time: 40h 26m 12s. Estimated total time: 42h 42m 40s. Time estimates for 10 more iterations: 8m 32s, 100 more iterations: 1h 25m 25s, 500 more iterations: 7h 7m 6s. -[2025-11-09 20:15:25,119][__main__][INFO] - Starting iteration 158. -[2025-11-09 20:15:25,584][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 15 and human policies 1. -[2025-11-09 20:15:25,585][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:15:37,466][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:15:40,793][__main__][INFO] - Number of regex retries in iteration 158: 1 -[2025-11-09 20:15:40,794][__main__][INFO] - agents played in iteration 158 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:15:41,590][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:15:41,613][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:15:41,635][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:15:41,657][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:15:41,658][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:15:41,658][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:15:42,316][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:15:42,775][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:15:43,279][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:15:43,779][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:15:44,280][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:15:44,779][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:15:45,280][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:15:45,779][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:15:46,278][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:15:46,781][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:15:47,282][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:15:47,783][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:15:48,285][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:15:48,784][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:15:49,282][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:15:49,780][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:15:50,279][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:15:50,782][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:15:51,281][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:15:51,778][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:15:52,280][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:15:52,777][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:15:53,280][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:15:53,783][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:15:54,281][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:15:54,782][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:15:55,278][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:15:55,777][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:15:56,276][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:15:56,777][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:15:57,278][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:15:57,774][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:15:58,271][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:15:58,769][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:15:59,265][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:15:59,769][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:16:00,268][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:16:00,763][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:16:01,259][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:16:01,755][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:16:02,252][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:16:02,749][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:16:03,241][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:16:03,751][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:16:04,250][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:16:04,745][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:16:05,243][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:16:05,738][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:16:06,246][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:16:06,743][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:16:07,240][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:16:07,738][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:16:08,236][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:16:08,735][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:16:09,232][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:16:09,729][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:16:10,242][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:16:10,743][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:16:11,246][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:16:11,748][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:16:12,254][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:16:12,758][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:16:13,261][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:16:13,765][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:16:14,273][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10840 tokens. -[2025-11-09 20:16:14,973][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.25%, Current % of VRAM taken: 55.75%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:32 -[2025-11-09 20:16:15,715][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:16:15,716][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:16:15,720][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:16:16,650][__main__][INFO] - Iteration 159 took 51s (29.78% Gen, 68.39% Train). Generation: 15s, Training: 34s. Estimated remaining time: 40h 15m 58s. Estimated total time: 42h 33m 18s. Time estimates for 10 more iterations: 8m 30s, 100 more iterations: 1h 25m 6s, 500 more iterations: 7h 5m 33s. -[2025-11-09 20:16:16,652][__main__][INFO] - Starting iteration 159. -[2025-11-09 20:16:17,113][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 15 and human policies 1. -[2025-11-09 20:16:17,113][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:16:20,146][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:16:24,715][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:16:31,333][__main__][INFO] - Number of regex retries in iteration 159: 2 -[2025-11-09 20:16:31,334][__main__][INFO] - agents played in iteration 159 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:16:32,113][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:16:32,137][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:16:32,160][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:16:32,182][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:16:32,182][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:16:32,183][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:16:32,816][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:16:33,273][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:16:33,780][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:16:34,279][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:16:34,779][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:16:35,282][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:16:35,781][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:16:36,280][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:16:36,781][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:16:37,279][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:16:37,778][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:16:38,275][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:16:38,775][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:16:39,274][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:16:39,771][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:16:40,270][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:16:40,770][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:16:41,272][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:16:41,771][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:16:42,269][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:16:42,769][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:16:43,268][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:16:43,769][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:16:44,272][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:16:44,779][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:16:45,281][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:16:45,780][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:16:46,278][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:16:46,776][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:16:47,274][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:16:47,772][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:16:48,270][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:16:48,767][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:16:49,264][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:16:49,761][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:16:50,260][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:16:50,758][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:16:51,256][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:16:51,756][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:16:52,254][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:16:52,751][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:16:53,247][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:16:53,760][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:16:54,258][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:16:54,755][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:16:55,256][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:16:55,751][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:16:56,248][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:16:56,747][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:16:57,243][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:16:57,756][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:16:58,254][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:16:58,752][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:16:59,251][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:16:59,746][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:17:00,257][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:17:00,754][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:17:01,251][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:17:01,769][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:17:02,268][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:17:02,770][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:17:03,272][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:17:03,779][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:17:04,281][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:17:04,782][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10851 tokens. -[2025-11-09 20:17:05,535][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 55.71%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 -[2025-11-09 20:17:06,254][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:17:06,256][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:17:06,257][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:17:07,390][__main__][INFO] - Iteration 160 took 50s (28.29% Gen, 69.46% Train). Generation: 14s, Training: 34s. Estimated remaining time: 39h 35m 43s. Estimated total time: 41h 53m 53s. Time estimates for 10 more iterations: 8m 22s, 100 more iterations: 1h 23m 47s, 500 more iterations: 6h 58m 58s. -[2025-11-09 20:17:07,392][__main__][INFO] - Starting iteration 160. -[2025-11-09 20:17:07,856][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 15 and human policies 1. -[2025-11-09 20:17:07,856][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:17:22,267][__main__][INFO] - Number of regex retries in iteration 160: 0 -[2025-11-09 20:17:22,268][__main__][INFO] - agents played in iteration 160 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:17:23,070][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:17:23,094][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:17:23,117][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:17:23,139][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:17:23,140][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:17:23,140][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:17:23,832][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:17:24,294][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:17:24,793][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:17:25,292][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:17:25,793][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:17:26,290][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:17:26,790][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:17:27,287][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:17:27,785][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:17:28,285][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:17:28,784][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:17:29,283][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:17:29,783][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:17:30,281][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:17:30,779][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:17:31,277][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:17:31,779][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:17:32,276][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:17:32,776][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:17:33,278][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:17:33,776][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:17:34,282][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:17:34,779][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:17:35,280][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:17:35,781][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:17:36,282][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:17:36,780][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:17:37,281][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:17:37,779][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:17:38,277][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:17:38,776][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:17:39,275][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:17:39,780][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:17:40,277][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:17:40,774][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:17:41,270][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:17:41,765][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:17:42,267][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:17:42,763][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:17:43,259][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:17:43,773][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:17:44,276][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:17:44,773][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:17:45,271][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:17:45,768][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:17:46,275][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:17:46,773][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:17:47,270][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:17:47,785][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:17:48,284][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:17:48,781][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:17:49,292][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:17:49,789][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:17:50,287][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:17:50,784][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:17:51,281][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:17:51,789][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:17:52,287][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:17:52,784][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:17:53,284][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:17:53,786][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:17:54,293][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:17:54,793][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:17:55,293][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:17:55,795][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10822 tokens. -[2025-11-09 20:17:56,516][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 55.72%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:32 -[2025-11-09 20:17:57,254][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:17:57,255][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:17:57,257][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:17:58,947][__main__][INFO] - Iteration 161 took 51s (28.21% Gen, 68.48% Train). Generation: 14s, Training: 34s. Estimated remaining time: 40h 15m 33s. Estimated total time: 42h 34m 35s. Time estimates for 10 more iterations: 8m 30s, 100 more iterations: 1h 25m 9s, 500 more iterations: 7h 5m 45s. -[2025-11-09 20:17:58,949][__main__][INFO] - Starting iteration 161. -[2025-11-09 20:17:59,419][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 16 and human policies 1. -[2025-11-09 20:17:59,419][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:18:03,402][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:18:14,016][__main__][INFO] - Number of regex retries in iteration 161: 1 -[2025-11-09 20:18:14,017][__main__][INFO] - agents played in iteration 161 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:18:14,806][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:18:14,833][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:18:14,859][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:18:14,882][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:18:14,883][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:18:14,883][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:18:15,584][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:18:16,043][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:18:16,564][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:18:17,062][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:18:17,562][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:18:18,066][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:18:18,568][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:18:19,068][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:18:19,567][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:18:20,065][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:18:20,564][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:18:21,061][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:18:21,561][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:18:22,065][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:18:22,565][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:18:23,078][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:18:23,575][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:18:24,078][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:18:24,578][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:18:25,077][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:18:25,575][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:18:26,075][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:18:26,576][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:18:27,078][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:18:27,578][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:18:28,075][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:18:28,569][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:18:29,067][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:18:29,577][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:18:30,073][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:18:30,568][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:18:31,069][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:18:31,563][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:18:32,074][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:18:32,574][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:18:33,072][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:18:33,571][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:18:34,069][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:18:34,564][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:18:35,061][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:18:35,559][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:18:36,087][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:18:36,590][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:18:37,088][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:18:37,585][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:18:38,083][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:18:38,581][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:18:39,079][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:18:39,578][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:18:40,078][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:18:40,576][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:18:41,074][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:18:41,578][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:18:42,078][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:18:42,576][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:18:43,075][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:18:43,573][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:18:44,087][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:18:44,583][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:18:45,080][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:18:45,581][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:18:46,082][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:18:46,589][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:18:47,088][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:18:47,585][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10867 tokens. -[2025-11-09 20:18:48,281][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 55.68%, Block Peak % of device VRAM: 62.42%, ΔTime: 00:00:32 -[2025-11-09 20:18:49,040][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:18:49,042][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:18:49,043][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:18:49,925][__main__][INFO] - Iteration 162 took 50s (28.90% Gen, 69.35% Train). Generation: 14s, Training: 35s. Estimated remaining time: 39h 45m 27s. Estimated total time: 42h 5m 20s. Time estimates for 10 more iterations: 8m 25s, 100 more iterations: 1h 24m 10s, 500 more iterations: 7h 0m 53s. -[2025-11-09 20:18:49,927][__main__][INFO] - Starting iteration 162. -[2025-11-09 20:18:50,393][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 16 and human policies 1. -[2025-11-09 20:18:50,393][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:18:55,098][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:18:55,887][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:19:04,629][__main__][INFO] - Number of regex retries in iteration 162: 2 -[2025-11-09 20:19:04,630][__main__][INFO] - agents played in iteration 162 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:19:05,436][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:19:05,459][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:19:05,481][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:19:05,504][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:19:05,504][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:19:05,505][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:19:06,158][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:19:06,613][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:19:07,116][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:19:07,620][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:19:08,121][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:19:08,622][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:19:09,121][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:19:09,619][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:19:10,121][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:19:10,622][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:19:11,125][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:19:11,625][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:19:12,125][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:19:12,625][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:19:13,123][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:19:13,622][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:19:14,124][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:19:14,622][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:19:15,121][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:19:15,625][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:19:16,126][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:19:16,627][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:19:17,126][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:19:17,623][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:19:18,124][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:19:18,621][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:19:19,116][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:19:19,611][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:19:20,107][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:19:20,601][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:19:21,096][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:19:21,592][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:19:22,085][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:19:22,581][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:19:23,078][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:19:23,575][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:19:24,072][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:19:24,569][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:19:25,066][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:19:25,562][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:19:26,069][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:19:26,564][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:19:27,062][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:19:27,557][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:19:28,053][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:19:28,549][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:19:29,044][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:19:29,541][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:19:30,036][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:19:30,535][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:19:31,036][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:19:31,537][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:19:32,034][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:19:32,535][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:19:33,036][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:19:33,534][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:19:34,033][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:19:34,530][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:19:35,028][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:19:35,528][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:19:36,025][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:19:36,522][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:19:37,020][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:19:37,522][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:19:38,025][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10872 tokens. -[2025-11-09 20:19:38,735][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 55.69%, Block Peak % of device VRAM: 62.42%, ΔTime: 00:00:32 -[2025-11-09 20:19:39,507][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:19:39,509][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:19:39,511][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:19:40,409][__main__][INFO] - Iteration 163 took 50s (28.46% Gen, 69.74% Train). Generation: 14s, Training: 34s. Estimated remaining time: 39h 20m 6s. Estimated total time: 41h 40m 49s. Time estimates for 10 more iterations: 8m 20s, 100 more iterations: 1h 23m 21s, 500 more iterations: 6h 56m 48s. -[2025-11-09 20:19:40,411][__main__][INFO] - Starting iteration 163. -[2025-11-09 20:19:40,876][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 16 and human policies 1. -[2025-11-09 20:19:40,877][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:19:55,017][__main__][INFO] - Number of regex retries in iteration 163: 0 -[2025-11-09 20:19:55,018][__main__][INFO] - agents played in iteration 163 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:19:55,806][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:19:55,844][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:19:55,874][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:19:55,900][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:19:55,901][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:19:55,902][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:19:56,536][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:19:56,989][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:19:57,496][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:19:57,995][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:19:58,493][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:19:58,992][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:19:59,489][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:19:59,987][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:20:00,484][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:20:00,989][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:20:01,491][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:20:01,991][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:20:02,511][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:20:03,026][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:20:03,530][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:20:04,027][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:20:04,523][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:20:05,025][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:20:05,524][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:20:06,023][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:20:06,524][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:20:07,019][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:20:07,515][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:20:08,011][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:20:08,506][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:20:09,014][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:20:09,510][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:20:10,004][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:20:10,500][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:20:10,995][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:20:11,501][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:20:11,996][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:20:12,491][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:20:13,004][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:20:13,500][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:20:13,997][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:20:14,497][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:20:14,993][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:20:15,499][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:20:15,995][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:20:16,491][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:20:16,992][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:20:17,491][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:20:17,988][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:20:18,485][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:20:18,977][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:20:19,481][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:20:19,978][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:20:20,477][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:20:20,987][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:20:21,485][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:20:21,987][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:20:22,484][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:20:22,979][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:20:23,478][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:20:23,974][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:20:24,471][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:20:24,978][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:20:25,475][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:20:25,974][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:20:26,474][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:20:26,972][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:20:27,471][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:20:27,973][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:20:28,469][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10852 tokens. -[2025-11-09 20:20:29,162][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 55.66%, Block Peak % of device VRAM: 62.42%, ΔTime: 00:00:32 -[2025-11-09 20:20:29,915][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:20:29,916][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:20:29,918][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:20:30,789][__main__][INFO] - Iteration 164 took 49s (28.33% Gen, 69.92% Train). Generation: 14s, Training: 34s. Estimated remaining time: 39h 14m 7s. Estimated total time: 41h 35m 41s. Time estimates for 10 more iterations: 8m 19s, 100 more iterations: 1h 23m 11s, 500 more iterations: 6h 55m 56s. -[2025-11-09 20:20:30,792][__main__][INFO] - Starting iteration 164. -[2025-11-09 20:20:31,259][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 16 and human policies 1. -[2025-11-09 20:20:31,259][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:20:36,451][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:20:36,621][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:20:45,219][__main__][INFO] - Number of regex retries in iteration 164: 2 -[2025-11-09 20:20:45,220][__main__][INFO] - agents played in iteration 164 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:20:45,996][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:20:46,023][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:20:46,049][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:20:46,072][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:20:46,072][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:20:46,073][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:20:46,707][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:20:47,171][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:20:47,680][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:20:48,179][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:20:48,680][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:20:49,182][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:20:49,680][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:20:50,181][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:20:50,683][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:20:51,186][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:20:51,687][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:20:52,197][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:20:52,699][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:20:53,199][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:20:53,705][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:20:54,208][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:20:54,720][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:20:55,222][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:20:55,723][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:20:56,224][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:20:56,722][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:20:57,222][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:20:57,732][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:20:58,229][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:20:58,726][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:20:59,226][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:20:59,723][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:21:00,232][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:21:00,730][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:21:01,227][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:21:01,724][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:21:02,225][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:21:02,722][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:21:03,219][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:21:03,716][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:21:04,217][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:21:04,715][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:21:05,211][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:21:05,710][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:21:06,207][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:21:06,725][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:21:07,221][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:21:07,718][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:21:08,218][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:21:08,714][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:21:09,210][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:21:09,707][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:21:10,204][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:21:10,716][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:21:11,212][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:21:11,710][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:21:12,211][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:21:12,707][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:21:13,215][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:21:13,713][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:21:14,211][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:21:14,719][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:21:15,218][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:21:15,715][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:21:16,218][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:21:16,721][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:21:17,228][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:21:17,730][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:21:18,230][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:21:18,740][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10835 tokens. -[2025-11-09 20:21:19,426][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.26%, Current % of VRAM taken: 55.76%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 -[2025-11-09 20:21:20,158][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:21:20,159][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:21:20,161][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:21:21,065][__main__][INFO] - Iteration 165 took 49s (28.03% Gen, 70.15% Train). Generation: 13s, Training: 34s. Estimated remaining time: 39h 7m 57s. Estimated total time: 41h 30m 21s. Time estimates for 10 more iterations: 8m 18s, 100 more iterations: 1h 23m 0s, 500 more iterations: 6h 55m 3s. -[2025-11-09 20:21:21,067][__main__][INFO] - Starting iteration 165. -[2025-11-09 20:21:21,529][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 16 and human policies 1. -[2025-11-09 20:21:21,530][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:21:35,762][__main__][INFO] - Number of regex retries in iteration 165: 0 -[2025-11-09 20:21:35,762][__main__][INFO] - agents played in iteration 165 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:21:36,537][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:21:36,559][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:21:36,585][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:21:36,607][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:21:36,608][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:21:36,608][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:21:37,254][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:21:37,706][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:21:38,222][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:21:38,719][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:21:39,218][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:21:39,720][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:21:40,216][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:21:40,713][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:21:41,210][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:21:41,712][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:21:42,228][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:21:42,727][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:21:43,228][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:21:43,728][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:21:44,227][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:21:44,729][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:21:45,229][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:21:45,729][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:21:46,237][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:21:46,737][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:21:47,241][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:21:47,737][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:21:48,237][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:21:48,740][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:21:49,240][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:21:49,738][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:21:50,242][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:21:50,739][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:21:51,236][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:21:51,736][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:21:52,233][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:21:52,733][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:21:53,230][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:21:53,728][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:21:54,229][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:21:54,725][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:21:55,222][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:21:55,720][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:21:56,217][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:21:56,725][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:21:57,222][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:21:57,719][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:21:58,218][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:21:58,716][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:21:59,216][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:21:59,715][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:22:00,215][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:22:00,727][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:22:01,222][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:22:01,720][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:22:02,220][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:22:02,718][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:22:03,216][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:22:03,712][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:22:04,210][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:22:04,709][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:22:05,205][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:22:05,701][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:22:06,198][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:22:06,695][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:22:07,193][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:22:07,691][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:22:08,190][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:22:08,689][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:22:09,188][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10860 tokens. -[2025-11-09 20:22:09,856][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 55.62%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:32 -[2025-11-09 20:22:10,593][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:22:10,594][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:22:10,596][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:22:11,488][__main__][INFO] - Iteration 166 took 49s (28.49% Gen, 69.72% Train). Generation: 14s, Training: 34s. Estimated remaining time: 39h 14m 42s. Estimated total time: 41h 37m 56s. Time estimates for 10 more iterations: 8m 19s, 100 more iterations: 1h 23m 15s, 500 more iterations: 6h 56m 19s. -[2025-11-09 20:22:11,491][__main__][INFO] - Starting iteration 166. -[2025-11-09 20:22:11,953][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 16 and human policies 1. -[2025-11-09 20:22:11,953][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:22:18,834][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:22:26,051][__main__][INFO] - Number of regex retries in iteration 166: 1 -[2025-11-09 20:22:26,052][__main__][INFO] - agents played in iteration 166 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:22:26,853][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:22:26,878][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:22:26,901][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:22:26,923][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:22:26,924][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:22:26,924][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:22:27,566][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:22:28,021][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:22:28,530][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:22:29,029][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:22:29,527][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:22:30,026][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:22:30,525][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:22:31,040][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:22:31,540][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:22:32,039][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:22:32,540][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:22:33,040][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:22:33,537][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:22:34,034][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:22:34,532][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:22:35,033][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:22:35,532][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:22:36,032][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:22:36,532][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:22:37,030][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:22:37,527][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:22:38,024][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:22:38,522][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:22:39,028][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:22:39,525][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:22:40,045][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:22:40,541][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:22:41,037][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:22:41,536][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:22:42,037][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:22:42,541][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:22:43,038][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:22:43,534][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:22:44,048][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:22:44,543][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:22:45,039][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:22:45,532][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:22:46,028][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:22:46,539][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:22:47,035][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:22:47,531][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:22:48,033][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:22:48,530][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:22:49,041][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:22:49,566][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:22:50,068][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:22:50,572][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:22:51,071][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:22:51,569][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:22:52,066][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:22:52,561][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:22:53,059][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:22:53,554][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:22:54,050][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:22:54,547][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:22:55,042][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:22:55,539][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:22:56,037][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:22:56,533][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:22:57,041][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:22:57,540][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:22:58,042][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:22:58,541][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:22:59,044][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:22:59,544][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10852 tokens. -[2025-11-09 20:23:00,223][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.28%, Current % of VRAM taken: 55.77%, Block Peak % of device VRAM: 62.50%, ΔTime: 00:00:32 -[2025-11-09 20:23:00,985][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:23:00,987][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:23:00,990][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:23:01,892][__main__][INFO] - Iteration 167 took 49s (28.23% Gen, 69.96% Train). Generation: 14s, Training: 34s. Estimated remaining time: 39h 12m 54s. Estimated total time: 41h 36m 59s. Time estimates for 10 more iterations: 8m 19s, 100 more iterations: 1h 23m 13s, 500 more iterations: 6h 56m 9s. -[2025-11-09 20:23:01,894][__main__][INFO] - Starting iteration 167. -[2025-11-09 20:23:02,359][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 16 and human policies 1. -[2025-11-09 20:23:02,360][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:23:15,940][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hatss, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:23:16,239][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:23:17,205][__main__][INFO] - Number of regex retries in iteration 167: 2 -[2025-11-09 20:23:17,205][__main__][INFO] - agents played in iteration 167 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:23:18,008][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:23:18,030][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:23:18,053][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:23:18,075][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:23:18,076][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:23:18,076][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:23:18,751][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:23:19,212][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:23:19,718][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:23:20,221][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:23:20,735][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:23:21,237][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:23:21,738][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:23:22,239][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:23:22,738][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:23:23,250][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:23:23,750][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:23:24,250][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:23:24,755][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:23:25,255][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:23:25,757][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:23:26,257][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:23:26,757][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:23:27,256][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:23:27,755][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:23:28,253][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:23:28,764][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:23:29,264][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:23:29,765][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:23:30,262][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:23:30,763][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:23:31,262][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:23:31,762][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:23:32,263][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:23:32,767][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:23:33,265][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:23:33,762][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:23:34,261][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:23:34,764][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:23:35,270][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:23:35,768][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:23:36,264][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:23:36,760][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:23:37,257][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:23:37,756][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:23:38,255][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:23:38,752][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:23:39,249][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:23:39,747][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:23:40,243][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:23:40,741][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:23:41,237][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:23:41,733][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:23:42,230][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:23:42,726][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:23:43,247][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:23:43,744][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:23:44,240][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:23:44,732][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:23:45,230][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:23:45,737][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:23:46,232][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:23:46,729][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:23:47,227][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:23:47,723][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:23:48,219][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:23:48,720][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:23:49,221][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:23:49,721][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:23:50,222][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:23:50,721][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10857 tokens. -[2025-11-09 20:23:51,406][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 55.70%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:32 -[2025-11-09 20:23:52,150][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:23:52,152][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:23:52,154][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:23:53,007][__main__][INFO] - Iteration 168 took 50s (29.31% Gen, 69.00% Train). Generation: 14s, Training: 34s. Estimated remaining time: 39h 47m 30s. Estimated total time: 42h 12m 26s. Time estimates for 10 more iterations: 8m 26s, 100 more iterations: 1h 24m 24s, 500 more iterations: 7h 2m 4s. -[2025-11-09 20:23:53,010][__main__][INFO] - Starting iteration 168. -[2025-11-09 20:23:53,485][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 16 and human policies 1. -[2025-11-09 20:23:53,486][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:24:01,332][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:24:06,924][__main__][INFO] - Number of regex retries in iteration 168: 1 -[2025-11-09 20:24:06,925][__main__][INFO] - agents played in iteration 168 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:24:07,709][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:24:07,738][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:24:07,765][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:24:07,789][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:24:07,790][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:24:07,791][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:24:08,481][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:24:08,939][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:24:09,443][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:24:09,947][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:24:10,445][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:24:10,964][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:24:11,463][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:24:11,963][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:24:12,464][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:24:12,966][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:24:13,478][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:24:13,980][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:24:14,483][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:24:14,988][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:24:15,485][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:24:15,984][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:24:16,483][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:24:16,981][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:24:17,480][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:24:17,981][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:24:18,479][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:24:18,981][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:24:19,478][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:24:19,977][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:24:20,475][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:24:20,973][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:24:21,471][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:24:21,969][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:24:22,466][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:24:22,975][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:24:23,471][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:24:23,968][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:24:24,470][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:24:24,967][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:24:25,460][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:24:25,959][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:24:26,457][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:24:26,974][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:24:27,476][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:24:27,971][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:24:28,467][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:24:28,963][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:24:29,471][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:24:29,967][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:24:30,463][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:24:30,960][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:24:31,457][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:24:31,953][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:24:32,449][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:24:32,945][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:24:33,458][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:24:33,955][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:24:34,453][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:24:34,949][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:24:35,444][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:24:35,946][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:24:36,442][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:24:36,941][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:24:37,441][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:24:37,940][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:24:38,440][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:24:38,950][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:24:39,446][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:24:39,960][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:24:40,460][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10865 tokens. -[2025-11-09 20:24:41,155][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 55.71%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:32 -[2025-11-09 20:24:41,894][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:24:41,895][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:24:41,898][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:24:42,813][__main__][INFO] - Iteration 169 took 49s (27.24% Gen, 70.90% Train). Generation: 13s, Training: 34s. Estimated remaining time: 38h 40m 40s. Estimated total time: 41h 6m 25s. Time estimates for 10 more iterations: 8m 13s, 100 more iterations: 1h 22m 12s, 500 more iterations: 6h 51m 4s. -[2025-11-09 20:24:42,815][__main__][INFO] - Starting iteration 169. -[2025-11-09 20:24:43,278][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 16 and human policies 1. -[2025-11-09 20:24:43,278][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:24:47,635][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:24:57,355][__main__][INFO] - Number of regex retries in iteration 169: 1 -[2025-11-09 20:24:57,356][__main__][INFO] - agents played in iteration 169 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:24:58,137][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:24:58,164][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:24:58,202][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:24:58,227][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:24:58,228][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:24:58,229][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:24:58,856][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:24:59,311][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:24:59,823][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:25:00,326][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:25:00,827][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:25:01,330][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:25:01,831][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:25:02,335][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:25:02,841][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:25:03,343][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:25:03,852][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:25:04,353][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:25:04,868][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:25:05,365][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:25:05,862][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:25:06,361][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:25:06,859][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:25:07,360][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:25:07,857][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:25:08,355][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:25:08,873][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:25:09,371][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:25:09,871][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:25:10,372][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:25:10,869][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:25:11,370][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:25:11,866][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:25:12,365][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:25:12,878][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:25:13,378][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:25:13,876][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:25:14,376][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:25:14,876][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:25:15,373][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:25:15,869][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:25:16,367][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:25:16,878][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:25:17,378][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:25:17,877][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:25:18,374][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:25:18,870][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:25:19,366][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:25:19,864][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:25:20,365][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:25:20,866][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:25:21,366][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:25:21,867][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:25:22,365][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:25:22,863][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:25:23,379][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:25:23,879][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:25:24,380][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:25:24,877][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:25:25,376][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:25:25,880][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:25:26,379][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:25:26,880][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:25:27,387][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:25:27,890][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:25:28,392][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:25:28,895][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:25:29,398][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:25:29,902][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:25:30,404][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:25:30,904][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10841 tokens. -[2025-11-09 20:25:31,608][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.39%, ΔTime: 00:00:32 -[2025-11-09 20:25:32,361][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:25:32,362][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:25:32,364][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:25:33,289][__main__][INFO] - Iteration 170 took 50s (28.15% Gen, 70.00% Train). Generation: 14s, Training: 35s. Estimated remaining time: 39h 14m 1s. Estimated total time: 41h 40m 37s. Time estimates for 10 more iterations: 8m 20s, 100 more iterations: 1h 23m 21s, 500 more iterations: 6h 56m 46s. -[2025-11-09 20:25:33,292][__main__][INFO] - Starting iteration 170. -[2025-11-09 20:25:33,758][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 16 and human policies 1. -[2025-11-09 20:25:33,759][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:25:45,809][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:25:48,918][__main__][INFO] - Number of regex retries in iteration 170: 1 -[2025-11-09 20:25:48,919][__main__][INFO] - agents played in iteration 170 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:25:49,697][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:25:49,719][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:25:49,744][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:25:49,766][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:25:49,766][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:25:49,768][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:25:50,385][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:25:50,854][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:25:51,357][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:25:51,856][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:25:52,357][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:25:52,856][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:25:53,358][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:25:53,858][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:25:54,359][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:25:54,877][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:25:55,380][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:25:55,880][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:25:56,382][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:25:56,885][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:25:57,387][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:25:57,887][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:25:58,387][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:25:58,894][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:25:59,391][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:25:59,891][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:26:00,389][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:26:00,888][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:26:01,385][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:26:01,886][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:26:02,386][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:26:02,891][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:26:03,389][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:26:03,887][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:26:04,384][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:26:04,884][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:26:05,386][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:26:05,887][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:26:06,386][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:26:06,903][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:26:07,405][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:26:07,903][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:26:08,402][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:26:08,902][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:26:09,402][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:26:09,903][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:26:10,402][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:26:10,902][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:26:11,400][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:26:11,899][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:26:12,397][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:26:12,899][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:26:13,402][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:26:13,902][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:26:14,401][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:26:14,898][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:26:15,398][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:26:15,897][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:26:16,394][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:26:16,893][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:26:17,394][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:26:17,892][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:26:18,392][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:26:18,896][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:26:19,398][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:26:19,900][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:26:20,401][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:26:20,904][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:26:21,405][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:26:21,904][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:26:22,406][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10856 tokens. -[2025-11-09 20:26:23,108][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.27%, Current % of VRAM taken: 55.77%, Block Peak % of device VRAM: 62.34%, ΔTime: 00:00:32 -[2025-11-09 20:26:23,870][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:26:23,872][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:26:23,876][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:26:25,558][__main__][INFO] - Iteration 171 took 51s (29.27% Gen, 67.48% Train). Generation: 15s, Training: 34s. Estimated remaining time: 40h 42m 36s. Estimated total time: 43h 10m 4s. Time estimates for 10 more iterations: 8m 38s, 100 more iterations: 1h 26m 20s, 500 more iterations: 7h 11m 40s. -[2025-11-09 20:26:25,560][__main__][INFO] - Starting iteration 171. -[2025-11-09 20:26:26,021][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 17 and human policies 1. -[2025-11-09 20:26:26,022][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:26:39,865][__main__][INFO] - Number of regex retries in iteration 171: 0 -[2025-11-09 20:26:39,866][__main__][INFO] - agents played in iteration 171 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:26:40,663][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:26:40,686][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:26:40,708][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:26:40,732][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:26:40,733][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:26:40,734][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:26:41,362][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:26:41,817][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:26:42,321][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:26:42,818][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:26:43,319][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:26:43,820][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:26:44,318][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:26:44,818][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:26:45,318][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:26:45,816][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:26:46,318][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:26:46,816][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:26:47,318][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:26:47,822][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:26:48,322][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:26:48,823][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:26:49,321][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:26:49,820][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:26:50,320][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:26:50,820][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:26:51,322][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:26:51,820][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:26:52,318][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:26:52,815][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:26:53,311][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:26:53,808][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:26:54,308][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:26:54,809][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:26:55,307][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:26:55,806][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:26:56,303][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:26:56,802][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:26:57,301][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:26:57,799][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:26:58,298][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:26:58,798][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:26:59,298][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:26:59,801][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:27:00,299][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:27:00,802][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:27:01,303][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:27:01,807][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:27:02,308][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:27:02,808][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:27:03,309][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:27:03,812][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:27:04,317][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:27:04,816][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:27:05,316][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:27:05,814][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:27:06,312][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:27:06,810][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:27:07,310][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:27:07,808][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:27:08,304][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:27:08,801][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:27:09,300][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:27:09,801][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:27:10,303][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:27:10,802][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:27:11,302][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:27:11,803][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:27:12,306][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:27:12,807][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:27:13,306][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10848 tokens. -[2025-11-09 20:27:14,020][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 55.69%, Block Peak % of device VRAM: 62.35%, ΔTime: 00:00:32 -[2025-11-09 20:27:14,759][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:27:14,760][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:27:14,761][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:27:15,662][__main__][INFO] - Iteration 172 took 49s (27.89% Gen, 70.30% Train). Generation: 13s, Training: 34s. Estimated remaining time: 38h 53m 44s. Estimated total time: 41h 22m 2s. Time estimates for 10 more iterations: 8m 16s, 100 more iterations: 1h 22m 44s, 500 more iterations: 6h 53m 40s. -[2025-11-09 20:27:15,664][__main__][INFO] - Starting iteration 172. -[2025-11-09 20:27:16,139][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 17 and human policies 1. -[2025-11-09 20:27:16,140][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:27:31,469][__main__][INFO] - Number of regex retries in iteration 172: 0 -[2025-11-09 20:27:31,470][__main__][INFO] - agents played in iteration 172 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:27:32,265][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:27:32,291][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:27:32,316][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:27:32,339][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:27:32,339][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:27:32,340][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:27:32,966][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:27:33,422][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:27:33,925][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:27:34,424][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:27:34,924][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:27:35,422][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:27:35,920][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:27:36,421][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:27:36,923][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:27:37,427][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:27:37,927][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:27:38,425][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:27:38,924][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:27:39,423][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:27:39,924][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:27:40,426][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:27:40,925][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:27:41,427][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:27:41,927][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:27:42,426][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:27:42,925][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:27:43,422][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:27:43,920][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:27:44,418][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:27:44,922][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:27:45,438][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:27:45,945][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:27:46,451][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:27:46,960][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:27:47,468][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:27:47,966][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:27:48,463][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:27:48,962][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:27:49,461][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:27:49,960][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:27:50,459][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:27:50,957][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:27:51,455][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:27:51,952][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:27:52,448][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:27:52,945][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:27:53,444][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:27:53,942][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:27:54,438][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:27:54,936][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:27:55,433][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:27:55,931][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:27:56,430][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:27:56,930][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:27:57,431][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:27:57,930][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:27:58,427][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:27:58,923][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:27:59,419][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:27:59,917][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:28:00,418][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:28:00,919][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:28:01,417][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:28:01,915][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:28:02,413][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:28:02,911][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:28:03,413][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:28:03,914][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:28:04,414][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:28:04,919][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10852 tokens. -[2025-11-09 20:28:05,607][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 55.65%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 -[2025-11-09 20:28:06,384][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:28:06,386][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:28:06,388][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:28:07,320][__main__][INFO] - Iteration 173 took 51s (29.95% Gen, 68.23% Train). Generation: 15s, Training: 34s. Estimated remaining time: 40h 9m 51s. Estimated total time: 42h 39m 2s. Time estimates for 10 more iterations: 8m 31s, 100 more iterations: 1h 25m 18s, 500 more iterations: 7h 6m 30s. -[2025-11-09 20:28:07,323][__main__][INFO] - Starting iteration 173. -[2025-11-09 20:28:07,792][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 17 and human policies 1. -[2025-11-09 20:28:07,793][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:28:11,172][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:28:22,102][__main__][INFO] - Number of regex retries in iteration 173: 1 -[2025-11-09 20:28:22,103][__main__][INFO] - agents played in iteration 173 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:28:22,878][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:28:22,905][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:28:22,931][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:28:22,954][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:28:22,954][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:28:22,955][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:28:23,595][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:28:24,050][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:28:24,555][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:28:25,057][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:28:25,556][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:28:26,057][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:28:26,558][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:28:27,059][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:28:27,560][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:28:28,057][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:28:28,556][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:28:29,056][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:28:29,556][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:28:30,054][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:28:30,556][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:28:31,056][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:28:31,560][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:28:32,061][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:28:32,562][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:28:33,062][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:28:33,561][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:28:34,063][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:28:34,563][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:28:35,063][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:28:35,562][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:28:36,063][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:28:36,566][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:28:37,066][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:28:37,564][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:28:38,062][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:28:38,560][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:28:39,060][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:28:39,557][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:28:40,054][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:28:40,553][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:28:41,050][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:28:41,548][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:28:42,046][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:28:42,544][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:28:43,044][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:28:43,541][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:28:44,043][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:28:44,543][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:28:45,043][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:28:45,544][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:28:46,042][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:28:46,542][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:28:47,043][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:28:47,542][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:28:48,040][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:28:48,542][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:28:49,041][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:28:49,540][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:28:50,042][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:28:50,544][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:28:51,047][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:28:51,550][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:28:52,052][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:28:52,555][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:28:53,056][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:28:53,559][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:28:54,062][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:28:54,562][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:28:55,067][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:28:55,570][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10843 tokens. -[2025-11-09 20:28:56,254][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.30%, Current % of VRAM taken: 55.79%, Block Peak % of device VRAM: 62.43%, ΔTime: 00:00:32 -[2025-11-09 20:28:57,018][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:28:57,019][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:28:57,021][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:28:57,969][__main__][INFO] - Iteration 174 took 50s (28.52% Gen, 69.59% Train). Generation: 14s, Training: 34s. Estimated remaining time: 39h 18m 49s. Estimated total time: 41h 48m 50s. Time estimates for 10 more iterations: 8m 21s, 100 more iterations: 1h 23m 37s, 500 more iterations: 6h 58m 8s. -[2025-11-09 20:28:57,971][__main__][INFO] - Starting iteration 174. -[2025-11-09 20:28:58,435][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 17 and human policies 1. -[2025-11-09 20:28:58,436][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:29:12,721][__main__][INFO] - Number of regex retries in iteration 174: 0 -[2025-11-09 20:29:12,722][__main__][INFO] - agents played in iteration 174 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:29:13,507][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:29:13,532][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:29:13,557][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:29:13,579][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:29:13,580][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:29:13,580][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:29:14,217][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:29:14,686][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:29:15,187][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:29:15,691][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:29:16,189][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:29:16,687][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:29:17,198][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:29:17,699][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:29:18,204][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:29:18,708][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:29:19,209][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:29:19,713][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:29:20,214][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:29:20,715][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:29:21,216][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:29:21,716][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:29:22,218][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:29:22,718][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:29:23,217][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:29:23,714][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:29:24,216][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:29:24,730][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:29:25,229][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:29:25,731][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:29:26,234][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:29:26,731][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:29:27,232][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:29:27,732][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:29:28,233][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:29:28,742][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:29:29,241][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:29:29,743][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:29:30,241][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:29:30,738][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:29:31,239][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:29:31,738][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:29:32,236][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:29:32,736][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:29:33,235][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:29:33,732][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:29:34,229][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:29:34,726][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:29:35,241][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:29:35,739][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:29:36,238][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:29:36,736][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:29:37,233][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:29:37,731][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:29:38,230][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:29:38,728][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:29:39,241][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:29:39,739][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:29:40,238][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:29:40,739][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:29:41,241][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:29:41,745][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:29:42,246][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:29:42,747][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:29:43,253][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:29:43,760][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:29:44,261][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:29:44,763][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:29:45,264][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:29:45,768][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:29:46,268][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10842 tokens. -[2025-11-09 20:29:46,954][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.30%, Current % of VRAM taken: 55.80%, Block Peak % of device VRAM: 62.52%, ΔTime: 00:00:32 -[2025-11-09 20:29:47,712][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:29:47,713][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:29:47,715][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:29:48,647][__main__][INFO] - Iteration 175 took 50s (28.45% Gen, 69.69% Train). Generation: 14s, Training: 34s. Estimated remaining time: 39h 19m 44s. Estimated total time: 41h 50m 35s. Time estimates for 10 more iterations: 8m 22s, 100 more iterations: 1h 23m 41s, 500 more iterations: 6h 58m 25s. -[2025-11-09 20:29:48,649][__main__][INFO] - Starting iteration 175. -[2025-11-09 20:29:49,116][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 17 and human policies 1. -[2025-11-09 20:29:49,116][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:30:04,303][__main__][INFO] - Number of regex retries in iteration 175: 0 -[2025-11-09 20:30:04,304][__main__][INFO] - agents played in iteration 175 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:30:05,080][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:30:05,103][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:30:05,125][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:30:05,147][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:30:05,148][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:30:05,149][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:30:05,798][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:30:06,257][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:30:06,766][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:30:07,267][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:30:07,765][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:30:08,265][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:30:08,763][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:30:09,261][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:30:09,761][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:30:10,258][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:30:10,759][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:30:11,256][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:30:11,753][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:30:12,255][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:30:12,754][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:30:13,251][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:30:13,750][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:30:14,246][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:30:14,740][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:30:15,241][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:30:15,738][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:30:16,249][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:30:16,750][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:30:17,252][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:30:17,749][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:30:18,252][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:30:18,757][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:30:19,259][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:30:19,759][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:30:20,263][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:30:20,763][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:30:21,264][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:30:21,765][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:30:22,264][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:30:22,764][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:30:23,262][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:30:23,761][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:30:24,259][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:30:24,758][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:30:25,259][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:30:25,758][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:30:26,256][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:30:26,757][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:30:27,256][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:30:27,757][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:30:28,256][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:30:28,754][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:30:29,258][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:30:29,758][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:30:30,258][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:30:30,759][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:30:31,261][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:30:31,763][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:30:32,264][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:30:32,764][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:30:33,267][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:30:33,766][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:30:34,269][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:30:34,772][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:30:35,272][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:30:35,778][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:30:36,281][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:30:36,783][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:30:37,285][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:30:37,784][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10844 tokens. -[2025-11-09 20:30:38,426][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 55.72%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 -[2025-11-09 20:30:39,175][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:30:39,177][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:30:39,180][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:30:40,101][__main__][INFO] - Iteration 176 took 50s (29.79% Gen, 68.40% Train). Generation: 15s, Training: 34s. Estimated remaining time: 39h 57m 35s. Estimated total time: 42h 29m 19s. Time estimates for 10 more iterations: 8m 29s, 100 more iterations: 1h 24m 58s, 500 more iterations: 7h 4m 53s. -[2025-11-09 20:30:40,104][__main__][INFO] - Starting iteration 176. -[2025-11-09 20:30:40,567][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 17 and human policies 1. -[2025-11-09 20:30:40,568][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:30:45,035][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:30:55,307][__main__][INFO] - Number of regex retries in iteration 176: 1 -[2025-11-09 20:30:55,308][__main__][INFO] - agents played in iteration 176 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:30:56,108][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:30:56,135][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:30:56,161][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:30:56,183][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:30:56,184][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:30:56,185][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:30:56,840][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:30:57,297][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:30:57,804][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:30:58,325][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:30:58,831][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:30:59,337][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:30:59,843][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:31:00,348][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:31:00,847][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:31:01,340][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:31:01,836][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:31:02,351][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:31:02,846][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:31:03,338][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:31:03,839][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:31:04,335][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:31:04,846][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:31:05,344][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:31:05,842][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:31:06,352][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:31:06,848][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:31:07,344][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:31:07,840][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:31:08,337][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:31:08,839][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:31:09,340][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:31:09,838][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:31:10,338][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:31:10,838][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:31:11,336][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:31:11,834][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:31:12,332][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:31:12,847][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:31:13,350][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:31:13,855][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:31:14,358][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:31:14,857][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:31:15,361][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:31:15,859][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:31:16,360][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:31:16,856][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:31:17,354][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:31:17,851][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:31:18,348][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:31:18,845][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:31:19,342][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:31:19,840][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:31:20,338][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:31:20,838][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:31:21,336][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:31:21,845][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:31:22,342][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:31:22,843][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:31:23,343][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:31:23,843][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:31:24,346][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:31:24,847][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:31:25,346][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:31:25,860][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:31:26,362][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:31:26,867][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:31:27,370][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:31:27,872][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:31:28,377][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:31:28,878][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10868 tokens. -[2025-11-09 20:31:29,580][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.29%, Current % of VRAM taken: 55.78%, Block Peak % of device VRAM: 62.50%, ΔTime: 00:00:32 -[2025-11-09 20:31:30,348][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:31:30,350][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:31:30,351][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:31:31,281][__main__][INFO] - Iteration 177 took 50s (29.07% Gen, 69.10% Train). Generation: 14s, Training: 35s. Estimated remaining time: 39h 43m 8s. Estimated total time: 42h 15m 43s. Time estimates for 10 more iterations: 8m 27s, 100 more iterations: 1h 24m 31s, 500 more iterations: 7h 2m 37s. -[2025-11-09 20:31:31,284][__main__][INFO] - Starting iteration 177. -[2025-11-09 20:31:31,775][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 17 and human policies 1. -[2025-11-09 20:31:31,776][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:31:47,444][__main__][INFO] - Number of regex retries in iteration 177: 0 -[2025-11-09 20:31:47,444][__main__][INFO] - agents played in iteration 177 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:31:48,254][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:31:48,283][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:31:48,310][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:31:48,333][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:31:48,334][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:31:48,335][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:31:48,956][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:31:49,431][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:31:49,936][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:31:50,435][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:31:50,937][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:31:51,434][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:31:51,931][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:31:52,428][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:31:52,925][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:31:53,441][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:31:53,939][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:31:54,436][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:31:54,934][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:31:55,431][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:31:55,944][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:31:56,441][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:31:56,937][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:31:57,437][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:31:57,936][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:31:58,445][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:31:58,941][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:31:59,441][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:31:59,955][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:32:00,455][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:32:00,956][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:32:01,455][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:32:01,955][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:32:02,462][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:32:02,959][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:32:03,461][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:32:03,961][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:32:04,461][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:32:04,963][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:32:05,463][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:32:05,965][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:32:06,470][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:32:06,966][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:32:07,464][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:32:07,968][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:32:08,464][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:32:08,966][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:32:09,466][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:32:09,962][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:32:10,462][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:32:10,960][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:32:11,458][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:32:11,966][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:32:12,464][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:32:12,963][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:32:13,468][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:32:13,972][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:32:14,481][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:32:14,983][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:32:15,487][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:32:15,986][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:32:16,487][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:32:16,988][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:32:17,489][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:32:17,989][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:32:18,499][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:32:18,999][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:32:19,498][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:32:19,996][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:32:20,494][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:32:20,998][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10846 tokens. -[2025-11-09 20:32:21,730][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 55.72%, Block Peak % of device VRAM: 62.28%, ΔTime: 00:00:32 -[2025-11-09 20:32:22,512][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:32:22,513][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:32:22,515][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:32:23,438][__main__][INFO] - Iteration 178 took 51s (30.33% Gen, 67.88% Train). Generation: 15s, Training: 35s. Estimated remaining time: 40h 29m 41s. Estimated total time: 43h 3m 7s. Time estimates for 10 more iterations: 8m 36s, 100 more iterations: 1h 26m 6s, 500 more iterations: 7h 10m 31s. -[2025-11-09 20:32:23,440][__main__][INFO] - Starting iteration 178. -[2025-11-09 20:32:23,924][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 17 and human policies 1. -[2025-11-09 20:32:23,925][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:32:40,168][__main__][INFO] - Number of regex retries in iteration 178: 0 -[2025-11-09 20:32:40,169][__main__][INFO] - agents played in iteration 178 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:32:40,960][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:32:40,987][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:32:41,013][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:32:41,036][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:32:41,036][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:32:41,037][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:32:41,653][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:32:42,110][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:32:42,612][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:32:43,107][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:32:43,606][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:32:44,102][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:32:44,601][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:32:45,103][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:32:45,600][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:32:46,101][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:32:46,599][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:32:47,101][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:32:47,618][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:32:48,117][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:32:48,615][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:32:49,117][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:32:49,614][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:32:50,111][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:32:50,608][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:32:51,105][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:32:51,604][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:32:52,101][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:32:52,599][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:32:53,093][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:32:53,589][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:32:54,091][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:32:54,591][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:32:55,092][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:32:55,594][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:32:56,095][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:32:56,594][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:32:57,093][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:32:57,591][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:32:58,092][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:32:58,593][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:32:59,093][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:32:59,595][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:33:00,095][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:33:00,597][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:33:01,094][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:33:01,593][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:33:02,087][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:33:02,590][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:33:03,091][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:33:03,593][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:33:04,094][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:33:04,596][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:33:05,096][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:33:05,598][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:33:06,100][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:33:06,601][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:33:07,104][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:33:07,604][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:33:08,103][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:33:08,608][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:33:09,106][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:33:09,607][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:33:10,108][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:33:10,607][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:33:11,105][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:33:11,603][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:33:12,103][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:33:12,609][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:33:13,109][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:33:13,610][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10851 tokens. -[2025-11-09 20:33:14,318][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.25%, Current % of VRAM taken: 55.75%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 -[2025-11-09 20:33:15,098][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:33:15,100][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:33:15,102][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:33:15,974][__main__][INFO] - Iteration 179 took 52s (31.21% Gen, 67.11% Train). Generation: 16s, Training: 34s. Estimated remaining time: 40h 48m 11s. Estimated total time: 43h 22m 30s. Time estimates for 10 more iterations: 8m 40s, 100 more iterations: 1h 26m 45s, 500 more iterations: 7h 13m 45s. -[2025-11-09 20:33:15,976][__main__][INFO] - Starting iteration 179. -[2025-11-09 20:33:16,449][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 17 and human policies 1. -[2025-11-09 20:33:16,450][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:33:19,930][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:33:20,118][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:33:31,726][__main__][INFO] - Number of regex retries in iteration 179: 2 -[2025-11-09 20:33:31,726][__main__][INFO] - agents played in iteration 179 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:33:32,498][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:33:32,521][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:33:32,543][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:33:32,565][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:33:32,565][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:33:32,566][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:33:33,178][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:33:33,635][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:33:34,137][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:33:34,635][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:33:35,134][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:33:35,631][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:33:36,128][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:33:36,628][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:33:37,124][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:33:37,622][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:33:38,120][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:33:38,616][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:33:39,116][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:33:39,613][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:33:40,111][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:33:40,608][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:33:41,109][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:33:41,610][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:33:42,106][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:33:42,604][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:33:43,103][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:33:43,602][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:33:44,103][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:33:44,603][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:33:45,101][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:33:45,600][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:33:46,096][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:33:46,591][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:33:47,086][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:33:47,582][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:33:48,092][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:33:48,588][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:33:49,089][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:33:49,591][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:33:50,090][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:33:50,589][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:33:51,088][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:33:51,586][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:33:52,099][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:33:52,600][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:33:53,101][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:33:53,601][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:33:54,102][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:33:54,604][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:33:55,106][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:33:55,607][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:33:56,111][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:33:56,611][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:33:57,114][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:33:57,618][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:33:58,119][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:33:58,620][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:33:59,120][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:33:59,620][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:34:00,122][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:34:00,622][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:34:01,122][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:34:01,621][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:34:02,122][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:34:02,626][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:34:03,127][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:34:03,628][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:34:04,131][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:34:04,636][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:34:05,141][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10853 tokens. -[2025-11-09 20:34:05,890][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 55.72%, Block Peak % of device VRAM: 62.42%, ΔTime: 00:00:32 -[2025-11-09 20:34:06,631][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:34:06,633][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:34:06,635][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:34:07,583][__main__][INFO] - Iteration 180 took 51s (29.87% Gen, 68.27% Train). Generation: 15s, Training: 34s. Estimated remaining time: 40h 1m 31s. Estimated total time: 42h 36m 41s. Time estimates for 10 more iterations: 8m 31s, 100 more iterations: 1h 25m 13s, 500 more iterations: 7h 6m 6s. -[2025-11-09 20:34:07,585][__main__][INFO] - Starting iteration 180. -[2025-11-09 20:34:08,055][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 17 and human policies 1. -[2025-11-09 20:34:08,055][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:34:23,086][__main__][INFO] - Number of regex retries in iteration 180: 0 -[2025-11-09 20:34:23,087][__main__][INFO] - agents played in iteration 180 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:34:23,861][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:34:23,883][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:34:23,905][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:34:23,927][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:34:23,928][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:34:23,929][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:34:24,549][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:34:25,004][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:34:25,507][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:34:26,013][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:34:26,510][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:34:27,007][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:34:27,511][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:34:28,009][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:34:28,508][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:34:29,003][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:34:29,500][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:34:30,008][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:34:30,504][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:34:31,001][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:34:31,499][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:34:31,995][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:34:32,492][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:34:32,989][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:34:33,486][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:34:33,991][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:34:34,487][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:34:34,985][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:34:35,483][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:34:35,980][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:34:36,477][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:34:36,977][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:34:37,475][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:34:37,973][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:34:38,472][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:34:38,970][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:34:39,478][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:34:39,977][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:34:40,476][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:34:40,972][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:34:41,472][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:34:41,971][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:34:42,471][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:34:42,970][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:34:43,471][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:34:43,966][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:34:44,470][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:34:44,978][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:34:45,481][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:34:45,995][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:34:46,499][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:34:47,000][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:34:47,504][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:34:48,003][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:34:48,509][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:34:49,010][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:34:49,513][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:34:50,014][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:34:50,513][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:34:51,015][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:34:51,512][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:34:52,011][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:34:52,513][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:34:53,014][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:34:53,515][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:34:54,015][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:34:54,518][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:34:55,020][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:34:55,520][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:34:56,022][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:34:56,525][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10861 tokens. -[2025-11-09 20:34:57,234][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 55.72%, Block Peak % of device VRAM: 62.50%, ΔTime: 00:00:32 -[2025-11-09 20:34:57,986][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:34:57,989][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:34:57,990][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:34:59,780][__main__][INFO] - Iteration 181 took 51s (29.06% Gen, 67.48% Train). Generation: 15s, Training: 34s. Estimated remaining time: 40h 30m 13s. Estimated total time: 43h 6m 16s. Time estimates for 10 more iterations: 8m 37s, 100 more iterations: 1h 26m 12s, 500 more iterations: 7h 11m 2s. -[2025-11-09 20:34:59,782][__main__][INFO] - Starting iteration 181. -[2025-11-09 20:35:00,255][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 18 and human policies 1. -[2025-11-09 20:35:00,256][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:35:07,217][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:35:15,278][__main__][INFO] - Number of regex retries in iteration 181: 1 -[2025-11-09 20:35:15,279][__main__][INFO] - agents played in iteration 181 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:35:16,065][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:35:16,093][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:35:16,130][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:35:16,155][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:35:16,156][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:35:16,157][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:35:16,769][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:35:17,228][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:35:17,732][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:35:18,232][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:35:18,731][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:35:19,232][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:35:19,733][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:35:20,236][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:35:20,736][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:35:21,236][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:35:21,734][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:35:22,232][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:35:22,729][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:35:23,226][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:35:23,727][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:35:24,228][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:35:24,726][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:35:25,226][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:35:25,726][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:35:26,227][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:35:26,728][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:35:27,227][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:35:27,724][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:35:28,222][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:35:28,719][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:35:29,222][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:35:29,720][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:35:30,222][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:35:30,724][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:35:31,222][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:35:31,721][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:35:32,225][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:35:32,725][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:35:33,223][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:35:33,723][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:35:34,224][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:35:34,727][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:35:35,228][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:35:35,727][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:35:36,227][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:35:36,725][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:35:37,226][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:35:37,724][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:35:38,225][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:35:38,729][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:35:39,229][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:35:39,730][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:35:40,234][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:35:40,736][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:35:41,238][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:35:41,740][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:35:42,240][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:35:42,744][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:35:43,246][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:35:43,747][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:35:44,249][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:35:44,750][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:35:45,252][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:35:45,752][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:35:46,253][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:35:46,757][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:35:47,259][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:35:47,759][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:35:48,262][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:35:48,761][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10851 tokens. -[2025-11-09 20:35:49,440][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 55.67%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 -[2025-11-09 20:35:50,184][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:35:50,186][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:35:50,187][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:35:51,110][__main__][INFO] - Iteration 182 took 50s (29.54% Gen, 68.64% Train). Generation: 15s, Training: 34s. Estimated remaining time: 39h 45m 54s. Estimated total time: 42h 22m 48s. Time estimates for 10 more iterations: 8m 28s, 100 more iterations: 1h 24m 45s, 500 more iterations: 7h 3m 48s. -[2025-11-09 20:35:51,113][__main__][INFO] - Starting iteration 182. -[2025-11-09 20:35:51,577][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 18 and human policies 1. -[2025-11-09 20:35:51,578][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:35:56,157][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:36:06,420][__main__][INFO] - Number of regex retries in iteration 182: 1 -[2025-11-09 20:36:06,420][__main__][INFO] - agents played in iteration 182 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:36:07,189][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:36:07,216][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:36:07,243][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:36:07,266][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:36:07,266][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:36:07,267][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:36:07,897][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:36:08,354][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:36:08,859][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:36:09,361][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:36:09,860][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:36:10,358][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:36:10,859][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:36:11,359][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:36:11,858][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:36:12,361][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:36:12,860][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:36:13,361][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:36:13,860][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:36:14,363][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:36:14,860][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:36:15,359][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:36:15,862][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:36:16,361][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:36:16,861][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:36:17,361][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:36:17,859][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:36:18,358][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:36:18,856][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:36:19,355][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:36:19,857][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:36:20,356][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:36:20,857][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:36:21,356][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:36:21,854][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:36:22,352][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:36:22,849][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:36:23,347][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:36:23,844][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:36:24,342][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:36:24,839][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:36:25,337][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:36:25,833][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:36:26,337][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:36:26,833][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:36:27,328][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:36:27,831][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:36:28,334][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:36:28,838][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:36:29,340][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:36:29,841][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:36:30,352][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:36:30,855][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:36:31,358][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:36:31,860][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:36:32,362][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:36:32,871][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:36:33,373][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:36:33,874][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:36:34,380][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:36:34,879][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:36:35,380][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:36:35,880][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:36:36,386][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:36:36,889][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:36:37,390][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:36:37,892][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:36:38,399][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:36:38,899][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:36:39,400][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:36:39,902][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10860 tokens. -[2025-11-09 20:36:40,593][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.32%, Current % of VRAM taken: 55.81%, Block Peak % of device VRAM: 62.51%, ΔTime: 00:00:32 -[2025-11-09 20:36:41,344][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:36:41,346][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:36:41,347][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:36:42,257][__main__][INFO] - Iteration 183 took 50s (29.29% Gen, 68.92% Train). Generation: 14s, Training: 34s. Estimated remaining time: 39h 36m 16s. Estimated total time: 42h 14m 1s. Time estimates for 10 more iterations: 8m 26s, 100 more iterations: 1h 24m 28s, 500 more iterations: 7h 2m 20s. -[2025-11-09 20:36:42,259][__main__][INFO] - Starting iteration 183. -[2025-11-09 20:36:42,721][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 18 and human policies 1. -[2025-11-09 20:36:42,722][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:36:46,320][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:36:57,299][__main__][INFO] - Number of regex retries in iteration 183: 1 -[2025-11-09 20:36:57,299][__main__][INFO] - agents played in iteration 183 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:36:58,073][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:36:58,098][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:36:58,122][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:36:58,144][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:36:58,144][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:36:58,145][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:36:58,767][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:36:59,236][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:36:59,741][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:37:00,238][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:37:00,739][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:37:01,237][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:37:01,746][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:37:02,245][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:37:02,743][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:37:03,258][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:37:03,755][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:37:04,253][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:37:04,752][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:37:05,249][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:37:05,752][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:37:06,248][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:37:06,745][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:37:07,246][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:37:07,745][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:37:08,242][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:37:08,741][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:37:09,240][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:37:09,749][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:37:10,247][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:37:10,745][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:37:11,243][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:37:11,740][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:37:12,238][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:37:12,736][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:37:13,233][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:37:13,746][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:37:14,243][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:37:14,740][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:37:15,257][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:37:15,756][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:37:16,253][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:37:16,753][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:37:17,253][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:37:17,757][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:37:18,255][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:37:18,753][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:37:19,279][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:37:19,783][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:37:20,288][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:37:20,786][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:37:21,288][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:37:21,791][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:37:22,297][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:37:22,801][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:37:23,311][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:37:23,815][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:37:24,330][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:37:24,834][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:37:25,336][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:37:25,842][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:37:26,346][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:37:26,857][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:37:27,359][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:37:27,860][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:37:28,365][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:37:28,867][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:37:29,369][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:37:29,870][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:37:30,369][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:37:30,869][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10869 tokens. -[2025-11-09 20:37:31,549][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 55.71%, Block Peak % of device VRAM: 62.42%, ΔTime: 00:00:32 -[2025-11-09 20:37:32,336][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:37:32,338][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:37:32,339][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:37:33,247][__main__][INFO] - Iteration 184 took 50s (28.85% Gen, 69.35% Train). Generation: 14s, Training: 35s. Estimated remaining time: 39h 27m 43s. Estimated total time: 42h 6m 19s. Time estimates for 10 more iterations: 8m 25s, 100 more iterations: 1h 24m 12s, 500 more iterations: 7h 1m 3s. -[2025-11-09 20:37:33,249][__main__][INFO] - Starting iteration 184. -[2025-11-09 20:37:33,714][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 18 and human policies 1. -[2025-11-09 20:37:33,715][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:37:47,189][__main__][INFO] - Number of regex retries in iteration 184: 0 -[2025-11-09 20:37:47,189][__main__][INFO] - agents played in iteration 184 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:37:47,960][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:37:47,985][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:37:48,011][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:37:48,034][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:37:48,035][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:37:48,036][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:37:48,657][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:37:49,113][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:37:49,616][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:37:50,113][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:37:50,610][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:37:51,107][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:37:51,605][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:37:52,107][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:37:52,604][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:37:53,098][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:37:53,595][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:37:54,091][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:37:54,592][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:37:55,088][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:37:55,585][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:37:56,081][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:37:56,579][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:37:57,079][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:37:57,575][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:37:58,072][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:37:58,573][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:37:59,071][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:37:59,568][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:38:00,066][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:38:00,562][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:38:01,060][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:38:01,557][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:38:02,053][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:38:02,566][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:38:03,062][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:38:03,560][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:38:04,060][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:38:04,557][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:38:05,064][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:38:05,562][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:38:06,060][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:38:06,558][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:38:07,056][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:38:07,552][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:38:08,049][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:38:08,545][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:38:09,068][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:38:09,568][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:38:10,066][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:38:10,579][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:38:11,080][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:38:11,578][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:38:12,079][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:38:12,580][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:38:13,085][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:38:13,586][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:38:14,086][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:38:14,588][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:38:15,088][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:38:15,587][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:38:16,087][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:38:16,588][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:38:17,095][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:38:17,596][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:38:18,098][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:38:18,599][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:38:19,098][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:38:19,602][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:38:20,102][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:38:20,604][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10837 tokens. -[2025-11-09 20:38:21,335][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 55.69%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 -[2025-11-09 20:38:22,098][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:38:22,100][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:38:22,105][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:38:22,984][__main__][INFO] - Iteration 185 took 49s (27.35% Gen, 70.86% Train). Generation: 13s, Training: 34s. Estimated remaining time: 38h 24m 5s. Estimated total time: 41h 3m 31s. Time estimates for 10 more iterations: 8m 12s, 100 more iterations: 1h 22m 7s, 500 more iterations: 6h 50m 35s. -[2025-11-09 20:38:22,986][__main__][INFO] - Starting iteration 185. -[2025-11-09 20:38:23,466][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 18 and human policies 1. -[2025-11-09 20:38:23,467][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:38:28,533][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:38:36,969][__main__][INFO] - Number of regex retries in iteration 185: 1 -[2025-11-09 20:38:36,970][__main__][INFO] - agents played in iteration 185 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:38:37,816][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:38:37,841][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:38:37,865][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:38:37,887][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:38:37,888][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:38:37,889][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:38:38,503][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:38:38,960][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:38:39,464][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:38:39,963][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:38:40,462][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:38:40,961][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:38:41,462][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:38:41,961][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:38:42,460][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:38:42,957][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:38:43,457][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:38:43,955][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:38:44,463][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:38:44,963][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:38:45,460][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:38:45,965][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:38:46,463][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:38:46,961][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:38:47,461][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:38:47,958][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:38:48,457][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:38:48,954][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:38:49,449][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:38:49,962][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:38:50,461][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:38:50,958][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:38:51,457][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:38:51,958][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:38:52,456][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:38:52,955][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:38:53,452][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:38:53,949][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:38:54,447][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:38:54,948][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:38:55,445][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:38:55,944][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:38:56,448][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:38:56,947][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:38:57,445][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:38:57,952][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:38:58,451][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:38:58,949][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:38:59,447][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:38:59,944][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:39:00,448][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:39:00,949][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:39:01,451][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:39:01,957][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:39:02,463][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:39:02,972][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:39:03,473][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:39:03,975][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:39:04,483][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:39:04,986][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:39:05,487][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:39:05,985][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:39:06,486][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:39:06,991][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:39:07,493][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:39:07,995][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:39:08,499][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:39:09,000][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:39:09,499][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:39:09,999][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:39:10,500][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10871 tokens. -[2025-11-09 20:39:11,216][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.28%, Current % of VRAM taken: 55.78%, Block Peak % of device VRAM: 62.42%, ΔTime: 00:00:32 -[2025-11-09 20:39:11,971][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:39:11,972][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:39:11,974][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:39:12,875][__main__][INFO] - Iteration 186 took 49s (27.33% Gen, 70.84% Train). Generation: 13s, Training: 35s. Estimated remaining time: 38h 30m 13s. Estimated total time: 41h 10m 29s. Time estimates for 10 more iterations: 8m 14s, 100 more iterations: 1h 22m 20s, 500 more iterations: 6h 51m 44s. -[2025-11-09 20:39:12,878][__main__][INFO] - Starting iteration 186. -[2025-11-09 20:39:13,345][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 18 and human policies 1. -[2025-11-09 20:39:13,346][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:39:26,819][__main__][INFO] - Number of regex retries in iteration 186: 0 -[2025-11-09 20:39:26,820][__main__][INFO] - agents played in iteration 186 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:39:27,608][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:39:27,636][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:39:27,664][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:39:27,700][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:39:27,700][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:39:27,702][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:39:28,335][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:39:28,793][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:39:29,302][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:39:29,936][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:39:30,447][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:39:30,948][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:39:31,449][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:39:31,955][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:39:32,454][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:39:32,966][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:39:33,464][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:39:33,964][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:39:34,467][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:39:34,966][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:39:35,474][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:39:35,972][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:39:36,470][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:39:36,972][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:39:37,468][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:39:37,966][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:39:38,464][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:39:38,962][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:39:39,472][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:39:39,968][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:39:40,464][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:39:40,965][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:39:41,463][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:39:41,971][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:39:42,469][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:39:42,966][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:39:43,467][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:39:43,965][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:39:44,464][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:39:44,962][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:39:45,460][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:39:45,972][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:39:46,470][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:39:46,971][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:39:47,472][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:39:47,971][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:39:48,470][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:39:48,967][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:39:49,465][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:39:49,966][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:39:50,470][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:39:50,970][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:39:51,474][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:39:51,976][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:39:52,475][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:39:52,977][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:39:53,477][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:39:53,979][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:39:54,482][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:39:54,985][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:39:55,486][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:39:55,988][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:39:56,498][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:39:57,000][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:39:57,502][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:39:58,002][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:39:58,504][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:39:59,008][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:39:59,508][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:40:00,008][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:40:00,508][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10863 tokens. -[2025-11-09 20:40:01,169][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 55.72%, Block Peak % of device VRAM: 62.42%, ΔTime: 00:00:32 -[2025-11-09 20:40:01,919][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:40:01,921][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:40:01,922][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:40:02,850][__main__][INFO] - Iteration 187 took 49s (27.22% Gen, 70.91% Train). Generation: 13s, Training: 35s. Estimated remaining time: 38h 34m 12s. Estimated total time: 41h 15m 18s. Time estimates for 10 more iterations: 8m 15s, 100 more iterations: 1h 22m 30s, 500 more iterations: 6h 52m 33s. -[2025-11-09 20:40:02,853][__main__][INFO] - Starting iteration 187. -[2025-11-09 20:40:03,315][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 18 and human policies 1. -[2025-11-09 20:40:03,316][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:40:07,466][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:40:07,468][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:40:18,554][__main__][INFO] - Number of regex retries in iteration 187: 2 -[2025-11-09 20:40:18,555][__main__][INFO] - agents played in iteration 187 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:40:19,334][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:40:19,361][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:40:19,388][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:40:19,411][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:40:19,412][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:40:19,413][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:40:20,037][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:40:20,491][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:40:20,996][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:40:21,497][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:40:21,999][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:40:22,504][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:40:23,002][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:40:23,499][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:40:23,999][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:40:24,497][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:40:24,997][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:40:25,495][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:40:25,994][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:40:26,503][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:40:27,000][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:40:27,497][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:40:27,998][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:40:28,495][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:40:28,991][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:40:29,488][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:40:29,983][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:40:30,479][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:40:30,974][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:40:31,470][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:40:31,977][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:40:32,478][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:40:32,976][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:40:33,473][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:40:33,970][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:40:34,474][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:40:34,968][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:40:35,465][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:40:35,962][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:40:36,459][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:40:36,959][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:40:37,456][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:40:37,952][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:40:38,449][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:40:38,945][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:40:39,441][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:40:39,940][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:40:40,443][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:40:40,944][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:40:41,446][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:40:41,948][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:40:42,450][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:40:42,950][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:40:43,450][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:40:43,951][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:40:44,451][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:40:44,951][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:40:45,451][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:40:45,955][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:40:46,456][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:40:46,956][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:40:47,455][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:40:47,958][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:40:48,459][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:40:48,959][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:40:49,459][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:40:49,962][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:40:50,462][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:40:50,963][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:40:51,461][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:40:51,968][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10858 tokens. -[2025-11-09 20:40:52,640][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 55.71%, Block Peak % of device VRAM: 62.28%, ΔTime: 00:00:32 -[2025-11-09 20:40:53,389][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:40:53,390][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:40:53,392][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:40:54,488][__main__][INFO] - Iteration 188 took 51s (29.78% Gen, 68.08% Train). Generation: 15s, Training: 34s. Estimated remaining time: 39h 56m 43s. Estimated total time: 42h 38m 40s. Time estimates for 10 more iterations: 8m 31s, 100 more iterations: 1h 25m 17s, 500 more iterations: 7h 6m 26s. -[2025-11-09 20:40:54,490][__main__][INFO] - Starting iteration 188. -[2025-11-09 20:40:54,971][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 18 and human policies 1. -[2025-11-09 20:40:54,972][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:41:01,385][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:41:09,504][__main__][INFO] - Number of regex retries in iteration 188: 1 -[2025-11-09 20:41:09,505][__main__][INFO] - agents played in iteration 188 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:41:10,293][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:41:10,318][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:41:10,342][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:41:10,364][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:41:10,364][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:41:10,365][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:41:11,039][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:41:11,502][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:41:12,005][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:41:12,504][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:41:13,008][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:41:13,508][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:41:14,010][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:41:14,509][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:41:15,007][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:41:15,507][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:41:16,006][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:41:16,504][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:41:17,004][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:41:17,501][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:41:18,009][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:41:18,510][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:41:19,007][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:41:19,505][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:41:20,002][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:41:20,500][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:41:20,997][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:41:21,494][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:41:21,992][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:41:22,488][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:41:22,987][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:41:23,492][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:41:23,989][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:41:24,485][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:41:24,982][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:41:25,480][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:41:25,978][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:41:26,474][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:41:26,972][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:41:27,483][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:41:27,981][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:41:28,483][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:41:28,983][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:41:29,482][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:41:29,982][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:41:30,486][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:41:30,988][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:41:31,487][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:41:31,987][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:41:32,486][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:41:32,986][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:41:33,486][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:41:33,995][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:41:34,494][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:41:35,005][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:41:35,509][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:41:36,011][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:41:36,524][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:41:37,025][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:41:37,522][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:41:38,023][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:41:38,524][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:41:39,032][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:41:39,532][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:41:40,032][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:41:40,536][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:41:41,036][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:41:41,536][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:41:42,035][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:41:42,539][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:41:43,040][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10835 tokens. -[2025-11-09 20:41:43,723][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 55.69%, Block Peak % of device VRAM: 62.42%, ΔTime: 00:00:32 -[2025-11-09 20:41:44,498][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:41:44,499][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:41:44,501][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:41:45,439][__main__][INFO] - Iteration 189 took 50s (28.79% Gen, 69.34% Train). Generation: 14s, Training: 34s. Estimated remaining time: 39h 20m 36s. Estimated total time: 42h 3m 24s. Time estimates for 10 more iterations: 8m 24s, 100 more iterations: 1h 24m 6s, 500 more iterations: 7h 0m 34s. -[2025-11-09 20:41:45,441][__main__][INFO] - Starting iteration 189. -[2025-11-09 20:41:45,929][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 18 and human policies 1. -[2025-11-09 20:41:45,930][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:42:00,145][__main__][INFO] - Number of regex retries in iteration 189: 0 -[2025-11-09 20:42:00,145][__main__][INFO] - agents played in iteration 189 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:42:00,919][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:42:00,942][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:42:00,964][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:42:00,986][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:42:00,986][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:42:00,987][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:42:01,621][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:42:02,073][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:42:02,581][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:42:03,086][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:42:03,586][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:42:04,088][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:42:04,587][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:42:05,089][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:42:05,599][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:42:06,102][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:42:06,626][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:42:07,145][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:42:07,646][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:42:08,145][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:42:08,646][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:42:09,146][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:42:09,644][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:42:10,141][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:42:10,640][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:42:11,139][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:42:11,639][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:42:12,137][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:42:12,635][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:42:13,137][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:42:13,637][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:42:14,136][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:42:14,634][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:42:15,132][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:42:15,631][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:42:16,128][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:42:16,626][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:42:17,124][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:42:17,621][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:42:18,118][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:42:18,616][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:42:19,113][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:42:19,609][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:42:20,105][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:42:20,601][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:42:21,098][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:42:21,593][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:42:22,093][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:42:22,596][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:42:23,099][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:42:23,617][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:42:24,115][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:42:24,616][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:42:25,120][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:42:25,621][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:42:26,126][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:42:26,625][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:42:27,127][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:42:27,634][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:42:28,135][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:42:28,638][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:42:29,143][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:42:29,644][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:42:30,143][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:42:30,644][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:42:31,145][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:42:31,647][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:42:32,149][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:42:32,645][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:42:33,143][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:42:33,643][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10861 tokens. -[2025-11-09 20:42:34,322][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.25%, Current % of VRAM taken: 55.75%, Block Peak % of device VRAM: 62.38%, ΔTime: 00:00:32 -[2025-11-09 20:42:35,059][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:42:35,060][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:42:35,062][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:42:35,998][__main__][INFO] - Iteration 190 took 50s (28.39% Gen, 69.74% Train). Generation: 14s, Training: 34s. Estimated remaining time: 38h 59m 48s. Estimated total time: 41h 43m 27s. Time estimates for 10 more iterations: 8m 20s, 100 more iterations: 1h 23m 26s, 500 more iterations: 6h 57m 14s. -[2025-11-09 20:42:36,000][__main__][INFO] - Starting iteration 190. -[2025-11-09 20:42:36,462][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 18 and human policies 1. -[2025-11-09 20:42:36,463][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:42:41,724][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:42:51,820][__main__][INFO] - Number of regex retries in iteration 190: 1 -[2025-11-09 20:42:51,821][__main__][INFO] - agents played in iteration 190 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:42:52,625][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:42:52,653][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:42:52,679][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:42:52,702][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:42:52,702][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:42:52,703][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:42:53,332][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:42:53,798][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:42:54,298][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:42:54,803][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:42:55,311][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:42:55,810][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:42:56,322][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:42:56,825][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:42:57,323][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:42:57,835][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:42:58,333][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:42:58,832][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:42:59,333][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:42:59,832][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:43:00,332][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:43:00,829][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:43:01,326][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:43:01,837][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:43:02,333][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:43:02,828][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:43:03,325][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:43:03,820][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:43:04,326][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:43:04,824][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:43:05,319][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:43:05,829][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:43:06,327][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:43:06,824][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:43:07,323][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:43:07,821][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:43:08,321][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:43:08,821][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:43:09,319][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:43:09,818][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:43:10,314][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:43:10,811][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:43:11,308][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:43:11,805][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:43:12,302][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:43:12,799][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:43:13,296][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:43:13,813][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:43:14,314][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:43:14,818][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:43:15,322][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:43:15,823][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:43:16,324][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:43:16,826][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:43:17,329][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:43:17,828][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:43:18,328][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:43:18,828][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:43:19,331][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:43:19,830][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:43:20,346][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:43:20,844][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:43:21,348][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:43:21,851][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:43:22,351][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:43:22,857][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:43:23,358][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:43:23,857][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:43:24,358][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:43:24,859][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:43:25,361][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10862 tokens. -[2025-11-09 20:43:26,041][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 55.69%, Block Peak % of device VRAM: 62.35%, ΔTime: 00:00:32 -[2025-11-09 20:43:26,776][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:43:26,778][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:43:26,780][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:43:28,632][__main__][INFO] - Iteration 191 took 52s (29.44% Gen, 67.01% Train). Generation: 15s, Training: 34s. Estimated remaining time: 40h 43m 59s. Estimated total time: 43h 28m 30s. Time estimates for 10 more iterations: 8m 41s, 100 more iterations: 1h 26m 57s, 500 more iterations: 7h 14m 45s. -[2025-11-09 20:43:28,634][__main__][INFO] - Starting iteration 191. -[2025-11-09 20:43:29,104][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 19 and human policies 1. -[2025-11-09 20:43:29,105][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:43:38,581][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:43:45,426][__main__][INFO] - Number of regex retries in iteration 191: 1 -[2025-11-09 20:43:45,427][__main__][INFO] - agents played in iteration 191 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:43:46,218][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:43:46,240][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:43:46,262][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:43:46,284][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:43:46,285][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:43:46,285][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:43:46,926][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:43:47,382][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:43:47,885][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:43:48,384][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:43:48,883][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:43:49,386][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:43:49,885][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:43:50,383][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:43:50,881][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:43:51,380][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:43:51,882][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:43:52,384][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:43:52,885][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:43:53,384][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:43:53,881][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:43:54,380][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:43:54,878][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:43:55,379][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:43:55,880][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:43:56,377][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:43:56,876][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:43:57,378][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:43:57,877][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:43:58,375][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:43:58,872][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:43:59,371][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:43:59,869][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:44:00,367][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:44:00,864][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:44:01,362][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:44:01,860][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:44:02,362][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:44:02,860][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:44:03,358][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:44:03,853][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:44:04,349][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:44:04,846][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:44:05,344][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:44:05,845][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:44:06,344][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:44:06,846][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:44:07,346][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:44:07,865][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:44:08,372][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:44:08,874][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:44:09,378][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:44:09,882][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:44:10,386][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:44:10,889][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:44:11,389][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:44:11,889][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:44:12,393][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:44:12,893][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:44:13,394][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:44:13,894][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:44:14,395][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:44:14,893][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:44:15,393][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:44:15,891][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:44:16,394][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:44:16,900][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:44:17,401][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:44:17,901][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:44:18,401][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:44:18,901][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10846 tokens. -[2025-11-09 20:44:19,599][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 55.72%, Block Peak % of device VRAM: 62.27%, ΔTime: 00:00:32 -[2025-11-09 20:44:20,353][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:44:20,355][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:44:20,356][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:44:21,283][__main__][INFO] - Iteration 192 took 52s (31.28% Gen, 66.94% Train). Generation: 16s, Training: 34s. Estimated remaining time: 40h 43m 35s. Estimated total time: 43h 28m 59s. Time estimates for 10 more iterations: 8m 41s, 100 more iterations: 1h 26m 57s, 500 more iterations: 7h 14m 49s. -[2025-11-09 20:44:21,285][__main__][INFO] - Starting iteration 192. -[2025-11-09 20:44:21,748][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 19 and human policies 1. -[2025-11-09 20:44:21,749][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:44:35,368][__main__][INFO] - Number of regex retries in iteration 192: 0 -[2025-11-09 20:44:35,369][__main__][INFO] - agents played in iteration 192 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:44:36,136][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:44:36,164][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:44:36,190][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:44:36,213][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:44:36,213][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:44:36,214][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:44:36,869][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:44:37,324][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:44:37,826][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:44:38,325][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:44:38,823][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:44:39,321][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:44:39,819][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:44:40,316][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:44:40,813][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:44:41,310][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:44:41,810][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:44:42,307][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:44:42,805][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:44:43,303][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:44:43,801][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:44:44,302][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:44:44,803][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:44:45,303][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:44:45,800][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:44:46,299][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:44:46,798][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:44:47,300][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:44:47,799][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:44:48,297][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:44:48,794][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:44:49,287][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:44:49,786][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:44:50,280][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:44:50,775][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:44:51,284][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:44:51,780][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:44:52,278][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:44:52,771][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:44:53,270][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:44:53,775][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:44:54,272][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:44:54,770][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:44:55,273][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:44:55,770][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:44:56,267][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:44:56,765][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:44:57,265][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:44:57,781][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:44:58,284][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:44:58,787][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:44:59,285][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:44:59,787][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:45:00,293][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:45:00,792][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:45:01,299][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:45:01,808][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:45:02,310][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:45:02,811][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:45:03,313][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:45:03,815][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:45:04,316][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:45:04,818][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:45:05,321][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:45:05,823][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:45:06,322][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:45:06,825][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:45:07,325][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:45:07,823][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:45:08,328][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:45:08,827][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10872 tokens. -[2025-11-09 20:45:09,526][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 55.69%, Block Peak % of device VRAM: 62.41%, ΔTime: 00:00:32 -[2025-11-09 20:45:10,265][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:45:10,267][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:45:10,269][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:45:11,194][__main__][INFO] - Iteration 193 took 49s (27.54% Gen, 70.58% Train). Generation: 13s, Training: 34s. Estimated remaining time: 38h 26m 5s. Estimated total time: 41h 12m 19s. Time estimates for 10 more iterations: 8m 14s, 100 more iterations: 1h 22m 24s, 500 more iterations: 6h 52m 3s. -[2025-11-09 20:45:11,197][__main__][INFO] - Starting iteration 193. -[2025-11-09 20:45:11,660][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 19 and human policies 1. -[2025-11-09 20:45:11,660][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:45:25,641][__main__][INFO] - Number of regex retries in iteration 193: 0 -[2025-11-09 20:45:25,641][__main__][INFO] - agents played in iteration 193 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:45:26,443][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:45:26,470][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:45:26,497][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:45:26,520][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:45:26,521][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:45:26,521][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:45:27,149][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:45:27,606][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:45:28,112][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:45:28,609][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:45:29,112][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:45:29,610][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:45:30,108][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:45:30,607][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:45:31,106][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:45:31,604][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:45:32,107][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:45:32,605][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:45:33,106][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:45:33,603][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:45:34,101][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:45:34,605][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:45:35,103][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:45:35,608][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:45:36,108][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:45:36,608][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:45:37,111][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:45:37,609][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:45:38,107][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:45:38,612][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:45:39,118][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:45:39,616][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:45:40,112][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:45:40,609][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:45:41,107][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:45:41,606][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:45:42,103][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:45:42,611][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:45:43,109][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:45:43,607][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:45:44,105][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:45:44,603][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:45:45,111][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:45:45,609][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:45:46,108][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:45:46,608][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:45:47,108][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:45:47,609][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:45:48,109][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:45:48,610][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:45:49,111][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:45:49,614][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:45:50,116][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:45:50,617][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:45:51,121][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:45:51,623][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:45:52,126][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:45:52,628][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:45:53,127][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:45:53,628][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:45:54,131][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:45:54,631][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:45:55,132][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:45:55,635][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:45:56,134][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:45:56,634][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:45:57,140][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:45:57,639][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:45:58,140][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:45:58,642][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:45:59,143][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10866 tokens. -[2025-11-09 20:45:59,818][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.28%, Current % of VRAM taken: 55.78%, Block Peak % of device VRAM: 62.50%, ΔTime: 00:00:32 -[2025-11-09 20:46:00,578][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:46:00,581][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:46:00,583][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:46:01,519][__main__][INFO] - Iteration 194 took 49s (28.04% Gen, 70.08% Train). Generation: 13s, Training: 34s. Estimated remaining time: 38h 45m 54s. Estimated total time: 41h 32m 58s. Time estimates for 10 more iterations: 8m 18s, 100 more iterations: 1h 23m 5s, 500 more iterations: 6h 55m 29s. -[2025-11-09 20:46:01,521][__main__][INFO] - Starting iteration 194. -[2025-11-09 20:46:01,990][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 19 and human policies 1. -[2025-11-09 20:46:01,991][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:46:17,164][__main__][INFO] - Number of regex retries in iteration 194: 0 -[2025-11-09 20:46:17,165][__main__][INFO] - agents played in iteration 194 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:46:17,968][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:46:17,991][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:46:18,014][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:46:18,037][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:46:18,037][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:46:18,039][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:46:18,689][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:46:19,146][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:46:19,651][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:46:20,153][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:46:20,654][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:46:21,152][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:46:21,651][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:46:22,150][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:46:22,648][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:46:23,148][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:46:23,645][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:46:24,143][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:46:24,641][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:46:25,139][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:46:25,640][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:46:26,137][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:46:26,634][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:46:27,135][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:46:27,633][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:46:28,133][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:46:28,637][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:46:29,136][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:46:29,639][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:46:30,139][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:46:30,639][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:46:31,138][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:46:31,635][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:46:32,138][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:46:32,638][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:46:33,139][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:46:33,645][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:46:34,144][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:46:34,641][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:46:35,137][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:46:35,634][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:46:36,129][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:46:36,624][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:46:37,121][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:46:37,621][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:46:38,117][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:46:38,623][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:46:39,143][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:46:39,647][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:46:40,147][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:46:40,651][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:46:41,153][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:46:41,656][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:46:42,161][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:46:42,667][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:46:43,170][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:46:43,672][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:46:44,174][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:46:44,680][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:46:45,185][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:46:45,688][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:46:46,192][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:46:46,694][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:46:47,197][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:46:47,709][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:46:48,210][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:46:48,708][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:46:49,211][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:46:49,710][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:46:50,219][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:46:50,720][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10843 tokens. -[2025-11-09 20:46:51,398][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.27%, ΔTime: 00:00:32 -[2025-11-09 20:46:52,142][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:46:52,144][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:46:52,145][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:46:53,059][__main__][INFO] - Iteration 195 took 51s (29.71% Gen, 68.50% Train). Generation: 15s, Training: 34s. Estimated remaining time: 39h 45m 30s. Estimated total time: 42h 33m 26s. Time estimates for 10 more iterations: 8m 30s, 100 more iterations: 1h 25m 6s, 500 more iterations: 7h 5m 34s. -[2025-11-09 20:46:53,061][__main__][INFO] - Starting iteration 195. -[2025-11-09 20:46:53,522][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 19 and human policies 1. -[2025-11-09 20:46:53,522][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:46:58,470][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 30 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:46:59,445][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 30 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 2/3 -[2025-11-09 20:47:02,688][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:47:08,927][__main__][INFO] - Number of regex retries in iteration 195: 3 -[2025-11-09 20:47:08,928][__main__][INFO] - agents played in iteration 195 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:47:09,718][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:47:09,743][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:47:09,768][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:47:09,790][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:47:09,790][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:47:09,791][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:47:10,473][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:47:10,933][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:47:11,441][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:47:11,943][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:47:12,441][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:47:12,939][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:47:13,438][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:47:13,941][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:47:14,437][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:47:14,936][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:47:15,438][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:47:15,935][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:47:16,433][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:47:16,929][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:47:17,427][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:47:17,927][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:47:18,424][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:47:18,925][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:47:19,433][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:47:19,931][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:47:20,432][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:47:20,929][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:47:21,427][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:47:21,931][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:47:22,429][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:47:22,927][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:47:23,427][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:47:23,924][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:47:24,419][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:47:24,918][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:47:25,417][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:47:25,933][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:47:26,433][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:47:26,932][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:47:27,430][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:47:27,926][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:47:28,428][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:47:28,928][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:47:29,427][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:47:29,928][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:47:30,428][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:47:30,927][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:47:31,424][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:47:31,925][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:47:32,439][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:47:32,941][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:47:33,445][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:47:33,945][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:47:34,446][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:47:34,949][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:47:35,450][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:47:35,951][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:47:36,454][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:47:36,954][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:47:37,455][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:47:37,957][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:47:38,456][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:47:38,956][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:47:39,452][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:47:39,950][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:47:40,449][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:47:40,947][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:47:41,446][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:47:41,944][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:47:42,443][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10848 tokens. -[2025-11-09 20:47:43,130][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.26%, Current % of VRAM taken: 55.76%, Block Peak % of device VRAM: 62.35%, ΔTime: 00:00:32 -[2025-11-09 20:47:43,881][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:47:43,883][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:47:43,884][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:47:44,795][__main__][INFO] - Iteration 196 took 51s (30.05% Gen, 68.17% Train). Generation: 15s, Training: 34s. Estimated remaining time: 39h 54m 54s. Estimated total time: 42h 43m 42s. Time estimates for 10 more iterations: 8m 32s, 100 more iterations: 1h 25m 27s, 500 more iterations: 7h 7m 17s. -[2025-11-09 20:47:44,797][__main__][INFO] - Starting iteration 196. -[2025-11-09 20:47:45,269][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 19 and human policies 1. -[2025-11-09 20:47:45,269][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:47:58,219][__main__][INFO] - Number of regex retries in iteration 196: 0 -[2025-11-09 20:47:58,220][__main__][INFO] - agents played in iteration 196 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:47:59,085][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:47:59,112][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:47:59,139][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:47:59,162][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:47:59,163][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:47:59,164][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:47:59,841][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:48:00,295][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:48:00,805][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:48:01,307][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:48:01,807][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:48:02,308][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:48:02,818][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:48:03,319][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:48:03,835][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:48:04,339][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:48:04,839][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:48:05,340][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:48:05,843][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:48:06,346][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:48:06,846][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:48:07,347][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:48:07,854][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:48:08,349][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:48:08,847][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:48:09,345][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:48:09,841][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:48:10,352][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:48:10,849][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:48:11,347][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:48:11,843][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:48:12,369][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:48:12,870][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:48:13,367][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:48:13,867][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:48:14,370][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:48:14,869][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:48:15,369][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:48:15,870][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:48:16,374][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:48:16,875][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:48:17,374][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:48:17,875][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:48:18,378][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:48:18,880][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:48:19,385][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:48:19,893][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:48:20,394][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:48:20,897][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:48:21,399][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:48:21,899][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:48:22,400][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:48:22,901][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:48:23,406][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:48:23,906][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:48:24,408][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:48:24,908][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:48:25,409][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:48:25,927][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:48:26,429][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:48:26,929][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:48:27,445][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:48:27,945][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:48:28,445][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:48:28,945][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:48:29,444][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:48:29,949][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:48:30,449][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:48:30,948][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:48:31,450][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:48:31,950][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10860 tokens. -[2025-11-09 20:48:32,661][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.42%, ΔTime: 00:00:32 -[2025-11-09 20:48:33,413][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:48:33,414][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:48:33,416][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:48:34,346][__main__][INFO] - Iteration 197 took 49s (26.39% Gen, 71.71% Train). Generation: 12s, Training: 35s. Estimated remaining time: 38h 4m 16s. Estimated total time: 40h 53m 54s. Time estimates for 10 more iterations: 8m 10s, 100 more iterations: 1h 21m 47s, 500 more iterations: 6h 48m 59s. -[2025-11-09 20:48:34,348][__main__][INFO] - Starting iteration 197. -[2025-11-09 20:48:34,816][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 19 and human policies 1. -[2025-11-09 20:48:34,817][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:48:49,407][__main__][INFO] - Number of regex retries in iteration 197: 0 -[2025-11-09 20:48:49,408][__main__][INFO] - agents played in iteration 197 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:48:50,175][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:48:50,204][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:48:50,231][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:48:50,254][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:48:50,255][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:48:50,255][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:48:50,890][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:48:51,349][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:48:51,858][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:48:52,361][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:48:52,864][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:48:53,366][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:48:53,869][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:48:54,372][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:48:54,877][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:48:55,379][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:48:55,881][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:48:56,385][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:48:56,886][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:48:57,386][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:48:57,885][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:48:58,384][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:48:58,885][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:48:59,383][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:48:59,882][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:49:00,381][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:49:00,881][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:49:01,381][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:49:01,880][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:49:02,380][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:49:02,882][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:49:03,380][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:49:03,879][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:49:04,377][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:49:04,872][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:49:05,373][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:49:05,870][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:49:06,367][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:49:06,864][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:49:07,362][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:49:07,859][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:49:08,358][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:49:08,859][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:49:09,364][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:49:09,861][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:49:10,362][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:49:10,861][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:49:11,365][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:49:11,868][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:49:12,371][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:49:12,874][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:49:13,379][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:49:13,883][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:49:14,392][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:49:14,894][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:49:15,397][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:49:15,899][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:49:16,400][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:49:16,903][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:49:17,406][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:49:17,907][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:49:18,409][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:49:18,911][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:49:19,411][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:49:19,913][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:49:20,410][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:49:20,909][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:49:21,407][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:49:21,907][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:49:22,407][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:49:22,907][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10858 tokens. -[2025-11-09 20:49:23,597][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 55.72%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 -[2025-11-09 20:49:24,345][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:49:24,347][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:49:24,349][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:49:25,276][__main__][INFO] - Iteration 198 took 50s (28.92% Gen, 69.24% Train). Generation: 14s, Training: 34s. Estimated remaining time: 39h 12m 32s. Estimated total time: 42h 3m 1s. Time estimates for 10 more iterations: 8m 24s, 100 more iterations: 1h 24m 6s, 500 more iterations: 7h 0m 30s. -[2025-11-09 20:49:25,278][__main__][INFO] - Starting iteration 198. -[2025-11-09 20:49:25,753][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 19 and human policies 1. -[2025-11-09 20:49:25,753][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:49:39,414][__main__][INFO] - Number of regex retries in iteration 198: 0 -[2025-11-09 20:49:39,414][__main__][INFO] - agents played in iteration 198 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:49:40,176][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:49:40,201][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:49:40,228][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:49:40,251][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:49:40,251][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:49:40,252][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:49:40,865][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:49:41,319][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:49:41,822][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:49:42,333][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:49:42,829][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:49:43,331][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:49:43,828][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:49:44,326][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:49:44,829][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:49:45,337][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:49:45,847][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:49:46,358][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:49:46,869][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:49:47,382][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:49:47,882][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:49:48,383][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:49:48,886][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:49:49,389][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:49:49,890][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:49:50,392][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:49:50,891][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:49:51,390][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:49:51,888][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:49:52,387][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:49:52,888][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:49:53,385][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:49:53,884][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:49:54,381][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:49:54,880][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:49:55,379][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:49:55,877][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:49:56,375][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:49:56,873][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:49:57,370][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:49:57,866][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:49:58,376][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:49:58,871][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:49:59,368][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:49:59,865][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:50:00,363][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:50:00,871][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:50:01,373][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:50:01,872][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:50:02,383][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:50:02,884][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:50:03,382][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:50:03,884][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:50:04,386][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:50:04,893][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:50:05,394][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:50:05,895][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:50:06,396][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:50:06,896][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:50:07,395][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:50:07,895][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:50:08,395][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:50:08,903][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:50:09,402][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:50:09,904][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:50:10,409][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:50:10,909][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:50:11,407][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:50:11,906][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:50:12,407][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:50:12,908][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10868 tokens. -[2025-11-09 20:50:13,573][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.26%, Current % of VRAM taken: 55.76%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 -[2025-11-09 20:50:14,339][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:50:14,341][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:50:14,343][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:50:15,280][__main__][INFO] - Iteration 199 took 49s (27.58% Gen, 70.52% Train). Generation: 13s, Training: 34s. Estimated remaining time: 38h 25m 5s. Estimated total time: 41h 16m 23s. Time estimates for 10 more iterations: 8m 15s, 100 more iterations: 1h 22m 32s, 500 more iterations: 6h 52m 43s. -[2025-11-09 20:50:15,285][__main__][INFO] - Starting iteration 199. -[2025-11-09 20:50:15,751][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 19 and human policies 1. -[2025-11-09 20:50:15,752][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:50:26,760][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:50:30,239][__main__][INFO] - Number of regex retries in iteration 199: 1 -[2025-11-09 20:50:30,240][__main__][INFO] - agents played in iteration 199 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:50:31,022][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:50:31,045][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:50:31,067][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:50:31,089][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:50:31,090][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:50:31,091][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:50:31,723][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:50:32,178][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:50:32,681][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:50:33,180][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:50:33,677][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:50:34,177][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:50:34,678][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:50:35,180][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:50:35,678][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:50:36,176][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:50:36,674][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:50:37,171][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:50:37,665][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:50:38,164][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:50:38,667][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:50:39,177][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:50:39,678][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:50:40,181][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:50:40,687][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:50:41,190][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:50:41,694][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:50:42,192][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:50:42,692][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:50:43,195][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:50:43,695][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:50:44,198][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:50:44,699][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:50:45,199][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:50:45,706][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:50:46,205][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:50:46,704][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:50:47,206][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:50:47,707][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:50:48,204][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:50:48,701][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:50:49,200][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:50:49,698][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:50:50,193][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:50:50,692][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:50:51,192][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:50:51,692][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:50:52,193][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:50:52,697][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:50:53,200][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:50:53,702][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:50:54,203][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:50:54,704][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:50:55,206][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:50:55,707][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:50:56,212][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:50:56,713][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:50:57,215][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:50:57,720][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:50:58,222][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:50:58,724][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:50:59,225][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:50:59,726][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:51:00,229][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:51:00,729][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:51:01,231][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:51:01,734][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:51:02,235][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:51:02,739][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:51:03,238][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:51:03,738][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10860 tokens. -[2025-11-09 20:51:04,429][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.25%, Current % of VRAM taken: 55.75%, Block Peak % of device VRAM: 62.27%, ΔTime: 00:00:32 -[2025-11-09 20:51:05,208][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:51:05,209][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:51:05,211][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:51:06,126][__main__][INFO] - Iteration 200 took 50s (28.76% Gen, 69.42% Train). Generation: 14s, Training: 34s. Estimated remaining time: 39h 6m 36s. Estimated total time: 41h 58m 45s. Time estimates for 10 more iterations: 8m 23s, 100 more iterations: 1h 23m 57s, 500 more iterations: 6h 59m 47s. -[2025-11-09 20:51:06,128][__main__][INFO] - Starting iteration 200. -[2025-11-09 20:51:06,620][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 19 and human policies 1. -[2025-11-09 20:51:06,620][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:51:21,853][__main__][INFO] - Number of regex retries in iteration 200: 0 -[2025-11-09 20:51:21,854][__main__][INFO] - agents played in iteration 200 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:51:22,623][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:51:22,647][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:51:22,671][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:51:22,693][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:51:22,694][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:51:22,695][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:51:23,329][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:51:23,789][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:51:24,292][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:51:24,797][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:51:25,297][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:51:25,798][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:51:26,303][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:51:26,801][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:51:27,299][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:51:27,797][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:51:28,297][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:51:28,801][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:51:29,301][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:51:29,802][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:51:30,307][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:51:30,807][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:51:31,311][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:51:31,811][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:51:32,313][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:51:32,816][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:51:33,314][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:51:33,816][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:51:34,318][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:51:34,821][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:51:35,324][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:51:35,828][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:51:36,329][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:51:36,829][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:51:37,328][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:51:37,827][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:51:38,326][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:51:38,822][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:51:39,322][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:51:39,820][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:51:40,318][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:51:40,817][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:51:41,315][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:51:41,813][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:51:42,314][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:51:42,812][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:51:43,314][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:51:43,816][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:51:44,322][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:51:44,828][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:51:45,329][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:51:45,832][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:51:46,335][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:51:46,836][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:51:47,342][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:51:47,848][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:51:48,352][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:51:48,856][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:51:49,360][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:51:49,862][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:51:50,366][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:51:50,871][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:51:51,394][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:51:51,919][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:51:52,419][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:51:52,919][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:51:53,421][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:51:53,923][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:51:54,422][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:51:54,922][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:51:55,425][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10842 tokens. -[2025-11-09 20:51:56,102][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.29%, Current % of VRAM taken: 55.78%, Block Peak % of device VRAM: 62.40%, ΔTime: 00:00:32 -[2025-11-09 20:51:56,852][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:51:56,854][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:51:56,856][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:51:58,900][__main__][INFO] - Iteration 201 took 52s (29.14% Gen, 66.95% Train). Generation: 15s, Training: 35s. Estimated remaining time: 40h 41m 0s. Estimated total time: 43h 34m 2s. Time estimates for 10 more iterations: 8m 42s, 100 more iterations: 1h 27m 8s, 500 more iterations: 7h 15m 40s. -[2025-11-09 20:51:58,902][__main__][INFO] - Starting iteration 201. -[2025-11-09 20:51:59,365][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 20 and human policies 1. -[2025-11-09 20:51:59,366][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:52:12,517][__main__][INFO] - Number of regex retries in iteration 201: 0 -[2025-11-09 20:52:12,518][__main__][INFO] - agents played in iteration 201 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:52:13,368][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:52:13,390][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:52:13,425][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:52:13,449][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:52:13,450][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:52:13,451][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:52:14,075][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:52:14,531][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:52:15,034][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:52:15,533][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:52:16,032][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:52:16,529][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:52:17,027][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:52:17,526][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:52:18,027][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:52:18,524][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:52:19,024][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:52:19,521][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:52:20,017][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:52:20,512][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:52:21,013][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:52:21,513][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:52:22,008][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:52:22,504][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:52:23,015][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:52:23,510][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:52:24,009][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:52:24,508][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:52:25,006][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:52:25,514][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:52:26,014][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:52:26,514][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:52:27,018][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:52:27,517][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:52:28,022][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:52:28,522][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:52:29,023][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:52:29,527][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:52:30,026][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:52:30,525][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:52:31,023][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:52:31,521][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:52:32,019][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:52:32,518][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:52:33,016][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:52:33,522][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:52:34,021][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:52:34,518][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:52:35,022][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:52:35,524][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:52:36,028][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:52:36,528][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:52:37,030][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:52:37,533][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:52:38,034][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:52:38,537][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:52:39,039][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:52:39,540][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:52:40,049][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:52:40,552][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:52:41,055][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:52:41,558][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:52:42,061][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:52:42,566][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:52:43,066][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:52:43,567][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:52:44,072][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:52:44,573][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:52:45,076][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:52:45,591][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:52:46,092][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10869 tokens. -[2025-11-09 20:52:46,773][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.25%, Current % of VRAM taken: 55.75%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 -[2025-11-09 20:52:47,519][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:52:47,520][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:52:47,522][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:52:48,460][__main__][INFO] - Iteration 202 took 49s (26.79% Gen, 71.30% Train). Generation: 13s, Training: 35s. Estimated remaining time: 38h 0m 54s. Estimated total time: 40h 54m 45s. Time estimates for 10 more iterations: 8m 10s, 100 more iterations: 1h 21m 49s, 500 more iterations: 6h 49m 7s. -[2025-11-09 20:52:48,462][__main__][INFO] - Starting iteration 202. -[2025-11-09 20:52:48,923][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 20 and human policies 1. -[2025-11-09 20:52:48,924][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:52:58,477][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:53:04,087][__main__][INFO] - Number of regex retries in iteration 202: 1 -[2025-11-09 20:53:04,088][__main__][INFO] - agents played in iteration 202 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:53:04,865][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:53:04,887][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:53:04,910][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:53:04,932][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:53:04,932][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:53:04,933][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:53:05,550][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:53:06,004][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:53:06,510][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:53:07,007][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:53:07,505][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:53:08,001][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:53:08,500][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:53:09,002][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:53:09,500][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:53:09,999][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:53:10,500][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:53:10,999][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:53:11,497][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:53:11,996][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:53:12,494][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:53:12,996][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:53:13,494][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:53:14,001][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:53:14,500][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:53:14,996][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:53:15,498][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:53:15,996][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:53:16,492][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:53:16,995][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:53:17,492][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:53:17,999][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:53:18,498][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:53:18,999][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:53:19,514][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:53:20,014][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:53:20,515][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:53:21,013][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:53:21,512][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:53:22,014][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:53:22,516][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:53:23,017][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:53:23,521][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:53:24,023][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:53:24,522][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:53:25,022][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:53:25,519][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:53:26,022][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:53:26,524][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:53:27,025][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:53:27,527][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:53:28,030][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:53:28,535][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:53:29,037][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:53:29,539][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:53:30,040][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:53:30,542][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:53:31,043][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:53:31,544][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:53:32,046][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:53:32,556][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:53:33,059][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:53:33,558][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:53:34,061][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:53:34,561][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:53:35,060][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:53:35,560][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:53:36,062][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:53:36,563][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:53:37,062][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:53:37,563][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10859 tokens. -[2025-11-09 20:53:38,285][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 55.69%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 -[2025-11-09 20:53:39,050][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:53:39,051][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:53:39,053][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:53:39,983][__main__][INFO] - Iteration 203 took 51s (29.70% Gen, 68.48% Train). Generation: 15s, Training: 34s. Estimated remaining time: 39h 38m 16s. Estimated total time: 42h 32m 59s. Time estimates for 10 more iterations: 8m 30s, 100 more iterations: 1h 25m 5s, 500 more iterations: 7h 5m 29s. -[2025-11-09 20:53:39,985][__main__][INFO] - Starting iteration 203. -[2025-11-09 20:53:40,451][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 20 and human policies 1. -[2025-11-09 20:53:40,452][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:53:54,464][__main__][INFO] - Number of regex retries in iteration 203: 0 -[2025-11-09 20:53:54,466][__main__][INFO] - agents played in iteration 203 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:53:55,235][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:53:55,264][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:53:55,291][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:53:55,314][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:53:55,314][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:53:55,315][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:53:55,955][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:53:56,410][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:53:56,914][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:53:57,413][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:53:57,910][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:53:58,409][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:53:58,908][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:53:59,405][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:53:59,912][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:54:00,412][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:54:00,910][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:54:01,410][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:54:01,908][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:54:02,406][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:54:02,903][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:54:03,401][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:54:03,921][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:54:04,423][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:54:04,922][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:54:05,420][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:54:05,919][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:54:06,423][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:54:06,922][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:54:07,420][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:54:07,920][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:54:08,417][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:54:08,914][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:54:09,412][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:54:09,910][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:54:10,415][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:54:10,913][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:54:11,409][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:54:11,910][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:54:12,412][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:54:12,911][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:54:13,411][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:54:13,911][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:54:14,424][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:54:14,926][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:54:15,426][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:54:15,926][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:54:16,428][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:54:16,938][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:54:17,441][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:54:17,944][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:54:18,452][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:54:18,959][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:54:19,461][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:54:19,963][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:54:20,464][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:54:20,970][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:54:21,471][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:54:21,971][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:54:22,476][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:54:22,975][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:54:23,475][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:54:23,975][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:54:24,476][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:54:24,980][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:54:25,482][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:54:25,981][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:54:26,492][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:54:26,991][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:54:27,490][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:54:27,989][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10876 tokens. -[2025-11-09 20:54:28,672][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 55.74%, Block Peak % of device VRAM: 62.35%, ΔTime: 00:00:32 -[2025-11-09 20:54:29,413][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:54:29,415][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:54:29,417][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:54:30,328][__main__][INFO] - Iteration 204 took 49s (28.10% Gen, 70.08% Train). Generation: 14s, Training: 34s. Estimated remaining time: 38h 38m 17s. Estimated total time: 41h 33m 50s. Time estimates for 10 more iterations: 8m 18s, 100 more iterations: 1h 23m 7s, 500 more iterations: 6h 55m 38s. -[2025-11-09 20:54:30,330][__main__][INFO] - Starting iteration 204. -[2025-11-09 20:54:30,797][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 20 and human policies 1. -[2025-11-09 20:54:30,797][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:54:34,139][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:54:37,932][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:54:38,245][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:54:40,393][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:54:44,739][__main__][INFO] - Number of regex retries in iteration 204: 4 -[2025-11-09 20:54:44,740][__main__][INFO] - agents played in iteration 204 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:54:45,508][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:54:45,535][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:54:45,568][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:54:45,590][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:54:45,591][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:54:45,591][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:54:46,218][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:54:46,672][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:54:47,181][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:54:47,680][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:54:48,181][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:54:48,680][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:54:49,174][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:54:49,672][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:54:50,168][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:54:50,668][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:54:51,167][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:54:51,688][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:54:52,191][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:54:52,690][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:54:53,186][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:54:53,687][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:54:54,185][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:54:54,682][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:54:55,181][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:54:55,678][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:54:56,174][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:54:56,671][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:54:57,167][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:54:57,667][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:54:58,164][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:54:58,667][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:54:59,170][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:54:59,670][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:55:00,170][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:55:00,668][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:55:01,168][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:55:01,668][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:55:02,168][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:55:02,668][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:55:03,168][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:55:03,666][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:55:04,163][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:55:04,662][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:55:05,164][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:55:05,666][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:55:06,166][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:55:06,668][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:55:07,168][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:55:07,670][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:55:08,172][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:55:08,674][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:55:09,177][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:55:09,681][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:55:10,185][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:55:10,684][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:55:11,200][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:55:11,703][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:55:12,221][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:55:12,725][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:55:13,228][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:55:13,731][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:55:14,232][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:55:14,736][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:55:15,232][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:55:15,728][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:55:16,230][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:55:16,729][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:55:17,228][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:55:17,737][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:55:18,236][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10839 tokens. -[2025-11-09 20:55:18,931][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 55.69%, Block Peak % of device VRAM: 62.39%, ΔTime: 00:00:32 -[2025-11-09 20:55:19,669][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:55:19,671][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:55:19,673][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:55:20,685][__main__][INFO] - Iteration 205 took 49s (27.94% Gen, 70.02% Train). Generation: 13s, Training: 34s. Estimated remaining time: 38h 38m 4s. Estimated total time: 41h 34m 27s. Time estimates for 10 more iterations: 8m 18s, 100 more iterations: 1h 23m 8s, 500 more iterations: 6h 55m 44s. -[2025-11-09 20:55:20,688][__main__][INFO] - Starting iteration 205. -[2025-11-09 20:55:21,168][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 20 and human policies 1. -[2025-11-09 20:55:21,168][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:55:30,112][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:55:37,052][__main__][INFO] - Number of regex retries in iteration 205: 1 -[2025-11-09 20:55:37,053][__main__][INFO] - agents played in iteration 205 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:55:37,830][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:55:37,857][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:55:37,884][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:55:37,906][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:55:37,907][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:55:37,907][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:55:38,534][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:55:38,990][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:55:39,495][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:55:39,997][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:55:40,495][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:55:40,992][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:55:41,490][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:55:41,988][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:55:42,490][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:55:42,988][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:55:43,485][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:55:43,984][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:55:44,483][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:55:44,981][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:55:45,479][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:55:45,979][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:55:46,477][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:55:46,975][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:55:47,472][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:55:47,967][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:55:48,464][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:55:48,958][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:55:49,456][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:55:49,952][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:55:50,450][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:55:50,946][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:55:51,441][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:55:51,938][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:55:52,435][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:55:52,929][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:55:53,425][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:55:53,922][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:55:54,427][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:55:54,925][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:55:55,422][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:55:55,922][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:55:56,420][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:55:56,915][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:55:57,415][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:55:57,912][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:55:58,408][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:55:58,905][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:55:59,403][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:55:59,904][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:56:00,405][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:56:00,905][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:56:01,407][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:56:01,910][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:56:02,413][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:56:02,912][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:56:03,413][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:56:03,916][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:56:04,414][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:56:04,931][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:56:05,428][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:56:05,927][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:56:06,441][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:56:06,939][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:56:07,437][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:56:07,933][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:56:08,430][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:56:08,932][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:56:09,429][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:56:09,928][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:56:10,428][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10831 tokens. -[2025-11-09 20:56:11,121][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 55.64%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 -[2025-11-09 20:56:11,907][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:56:11,909][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:56:11,911][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:56:12,867][__main__][INFO] - Iteration 206 took 51s (30.72% Gen, 67.42% Train). Generation: 15s, Training: 34s. Estimated remaining time: 40h 7m 42s. Estimated total time: 43h 4m 58s. Time estimates for 10 more iterations: 8m 36s, 100 more iterations: 1h 26m 9s, 500 more iterations: 7h 10m 49s. -[2025-11-09 20:56:12,869][__main__][INFO] - Starting iteration 206. -[2025-11-09 20:56:13,332][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 20 and human policies 1. -[2025-11-09 20:56:13,332][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:56:17,196][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:56:28,246][__main__][INFO] - Number of regex retries in iteration 206: 1 -[2025-11-09 20:56:28,246][__main__][INFO] - agents played in iteration 206 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:56:29,019][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:56:29,046][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:56:29,073][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:56:29,095][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:56:29,096][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:56:29,096][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:56:29,719][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:56:30,178][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:56:30,683][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:56:31,182][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:56:31,675][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:56:32,173][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:56:32,670][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:56:33,165][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:56:33,663][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:56:34,161][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:56:34,658][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:56:35,155][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:56:35,652][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:56:36,148][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:56:36,645][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:56:37,144][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:56:37,642][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:56:38,139][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:56:38,636][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:56:39,132][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:56:39,629][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:56:40,125][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:56:40,623][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:56:41,120][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:56:41,630][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:56:42,128][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:56:42,625][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:56:43,125][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:56:43,623][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:56:44,121][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:56:44,621][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:56:45,118][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:56:45,629][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:56:46,127][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:56:46,627][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:56:47,149][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:56:47,648][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:56:48,147][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:56:48,643][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:56:49,144][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:56:49,642][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:56:50,139][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:56:50,638][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:56:51,135][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:56:51,634][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:56:52,137][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:56:52,634][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:56:53,137][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:56:53,643][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:56:54,149][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:56:54,656][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:56:55,162][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:56:55,665][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:56:56,169][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:56:56,672][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:56:57,173][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:56:57,681][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:56:58,183][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:56:58,684][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:56:59,187][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:56:59,688][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:57:00,195][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:57:00,696][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:57:01,200][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:57:01,715][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10829 tokens. -[2025-11-09 20:57:02,392][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.27%, Current % of VRAM taken: 55.77%, Block Peak % of device VRAM: 62.50%, ΔTime: 00:00:32 -[2025-11-09 20:57:03,157][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:57:03,158][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:57:03,160][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:57:04,162][__main__][INFO] - Iteration 207 took 50s (29.34% Gen, 68.69% Train). Generation: 14s, Training: 34s. Estimated remaining time: 39h 23m 26s. Estimated total time: 42h 21m 33s. Time estimates for 10 more iterations: 8m 28s, 100 more iterations: 1h 24m 43s, 500 more iterations: 7h 3m 35s. -[2025-11-09 20:57:04,165][__main__][INFO] - Starting iteration 207. -[2025-11-09 20:57:04,632][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 20 and human policies 1. -[2025-11-09 20:57:04,633][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:57:18,396][__main__][INFO] - Number of regex retries in iteration 207: 0 -[2025-11-09 20:57:18,397][__main__][INFO] - agents played in iteration 207 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:57:19,168][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:57:19,191][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:57:19,213][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:57:19,235][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:57:19,235][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:57:19,236][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:57:19,868][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:57:20,321][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:57:20,823][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:57:21,324][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:57:21,819][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:57:22,319][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:57:22,823][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:57:23,320][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:57:23,844][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:57:24,340][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:57:24,838][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:57:25,338][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:57:25,834][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:57:26,332][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:57:26,829][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:57:27,325][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:57:27,839][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:57:28,339][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:57:28,836][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:57:29,339][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:57:29,837][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:57:30,334][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:57:30,834][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:57:31,332][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:57:31,827][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:57:32,337][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:57:32,833][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:57:33,346][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:57:33,843][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:57:34,341][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:57:34,841][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:57:35,339][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:57:35,837][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:57:36,335][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:57:36,831][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:57:37,335][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:57:37,832][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:57:38,328][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:57:38,826][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:57:39,323][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:57:39,820][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:57:40,317][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:57:40,813][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:57:41,311][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:57:41,807][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:57:42,304][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:57:42,803][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:57:43,301][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:57:43,800][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:57:44,301][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:57:44,800][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:57:45,301][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:57:45,799][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:57:46,299][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:57:46,802][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:57:47,302][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:57:47,805][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:57:48,337][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:57:48,841][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:57:49,348][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:57:49,850][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:57:50,361][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:57:50,864][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:57:51,369][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:57:51,874][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10843 tokens. -[2025-11-09 20:57:52,599][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.32%, Current % of VRAM taken: 55.81%, Block Peak % of device VRAM: 62.50%, ΔTime: 00:00:32 -[2025-11-09 20:57:53,344][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:57:53,346][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:57:53,349][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:57:54,285][__main__][INFO] - Iteration 208 took 49s (27.72% Gen, 70.39% Train). Generation: 13s, Training: 34s. Estimated remaining time: 38h 23m 45s. Estimated total time: 41h 22m 43s. Time estimates for 10 more iterations: 8m 16s, 100 more iterations: 1h 22m 45s, 500 more iterations: 6h 53m 47s. -[2025-11-09 20:57:54,287][__main__][INFO] - Starting iteration 208. -[2025-11-09 20:57:54,746][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 20 and human policies 1. -[2025-11-09 20:57:54,747][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:57:58,495][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:58:03,778][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 20:58:09,423][__main__][INFO] - Number of regex retries in iteration 208: 2 -[2025-11-09 20:58:09,424][__main__][INFO] - agents played in iteration 208 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:58:10,191][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:58:10,218][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:58:10,244][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:58:10,267][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:58:10,268][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:58:10,269][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:58:10,885][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:58:11,340][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:58:11,847][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:58:12,343][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:58:12,839][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:58:13,336][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:58:13,832][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:58:14,327][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:58:14,823][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:58:15,325][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:58:15,822][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:58:16,320][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:58:16,818][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:58:17,318][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:58:17,815][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:58:18,320][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:58:18,819][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:58:19,314][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:58:19,810][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:58:20,305][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:58:20,803][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:58:21,297][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:58:21,793][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:58:22,288][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:58:22,785][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:58:23,290][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:58:23,787][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:58:24,285][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:58:24,785][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:58:25,286][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:58:25,785][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:58:26,282][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:58:26,782][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:58:27,282][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:58:27,780][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:58:28,277][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:58:28,772][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:58:29,268][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:58:29,772][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:58:30,269][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:58:30,765][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:58:31,266][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:58:31,763][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:58:32,260][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:58:32,758][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:58:33,258][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:58:33,760][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:58:34,261][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:58:34,766][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:58:35,278][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:58:35,777][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:58:36,278][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:58:36,773][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:58:37,272][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:58:37,779][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:58:38,280][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:58:38,782][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:58:39,284][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:58:39,786][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:58:40,291][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:58:40,795][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:58:41,296][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:58:41,799][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:58:42,301][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:58:42,805][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10827 tokens. -[2025-11-09 20:58:43,552][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.34%, ΔTime: 00:00:32 -[2025-11-09 20:58:44,306][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:58:44,311][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:58:44,315][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:58:45,204][__main__][INFO] - Iteration 209 took 50s (29.09% Gen, 69.15% Train). Generation: 14s, Training: 34s. Estimated remaining time: 39h 3m 6s. Estimated total time: 42h 2m 54s. Time estimates for 10 more iterations: 8m 24s, 100 more iterations: 1h 24m 5s, 500 more iterations: 7h 0m 29s. -[2025-11-09 20:58:45,206][__main__][INFO] - Starting iteration 209. -[2025-11-09 20:58:45,675][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 20 and human policies 1. -[2025-11-09 20:58:45,675][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:59:01,481][__main__][INFO] - Number of regex retries in iteration 209: 0 -[2025-11-09 20:59:01,482][__main__][INFO] - agents played in iteration 209 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:59:02,254][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:59:02,281][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:59:02,307][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:59:02,329][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:59:02,330][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:59:02,330][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:59:02,945][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:59:03,402][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:59:03,906][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:59:04,403][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:59:04,902][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:59:05,402][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:59:05,902][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:59:06,404][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:59:06,904][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:59:07,405][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:59:07,903][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 20:59:08,400][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 20:59:08,896][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 20:59:09,396][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 20:59:09,894][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 20:59:10,391][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 20:59:10,889][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 20:59:11,385][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 20:59:11,886][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 20:59:12,383][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 20:59:12,882][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 20:59:13,382][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 20:59:13,881][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 20:59:14,379][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 20:59:14,877][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 20:59:15,378][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 20:59:15,881][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 20:59:16,381][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 20:59:16,880][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 20:59:17,378][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 20:59:17,876][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 20:59:18,379][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 20:59:18,877][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 20:59:19,375][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 20:59:19,874][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 20:59:20,371][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 20:59:20,868][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 20:59:21,365][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 20:59:21,862][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 20:59:22,377][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 20:59:22,875][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 20:59:23,375][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 20:59:23,880][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 20:59:24,381][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 20:59:24,887][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 20:59:25,386][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 20:59:25,887][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 20:59:26,389][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 20:59:26,888][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 20:59:27,389][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 20:59:27,893][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 20:59:28,395][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 20:59:28,900][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 20:59:29,404][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 20:59:29,905][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 20:59:30,411][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 20:59:30,916][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 20:59:31,417][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 20:59:31,919][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 20:59:32,420][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 20:59:32,922][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 20:59:33,424][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 20:59:33,924][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 20:59:34,426][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 20:59:34,926][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10848 tokens. -[2025-11-09 20:59:35,678][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 55.69%, Block Peak % of device VRAM: 62.41%, ΔTime: 00:00:32 -[2025-11-09 20:59:36,417][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 20:59:36,419][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 20:59:36,422][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 20:59:37,559][__main__][INFO] - Iteration 210 took 51s (30.46% Gen, 67.34% Train). Generation: 15s, Training: 34s. Estimated remaining time: 40h 13m 33s. Estimated total time: 43h 14m 13s. Time estimates for 10 more iterations: 8m 38s, 100 more iterations: 1h 26m 28s, 500 more iterations: 7h 12m 22s. -[2025-11-09 20:59:37,562][__main__][INFO] - Starting iteration 210. -[2025-11-09 20:59:38,054][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 20 and human policies 1. -[2025-11-09 20:59:38,055][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 20:59:53,473][__main__][INFO] - Number of regex retries in iteration 210: 0 -[2025-11-09 20:59:53,474][__main__][INFO] - agents played in iteration 210 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 20:59:54,253][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:59:54,280][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:59:54,306][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:59:54,329][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 20:59:54,330][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 20:59:54,330][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 20:59:54,984][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 20:59:55,456][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 20:59:55,963][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 20:59:56,467][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 20:59:56,970][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 20:59:57,468][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 20:59:57,970][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 20:59:58,470][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 20:59:58,969][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 20:59:59,472][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 20:59:59,970][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:00:00,468][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:00:00,968][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:00:01,468][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:00:01,985][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:00:02,485][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:00:02,985][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:00:03,497][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:00:03,997][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:00:04,499][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:00:04,997][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:00:05,495][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:00:06,000][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:00:06,499][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:00:07,000][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:00:07,495][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:00:07,995][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:00:08,494][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:00:08,992][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:00:09,488][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:00:09,996][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:00:10,492][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:00:10,990][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:00:11,489][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:00:11,987][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:00:12,492][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:00:12,990][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:00:13,489][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:00:13,998][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:00:14,497][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:00:14,997][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:00:15,494][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:00:15,993][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:00:16,496][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:00:16,995][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:00:17,498][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:00:18,006][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:00:18,506][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:00:19,008][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:00:19,510][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:00:20,011][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:00:20,513][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:00:21,017][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:00:21,520][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:00:22,025][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:00:22,528][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:00:23,030][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:00:23,532][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:00:24,032][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:00:24,534][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:00:25,035][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:00:25,537][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:00:26,039][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:00:26,541][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:00:27,044][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10844 tokens. -[2025-11-09 21:00:27,747][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.35%, ΔTime: 00:00:32 -[2025-11-09 21:00:28,499][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:00:28,501][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:00:28,504][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:00:30,506][__main__][INFO] - Iteration 211 took 52s (29.39% Gen, 66.78% Train). Generation: 15s, Training: 35s. Estimated remaining time: 40h 41m 4s. Estimated total time: 43h 42m 37s. Time estimates for 10 more iterations: 8m 44s, 100 more iterations: 1h 27m 25s, 500 more iterations: 7h 17m 6s. -[2025-11-09 21:00:30,508][__main__][INFO] - Starting iteration 211. -[2025-11-09 21:00:30,977][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 21 and human policies 1. -[2025-11-09 21:00:30,977][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:00:39,648][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 21:00:46,728][__main__][INFO] - Number of regex retries in iteration 211: 1 -[2025-11-09 21:00:46,728][__main__][INFO] - agents played in iteration 211 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:00:47,581][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:00:47,606][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:00:47,631][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:00:47,653][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:00:47,654][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:00:47,654][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:00:48,332][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:00:48,789][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:00:49,297][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:00:49,797][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:00:50,297][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:00:50,796][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:00:51,303][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:00:51,805][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:00:52,307][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:00:52,807][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:00:53,304][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:00:53,806][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:00:54,307][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:00:54,806][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:00:55,305][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:00:55,803][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:00:56,306][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:00:56,805][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:00:57,306][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:00:57,802][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:00:58,299][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:00:58,802][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:00:59,301][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:00:59,798][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:01:00,300][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:01:00,799][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:01:01,296][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:01:01,796][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:01:02,294][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:01:02,804][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:01:03,302][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:01:03,802][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:01:04,303][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:01:04,801][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:01:05,300][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:01:05,802][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:01:06,300][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:01:06,800][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:01:07,299][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:01:07,798][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:01:08,308][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:01:08,810][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:01:09,311][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:01:09,807][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:01:10,311][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:01:10,816][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:01:11,316][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:01:11,817][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:01:12,317][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:01:12,816][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:01:13,318][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:01:13,819][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:01:14,319][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:01:14,825][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:01:15,327][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:01:15,829][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:01:16,333][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:01:16,832][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:01:17,335][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:01:17,836][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:01:18,336][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:01:18,837][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:01:19,338][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:01:19,844][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:01:20,345][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10836 tokens. -[2025-11-09 21:01:21,071][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 55.69%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 -[2025-11-09 21:01:21,813][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:01:21,815][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:01:21,817][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:01:22,864][__main__][INFO] - Iteration 212 took 51s (30.36% Gen, 67.62% Train). Generation: 15s, Training: 35s. Estimated remaining time: 40h 11m 57s. Estimated total time: 43h 14m 23s. Time estimates for 10 more iterations: 8m 38s, 100 more iterations: 1h 26m 28s, 500 more iterations: 7h 12m 23s. -[2025-11-09 21:01:22,866][__main__][INFO] - Starting iteration 212. -[2025-11-09 21:01:23,349][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 21 and human policies 1. -[2025-11-09 21:01:23,349][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:01:26,657][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 21:01:37,060][__main__][INFO] - Number of regex retries in iteration 212: 1 -[2025-11-09 21:01:37,061][__main__][INFO] - agents played in iteration 212 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:01:37,835][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:01:37,861][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:01:37,886][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:01:37,908][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:01:37,909][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:01:37,910][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:01:38,553][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:01:39,008][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:01:39,518][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:01:40,017][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:01:40,514][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:01:41,015][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:01:41,514][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:01:42,013][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:01:42,511][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:01:43,008][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:01:43,513][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:01:44,010][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:01:44,510][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:01:45,012][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:01:45,510][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:01:46,008][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:01:46,504][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:01:47,003][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:01:47,517][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:01:48,014][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:01:48,511][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:01:49,023][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:01:49,520][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:01:50,019][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:01:50,517][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:01:51,013][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:01:51,517][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:01:52,013][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:01:52,509][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:01:53,005][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:01:53,502][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:01:54,007][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:01:54,503][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:01:54,999][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:01:55,501][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:01:56,000][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:01:56,501][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:01:57,000][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:01:57,496][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:01:58,009][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:01:58,510][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:01:59,013][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:01:59,511][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:02:00,015][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:02:00,519][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:02:01,016][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:02:01,516][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:02:02,018][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:02:02,520][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:02:03,024][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:02:03,525][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:02:04,023][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:02:04,526][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:02:05,025][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:02:05,525][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:02:06,023][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:02:06,522][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:02:07,020][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:02:07,517][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:02:08,013][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:02:08,517][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:02:09,023][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:02:09,524][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:02:10,025][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:02:10,525][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10847 tokens. -[2025-11-09 21:02:11,231][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 55.66%, Block Peak % of device VRAM: 62.27%, ΔTime: 00:00:32 -[2025-11-09 21:02:11,956][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:02:11,958][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:02:11,959][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:02:12,943][__main__][INFO] - Iteration 213 took 49s (27.65% Gen, 70.37% Train). Generation: 13s, Training: 34s. Estimated remaining time: 38h 16m 30s. Estimated total time: 41h 19m 46s. Time estimates for 10 more iterations: 8m 15s, 100 more iterations: 1h 22m 39s, 500 more iterations: 6h 53m 17s. -[2025-11-09 21:02:12,945][__main__][INFO] - Starting iteration 213. -[2025-11-09 21:02:13,406][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 21 and human policies 1. -[2025-11-09 21:02:13,407][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:02:28,397][__main__][INFO] - Number of regex retries in iteration 213: 0 -[2025-11-09 21:02:28,397][__main__][INFO] - agents played in iteration 213 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:02:29,179][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:02:29,206][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:02:29,233][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:02:29,255][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:02:29,256][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:02:29,257][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:02:29,902][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:02:30,355][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:02:30,863][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:02:31,362][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:02:31,861][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:02:32,368][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:02:32,867][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:02:33,366][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:02:33,866][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:02:34,365][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:02:34,862][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:02:35,362][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:02:35,860][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:02:36,361][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:02:36,859][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:02:37,361][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:02:37,860][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:02:38,359][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:02:38,858][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:02:39,357][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:02:39,857][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:02:40,363][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:02:40,863][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:02:41,361][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:02:41,860][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:02:42,358][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:02:42,860][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:02:43,360][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:02:43,858][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:02:44,360][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:02:44,859][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:02:45,359][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:02:45,859][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:02:46,357][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:02:46,854][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:02:47,352][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:02:47,850][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:02:48,347][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:02:48,844][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:02:49,341][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:02:49,839][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:02:50,368][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:02:50,873][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:02:51,375][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:02:51,875][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:02:52,387][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:02:52,888][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:02:53,400][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:02:53,901][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:02:54,402][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:02:54,924][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:02:55,425][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:02:55,926][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:02:56,426][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:02:56,925][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:02:57,430][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:02:57,930][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:02:58,430][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:02:58,927][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:02:59,427][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:02:59,932][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:03:00,432][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:03:00,935][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:03:01,435][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:03:01,936][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10866 tokens. -[2025-11-09 21:03:02,688][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.38%, ΔTime: 00:00:32 -[2025-11-09 21:03:03,426][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:03:03,428][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:03:03,430][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:03:04,418][__main__][INFO] - Iteration 214 took 51s (29.39% Gen, 68.67% Train). Generation: 14s, Training: 35s. Estimated remaining time: 39h 26m 30s. Estimated total time: 42h 30m 37s. Time estimates for 10 more iterations: 8m 30s, 100 more iterations: 1h 25m 1s, 500 more iterations: 7h 5m 6s. -[2025-11-09 21:03:04,420][__main__][INFO] - Starting iteration 214. -[2025-11-09 21:03:04,881][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 21 and human policies 1. -[2025-11-09 21:03:04,882][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:03:13,593][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 21:03:20,100][__main__][INFO] - Number of regex retries in iteration 214: 1 -[2025-11-09 21:03:20,101][__main__][INFO] - agents played in iteration 214 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:03:20,882][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:03:20,909][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:03:20,935][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:03:20,958][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:03:20,958][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:03:20,959][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:03:21,582][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:03:22,042][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:03:22,545][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:03:23,043][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:03:23,546][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:03:24,044][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:03:24,544][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:03:25,044][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:03:25,543][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:03:26,040][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:03:26,535][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:03:27,029][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:03:27,528][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:03:28,023][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:03:28,518][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:03:29,015][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:03:29,511][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:03:30,020][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:03:30,518][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:03:31,018][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:03:31,520][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:03:32,017][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:03:32,515][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:03:33,011][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:03:33,508][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:03:34,031][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:03:34,531][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:03:35,032][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:03:35,531][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:03:36,028][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:03:36,530][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:03:37,026][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:03:37,525][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:03:38,022][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:03:38,518][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:03:39,014][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:03:39,511][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:03:40,011][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:03:40,525][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:03:41,025][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:03:41,526][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:03:42,035][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:03:42,535][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:03:43,041][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:03:43,538][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:03:44,040][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:03:44,543][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:03:45,042][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:03:45,542][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:03:46,048][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:03:46,550][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:03:47,049][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:03:47,553][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:03:48,050][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:03:48,551][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:03:49,050][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:03:49,549][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:03:50,050][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:03:50,547][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:03:51,045][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:03:51,544][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:03:52,044][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:03:52,548][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:03:53,046][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:03:53,546][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10851 tokens. -[2025-11-09 21:03:54,254][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 55.71%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 -[2025-11-09 21:03:55,003][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:03:55,005][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:03:55,006][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:03:55,998][__main__][INFO] - Iteration 215 took 51s (29.77% Gen, 68.29% Train). Generation: 15s, Training: 34s. Estimated remaining time: 39h 30m 53s. Estimated total time: 42h 35m 52s. Time estimates for 10 more iterations: 8m 31s, 100 more iterations: 1h 25m 11s, 500 more iterations: 7h 5m 58s. -[2025-11-09 21:03:56,000][__main__][INFO] - Starting iteration 215. -[2025-11-09 21:03:56,463][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 21 and human policies 1. -[2025-11-09 21:03:56,463][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:04:05,774][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 21:04:10,332][__main__][INFO] - Number of regex retries in iteration 215: 1 -[2025-11-09 21:04:10,333][__main__][INFO] - agents played in iteration 215 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:04:11,149][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:04:11,173][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:04:11,195][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:04:11,218][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:04:11,218][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:04:11,219][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:04:11,946][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:04:12,403][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:04:12,909][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:04:13,411][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:04:13,912][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:04:14,415][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:04:14,914][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:04:15,413][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:04:15,928][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:04:16,424][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:04:16,922][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:04:17,424][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:04:17,921][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:04:18,417][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:04:18,915][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:04:19,412][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:04:19,908][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:04:20,409][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:04:20,907][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:04:21,406][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:04:21,905][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:04:22,428][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:04:22,928][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:04:23,426][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:04:23,944][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:04:24,446][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:04:24,947][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:04:25,446][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:04:25,949][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:04:26,453][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:04:26,952][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:04:27,452][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:04:27,951][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:04:28,454][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:04:28,988][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:04:29,492][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:04:29,995][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:04:30,496][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:04:31,005][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:04:31,513][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:04:32,014][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:04:32,512][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:04:33,013][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:04:33,512][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:04:34,013][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:04:34,518][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:04:35,018][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:04:35,530][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:04:36,029][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:04:36,529][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:04:37,027][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:04:37,525][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:04:38,022][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:04:38,520][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:04:39,018][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:04:39,516][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:04:40,014][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:04:40,509][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:04:41,008][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:04:41,506][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:04:42,009][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:04:42,509][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:04:43,009][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:04:43,509][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:04:44,011][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10848 tokens. -[2025-11-09 21:04:44,773][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 55.71%, Block Peak % of device VRAM: 62.42%, ΔTime: 00:00:32 -[2025-11-09 21:04:45,486][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:04:45,487][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:04:45,489][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:04:46,430][__main__][INFO] - Iteration 216 took 49s (27.76% Gen, 70.36% Train). Generation: 13s, Training: 35s. Estimated remaining time: 38h 32m 33s. Estimated total time: 41h 38m 22s. Time estimates for 10 more iterations: 8m 19s, 100 more iterations: 1h 23m 16s, 500 more iterations: 6h 56m 23s. -[2025-11-09 21:04:46,432][__main__][INFO] - Starting iteration 216. -[2025-11-09 21:04:46,897][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 21 and human policies 1. -[2025-11-09 21:04:46,898][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:04:51,716][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 21:05:01,835][__main__][INFO] - Number of regex retries in iteration 216: 1 -[2025-11-09 21:05:01,836][__main__][INFO] - agents played in iteration 216 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:05:02,634][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:05:02,669][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:05:02,695][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:05:02,717][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:05:02,718][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:05:02,719][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:05:03,377][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:05:03,832][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:05:04,352][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:05:04,851][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:05:05,354][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:05:05,855][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:05:06,355][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:05:06,862][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:05:07,364][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:05:07,866][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:05:08,366][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:05:08,872][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:05:09,373][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:05:09,872][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:05:10,371][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:05:10,871][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:05:11,370][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:05:11,870][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:05:12,369][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:05:12,869][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:05:13,369][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:05:13,870][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:05:14,372][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:05:14,872][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:05:15,372][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:05:15,876][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:05:16,377][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:05:16,879][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:05:17,381][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:05:17,885][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:05:18,388][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:05:18,892][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:05:19,394][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:05:19,894][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:05:20,394][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:05:20,892][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:05:21,397][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:05:21,904][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:05:22,409][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:05:22,912][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:05:23,415][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:05:23,915][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:05:24,416][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:05:24,922][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:05:25,423][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:05:25,924][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:05:26,427][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:05:26,927][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:05:27,427][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:05:27,926][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:05:28,425][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:05:28,937][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:05:29,439][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:05:29,936][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:05:30,438][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:05:30,935][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:05:31,446][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:05:31,947][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:05:32,447][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:05:32,947][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:05:33,445][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:05:33,947][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:05:34,449][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:05:34,949][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:05:35,459][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10857 tokens. -[2025-11-09 21:05:36,164][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.26%, Current % of VRAM taken: 55.76%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 -[2025-11-09 21:05:36,904][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:05:36,905][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:05:36,907][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:05:37,952][__main__][INFO] - Iteration 217 took 51s (29.26% Gen, 68.69% Train). Generation: 14s, Training: 35s. Estimated remaining time: 39h 26m 4s. Estimated total time: 42h 32m 45s. Time estimates for 10 more iterations: 8m 30s, 100 more iterations: 1h 25m 5s, 500 more iterations: 7h 5m 27s. -[2025-11-09 21:05:37,955][__main__][INFO] - Starting iteration 217. -[2025-11-09 21:05:38,420][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 21 and human policies 1. -[2025-11-09 21:05:38,420][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:05:53,260][__main__][INFO] - Number of regex retries in iteration 217: 0 -[2025-11-09 21:05:53,261][__main__][INFO] - agents played in iteration 217 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:05:54,054][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:05:54,081][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:05:54,107][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:05:54,129][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:05:54,130][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:05:54,131][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:05:54,773][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:05:55,235][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:05:55,742][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:05:56,244][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:05:56,764][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:05:57,264][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:05:57,767][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:05:58,269][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:05:58,769][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:05:59,279][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:05:59,789][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:06:00,299][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:06:00,802][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:06:01,303][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:06:01,802][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:06:02,303][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:06:02,804][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:06:03,308][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:06:03,808][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:06:04,309][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:06:04,819][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:06:05,320][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:06:05,829][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:06:06,329][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:06:06,829][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:06:07,328][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:06:07,827][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:06:08,326][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:06:08,828][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:06:09,327][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:06:09,842][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:06:10,342][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:06:10,842][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:06:11,345][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:06:11,849][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:06:12,357][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:06:12,859][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:06:13,364][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:06:13,865][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:06:14,371][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:06:14,873][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:06:15,375][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:06:15,878][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:06:16,382][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:06:16,883][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:06:17,384][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:06:17,886][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:06:18,386][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:06:18,886][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:06:19,386][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:06:19,883][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:06:20,385][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:06:20,885][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:06:21,385][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:06:21,885][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:06:22,388][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:06:22,887][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:06:23,387][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:06:23,887][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:06:24,388][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:06:24,890][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:06:25,393][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:06:25,895][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:06:26,397][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:06:26,901][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10858 tokens. -[2025-11-09 21:06:27,649][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.25%, Current % of VRAM taken: 55.75%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 -[2025-11-09 21:06:28,386][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:06:28,388][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:06:28,390][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:06:29,403][__main__][INFO] - Iteration 218 took 50s (29.11% Gen, 68.90% Train). Generation: 14s, Training: 35s. Estimated remaining time: 39h 21m 38s. Estimated total time: 42h 29m 11s. Time estimates for 10 more iterations: 8m 29s, 100 more iterations: 1h 24m 58s, 500 more iterations: 7h 4m 51s. -[2025-11-09 21:06:29,405][__main__][INFO] - Starting iteration 218. -[2025-11-09 21:06:29,873][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 21 and human policies 1. -[2025-11-09 21:06:29,874][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:06:34,154][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 21:06:41,839][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 21:06:45,177][__main__][INFO] - Number of regex retries in iteration 218: 2 -[2025-11-09 21:06:45,178][__main__][INFO] - agents played in iteration 218 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:06:45,951][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:06:45,974][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:06:45,997][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:06:46,018][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:06:46,019][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:06:46,020][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:06:46,664][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:06:47,121][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:06:47,626][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:06:48,127][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:06:48,627][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:06:49,138][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:06:49,634][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:06:50,135][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:06:50,643][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:06:51,148][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:06:51,661][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:06:52,163][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:06:52,665][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:06:53,169][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:06:53,672][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:06:54,174][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:06:54,672][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:06:55,171][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:06:55,669][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:06:56,169][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:06:56,669][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:06:57,176][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:06:57,673][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:06:58,174][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:06:58,672][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:06:59,172][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:06:59,670][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:07:00,170][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:07:00,667][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:07:01,167][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:07:01,667][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:07:02,165][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:07:02,665][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:07:03,169][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:07:03,674][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:07:04,174][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:07:04,674][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:07:05,175][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:07:05,672][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:07:06,175][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:07:06,674][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:07:07,177][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:07:07,681][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:07:08,184][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:07:08,688][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:07:09,191][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:07:09,688][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:07:10,187][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:07:10,686][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:07:11,193][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:07:11,690][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:07:12,189][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:07:12,689][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:07:13,188][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:07:13,688][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:07:14,187][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:07:14,687][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:07:15,187][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:07:15,687][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:07:16,187][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:07:16,689][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:07:17,187][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:07:17,687][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:07:18,190][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:07:18,689][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10823 tokens. -[2025-11-09 21:07:19,439][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 55.62%, Block Peak % of device VRAM: 62.35%, ΔTime: 00:00:32 -[2025-11-09 21:07:20,161][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:07:20,164][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:07:20,166][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:07:21,191][__main__][INFO] - Iteration 219 took 51s (29.82% Gen, 68.18% Train). Generation: 15s, Training: 34s. Estimated remaining time: 39h 37m 30s. Estimated total time: 42h 45m 54s. Time estimates for 10 more iterations: 8m 33s, 100 more iterations: 1h 25m 31s, 500 more iterations: 7h 7m 39s. -[2025-11-09 21:07:21,193][__main__][INFO] - Starting iteration 219. -[2025-11-09 21:07:21,657][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 21 and human policies 1. -[2025-11-09 21:07:21,658][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:07:35,416][__main__][INFO] - Number of regex retries in iteration 219: 0 -[2025-11-09 21:07:35,417][__main__][INFO] - agents played in iteration 219 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:07:36,185][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:07:36,212][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:07:36,239][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:07:36,263][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:07:36,264][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:07:36,265][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:07:36,922][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:07:37,378][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:07:37,884][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:07:38,384][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:07:38,881][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:07:39,381][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:07:39,882][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:07:40,387][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:07:40,887][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:07:41,388][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:07:41,891][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:07:42,391][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:07:42,888][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:07:43,391][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:07:43,892][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:07:44,396][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:07:44,897][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:07:45,399][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:07:45,901][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:07:46,401][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:07:46,900][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:07:47,403][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:07:47,904][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:07:48,404][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:07:48,903][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:07:49,403][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:07:49,902][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:07:50,402][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:07:50,902][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:07:51,402][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:07:51,904][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:07:52,408][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:07:52,909][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:07:53,410][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:07:53,913][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:07:54,416][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:07:54,921][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:07:55,421][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:07:55,923][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:07:56,427][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:07:56,928][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:07:57,429][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:07:57,929][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:07:58,430][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:07:58,940][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:07:59,446][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:07:59,948][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:08:00,454][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:08:00,957][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:08:01,456][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:08:01,958][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:08:02,457][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:08:02,977][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:08:03,480][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:08:03,981][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:08:04,482][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:08:04,983][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:08:05,486][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:08:05,990][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:08:06,491][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:08:06,994][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:08:07,495][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:08:07,998][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:08:08,498][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:08:09,000][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10856 tokens. -[2025-11-09 21:08:09,769][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 55.74%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:32 -[2025-11-09 21:08:10,500][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:08:10,501][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:08:10,503][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:08:11,517][__main__][INFO] - Iteration 220 took 49s (27.60% Gen, 70.37% Train). Generation: 13s, Training: 35s. Estimated remaining time: 38h 23m 45s. Estimated total time: 41h 32m 59s. Time estimates for 10 more iterations: 8m 18s, 100 more iterations: 1h 23m 5s, 500 more iterations: 6h 55m 29s. -[2025-11-09 21:08:11,519][__main__][INFO] - Starting iteration 220. -[2025-11-09 21:08:11,990][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 21 and human policies 1. -[2025-11-09 21:08:11,990][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:08:16,735][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 21:08:27,466][__main__][INFO] - Number of regex retries in iteration 220: 1 -[2025-11-09 21:08:27,467][__main__][INFO] - agents played in iteration 220 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:08:28,242][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:08:28,270][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:08:28,296][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:08:28,319][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:08:28,319][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:08:28,320][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:08:28,975][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:08:29,432][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:08:29,934][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:08:30,432][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:08:30,929][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:08:31,427][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:08:31,926][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:08:32,427][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:08:32,927][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:08:33,428][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:08:33,930][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:08:34,430][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:08:34,930][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:08:35,431][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:08:35,930][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:08:36,430][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:08:36,942][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:08:37,447][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:08:37,955][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:08:38,457][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:08:38,959][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:08:39,457][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:08:39,959][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:08:40,459][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:08:40,959][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:08:41,457][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:08:41,955][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:08:42,452][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:08:42,950][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:08:43,448][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:08:43,950][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:08:44,455][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:08:44,955][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:08:45,456][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:08:45,954][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:08:46,454][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:08:46,952][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:08:47,449][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:08:47,947][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:08:48,468][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:08:48,969][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:08:49,470][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:08:49,977][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:08:50,479][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:08:50,990][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:08:51,491][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:08:51,992][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:08:52,491][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:08:52,995][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:08:53,496][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:08:53,995][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:08:54,496][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:08:54,998][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:08:55,498][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:08:56,001][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:08:56,503][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:08:57,002][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:08:57,501][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:08:58,001][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:08:58,499][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:08:59,004][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:08:59,505][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:09:00,007][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:09:00,514][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:09:01,015][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10836 tokens. -[2025-11-09 21:09:01,745][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:32 -[2025-11-09 21:09:02,487][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:09:02,488][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:09:02,490][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:09:04,441][__main__][INFO] - Iteration 221 took 52s (29.51% Gen, 66.77% Train). Generation: 15s, Training: 35s. Estimated remaining time: 40h 32m 28s. Estimated total time: 43h 42m 35s. Time estimates for 10 more iterations: 8m 44s, 100 more iterations: 1h 27m 25s, 500 more iterations: 7h 17m 5s. -[2025-11-09 21:09:04,443][__main__][INFO] - Starting iteration 221. -[2025-11-09 21:09:04,911][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 22 and human policies 1. -[2025-11-09 21:09:04,912][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:09:21,847][__main__][INFO] - Number of regex retries in iteration 221: 0 -[2025-11-09 21:09:21,847][__main__][INFO] - agents played in iteration 221 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:09:22,643][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:09:22,668][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:09:22,693][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:09:22,715][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:09:22,716][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:09:22,717][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:09:23,351][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:09:23,812][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:09:24,316][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:09:24,817][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:09:25,318][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:09:25,817][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:09:26,316][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:09:26,811][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:09:27,311][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:09:27,811][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:09:28,313][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:09:28,818][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:09:29,324][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:09:29,842][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:09:30,342][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:09:30,845][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:09:31,345][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:09:31,852][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:09:32,355][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:09:32,857][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:09:33,373][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:09:33,874][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:09:34,390][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:09:34,890][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:09:35,389][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:09:35,893][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:09:36,393][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:09:36,893][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:09:37,397][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:09:37,897][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:09:38,404][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:09:38,908][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:09:39,410][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:09:39,915][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:09:40,415][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:09:40,920][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:09:41,421][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:09:41,924][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:09:42,430][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:09:42,931][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:09:43,432][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:09:43,934][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:09:44,436][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:09:44,938][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:09:45,440][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:09:45,939][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:09:46,439][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:09:46,938][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:09:47,438][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:09:47,943][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:09:48,446][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:09:48,947][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:09:49,448][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:09:49,947][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:09:50,447][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:09:50,949][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:09:51,448][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:09:51,947][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:09:52,447][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:09:52,951][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:09:53,450][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:09:53,950][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:09:54,453][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:09:54,955][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:09:55,461][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10816 tokens. -[2025-11-09 21:09:56,173][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.27%, Current % of VRAM taken: 55.77%, Block Peak % of device VRAM: 62.42%, ΔTime: 00:00:32 -[2025-11-09 21:09:56,910][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:09:56,912][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:09:56,914][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:09:57,886][__main__][INFO] - Iteration 222 took 52s (31.97% Gen, 66.20% Train). Generation: 16s, Training: 35s. Estimated remaining time: 40h 57m 45s. Estimated total time: 44h 8m 46s. Time estimates for 10 more iterations: 8m 49s, 100 more iterations: 1h 28m 17s, 500 more iterations: 7h 21m 27s. -[2025-11-09 21:09:57,888][__main__][INFO] - Starting iteration 222. -[2025-11-09 21:09:58,354][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 22 and human policies 1. -[2025-11-09 21:09:58,355][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:10:10,879][__main__][INFO] - Number of regex retries in iteration 222: 0 -[2025-11-09 21:10:10,880][__main__][INFO] - agents played in iteration 222 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:10:11,795][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:10:11,822][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:10:11,848][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:10:11,870][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:10:11,871][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:10:11,872][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:10:12,514][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:10:12,969][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:10:13,473][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:10:13,976][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:10:14,475][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:10:14,973][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:10:15,476][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:10:15,973][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:10:16,469][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:10:16,966][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:10:17,464][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:10:17,966][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:10:18,463][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:10:18,961][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:10:19,459][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:10:19,958][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:10:20,459][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:10:20,960][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:10:21,462][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:10:21,968][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:10:22,469][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:10:22,969][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:10:23,471][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:10:23,972][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:10:24,477][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:10:24,977][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:10:25,477][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:10:25,981][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:10:26,481][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:10:26,982][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:10:27,485][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:10:27,986][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:10:28,507][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:10:29,013][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:10:29,515][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:10:30,016][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:10:30,515][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:10:31,018][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:10:31,519][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:10:32,021][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:10:32,532][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:10:33,037][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:10:33,541][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:10:34,050][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:10:34,554][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:10:35,054][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:10:35,561][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:10:36,064][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:10:36,569][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:10:37,067][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:10:37,567][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:10:38,069][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:10:38,570][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:10:39,070][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:10:39,569][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:10:40,067][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:10:40,569][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:10:41,069][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:10:41,570][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:10:42,072][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:10:42,571][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:10:43,071][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:10:43,570][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:10:44,071][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:10:44,575][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10869 tokens. -[2025-11-09 21:10:45,306][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 55.69%, Block Peak % of device VRAM: 62.53%, ΔTime: 00:00:32 -[2025-11-09 21:10:46,094][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:10:46,095][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:10:46,097][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:10:47,111][__main__][INFO] - Iteration 223 took 48s (25.69% Gen, 72.23% Train). Generation: 12s, Training: 35s. Estimated remaining time: 37h 26m 1s. Estimated total time: 40h 37m 51s. Time estimates for 10 more iterations: 8m 7s, 100 more iterations: 1h 21m 15s, 500 more iterations: 6h 46m 18s. -[2025-11-09 21:10:47,113][__main__][INFO] - Starting iteration 223. -[2025-11-09 21:10:47,585][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 22 and human policies 1. -[2025-11-09 21:10:47,586][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:11:02,326][__main__][INFO] - Number of regex retries in iteration 223: 0 -[2025-11-09 21:11:02,327][__main__][INFO] - agents played in iteration 223 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:11:03,235][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:11:03,260][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:11:03,284][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:11:03,306][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:11:03,306][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:11:03,308][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:11:03,934][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:11:04,391][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:11:04,895][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:11:05,395][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:11:05,894][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:11:06,392][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:11:06,898][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:11:07,400][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:11:07,898][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:11:08,396][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:11:08,896][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:11:09,396][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:11:09,898][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:11:10,400][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:11:10,910][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:11:11,407][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:11:11,905][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:11:12,402][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:11:12,898][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:11:13,400][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:11:13,901][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:11:14,400][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:11:14,908][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:11:15,407][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:11:15,907][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:11:16,405][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:11:16,906][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:11:17,406][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:11:17,905][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:11:18,404][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:11:18,914][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:11:19,412][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:11:19,911][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:11:20,411][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:11:20,910][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:11:21,412][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:11:21,913][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:11:22,418][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:11:22,918][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:11:23,418][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:11:23,933][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:11:24,433][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:11:24,932][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:11:25,436][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:11:25,934][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:11:26,443][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:11:26,941][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:11:27,441][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:11:27,941][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:11:28,443][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:11:28,944][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:11:29,444][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:11:29,944][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:11:30,444][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:11:30,944][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:11:31,444][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:11:31,961][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:11:32,462][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:11:32,961][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:11:33,460][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:11:33,958][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:11:34,457][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:11:34,959][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:11:35,463][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:11:35,964][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10848 tokens. -[2025-11-09 21:11:36,665][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.28%, ΔTime: 00:00:32 -[2025-11-09 21:11:37,428][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:11:37,430][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:11:37,431][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:11:38,406][__main__][INFO] - Iteration 224 took 50s (29.00% Gen, 69.07% Train). Generation: 14s, Training: 35s. Estimated remaining time: 39h 8m 24s. Estimated total time: 42h 21m 5s. Time estimates for 10 more iterations: 8m 28s, 100 more iterations: 1h 24m 42s, 500 more iterations: 7h 3m 30s. -[2025-11-09 21:11:38,408][__main__][INFO] - Starting iteration 224. -[2025-11-09 21:11:38,883][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 22 and human policies 1. -[2025-11-09 21:11:38,884][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:11:52,753][__main__][INFO] - Number of regex retries in iteration 224: 0 -[2025-11-09 21:11:52,754][__main__][INFO] - agents played in iteration 224 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:11:53,537][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:11:53,565][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:11:53,591][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:11:53,615][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:11:53,616][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:11:53,617][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:11:54,323][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:11:54,782][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:11:55,296][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:11:55,796][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:11:56,297][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:11:56,800][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:11:57,300][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:11:57,800][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:11:58,303][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:11:58,800][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:11:59,300][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:11:59,798][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:12:00,297][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:12:00,798][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:12:01,297][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:12:01,797][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:12:02,295][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:12:02,793][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:12:03,293][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:12:03,791][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:12:04,290][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:12:04,792][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:12:05,291][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:12:05,794][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:12:06,293][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:12:06,790][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:12:07,293][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:12:07,798][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:12:08,303][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:12:08,809][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:12:09,311][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:12:09,827][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:12:10,328][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:12:10,831][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:12:11,338][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:12:11,840][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:12:12,350][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:12:12,851][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:12:13,353][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:12:13,864][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:12:14,366][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:12:14,869][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:12:15,365][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:12:15,863][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:12:16,361][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:12:16,860][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:12:17,357][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:12:17,855][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:12:18,354][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:12:18,857][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:12:19,360][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:12:19,861][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:12:20,361][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:12:20,861][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:12:21,360][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:12:21,862][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:12:22,362][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:12:22,862][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:12:23,362][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:12:23,860][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:12:24,368][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:12:24,867][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:12:25,369][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:12:25,870][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:12:26,370][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10856 tokens. -[2025-11-09 21:12:27,135][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.27%, Current % of VRAM taken: 55.77%, Block Peak % of device VRAM: 62.41%, ΔTime: 00:00:32 -[2025-11-09 21:12:27,873][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:12:27,874][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:12:27,876][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:12:28,901][__main__][INFO] - Iteration 225 took 50s (27.73% Gen, 70.22% Train). Generation: 13s, Training: 35s. Estimated remaining time: 38h 27m 24s. Estimated total time: 41h 40m 55s. Time estimates for 10 more iterations: 8m 20s, 100 more iterations: 1h 23m 21s, 500 more iterations: 6h 56m 49s. -[2025-11-09 21:12:28,904][__main__][INFO] - Starting iteration 225. -[2025-11-09 21:12:29,386][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 22 and human policies 1. -[2025-11-09 21:12:29,387][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:12:45,411][__main__][INFO] - Number of regex retries in iteration 225: 0 -[2025-11-09 21:12:45,412][__main__][INFO] - agents played in iteration 225 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:12:46,231][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:12:46,253][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:12:46,276][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:12:46,298][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:12:46,299][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:12:46,300][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:12:47,084][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:12:47,550][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:12:48,057][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:12:48,556][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:12:49,052][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:12:49,558][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:12:50,056][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:12:50,552][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:12:51,066][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:12:51,568][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:12:52,067][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:12:52,567][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:12:53,067][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:12:53,566][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:12:54,067][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:12:54,570][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:12:55,072][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:12:55,571][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:12:56,070][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:12:56,567][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:12:57,064][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:12:57,567][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:12:58,066][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:12:58,563][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:12:59,068][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:12:59,573][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:13:00,080][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:13:00,582][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:13:01,084][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:13:01,590][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:13:02,093][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:13:02,597][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:13:03,096][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:13:03,595][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:13:04,099][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:13:04,601][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:13:05,103][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:13:05,606][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:13:06,106][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:13:06,609][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:13:07,113][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:13:07,615][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:13:08,123][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:13:08,622][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:13:09,122][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:13:09,621][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:13:10,120][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:13:10,628][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:13:11,126][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:13:11,625][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:13:12,134][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:13:12,631][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:13:13,134][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:13:13,632][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:13:14,129][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:13:14,629][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:13:15,127][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:13:15,626][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:13:16,125][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:13:16,624][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:13:17,133][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:13:17,633][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:13:18,132][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:13:18,641][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:13:19,143][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10846 tokens. -[2025-11-09 21:13:19,915][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 55.71%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:32 -[2025-11-09 21:13:20,677][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:13:20,679][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:13:20,681][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:13:21,728][__main__][INFO] - Iteration 226 took 52s (30.62% Gen, 67.38% Train). Generation: 16s, Training: 35s. Estimated remaining time: 40h 22m 43s. Estimated total time: 43h 37m 7s. Time estimates for 10 more iterations: 8m 43s, 100 more iterations: 1h 27m 14s, 500 more iterations: 7h 16m 11s. -[2025-11-09 21:13:21,731][__main__][INFO] - Starting iteration 226. -[2025-11-09 21:13:22,209][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 22 and human policies 1. -[2025-11-09 21:13:22,209][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:13:38,594][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 21:13:39,372][__main__][INFO] - Number of regex retries in iteration 226: 1 -[2025-11-09 21:13:39,373][__main__][INFO] - agents played in iteration 226 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:13:40,150][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:13:40,173][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:13:40,195][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:13:40,217][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:13:40,217][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:13:40,219][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:13:40,873][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:13:41,330][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:13:41,835][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:13:42,346][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:13:42,841][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:13:43,340][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:13:43,840][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:13:44,342][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:13:44,855][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:13:45,352][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:13:45,849][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:13:46,353][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:13:46,849][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:13:47,360][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:13:47,858][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:13:48,353][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:13:48,857][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:13:49,358][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:13:49,857][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:13:50,358][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:13:50,856][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:13:51,365][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:13:51,864][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:13:52,364][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:13:52,863][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:13:53,366][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:13:53,869][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:13:54,371][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:13:54,874][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:13:55,385][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:13:55,888][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:13:56,389][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:13:56,890][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:13:57,391][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:13:57,895][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:13:58,397][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:13:58,897][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:13:59,399][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:13:59,901][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:14:00,397][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:14:00,895][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:14:01,392][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:14:01,892][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:14:02,390][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:14:02,888][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:14:03,387][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:14:03,886][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:14:04,384][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:14:04,884][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:14:05,382][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:14:05,886][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:14:06,379][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:14:06,878][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:14:07,376][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:14:07,873][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:14:08,374][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:14:08,873][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:14:09,376][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:14:09,884][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:14:10,384][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:14:10,886][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:14:11,388][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:14:11,891][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:14:12,396][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:14:12,905][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10790 tokens. -[2025-11-09 21:14:13,695][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.28%, Current % of VRAM taken: 55.78%, Block Peak % of device VRAM: 62.41%, ΔTime: 00:00:32 -[2025-11-09 21:14:14,430][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:14:14,431][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:14:14,433][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:14:15,419][__main__][INFO] - Iteration 227 took 53s (32.26% Gen, 65.89% Train). Generation: 17s, Training: 35s. Estimated remaining time: 41h 5m 14s. Estimated total time: 44h 20m 33s. Time estimates for 10 more iterations: 8m 52s, 100 more iterations: 1h 28m 41s, 500 more iterations: 7h 23m 25s. -[2025-11-09 21:14:15,421][__main__][INFO] - Starting iteration 227. -[2025-11-09 21:14:15,892][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 22 and human policies 1. -[2025-11-09 21:14:15,893][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:14:20,511][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 21:14:30,803][__main__][INFO] - Number of regex retries in iteration 227: 1 -[2025-11-09 21:14:30,804][__main__][INFO] - agents played in iteration 227 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:14:31,577][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:14:31,602][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:14:31,626][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:14:31,648][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:14:31,649][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:14:31,649][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:14:32,298][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:14:32,756][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:14:33,259][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:14:33,758][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:14:34,257][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:14:34,769][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:14:35,269][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:14:35,768][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:14:36,269][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:14:36,767][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:14:37,272][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:14:37,770][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:14:38,268][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:14:38,766][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:14:39,265][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:14:39,764][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:14:40,268][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:14:40,767][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:14:41,274][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:14:41,772][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:14:42,268][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:14:42,765][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:14:43,262][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:14:43,760][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:14:44,256][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:14:44,756][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:14:45,277][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:14:45,779][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:14:46,283][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:14:46,786][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:14:47,288][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:14:47,794][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:14:48,293][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:14:48,794][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:14:49,299][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:14:49,801][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:14:50,304][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:14:50,805][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:14:51,306][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:14:51,809][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:14:52,309][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:14:52,810][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:14:53,306][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:14:53,807][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:14:54,310][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:14:54,809][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:14:55,311][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:14:55,811][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:14:56,311][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:14:56,811][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:14:57,314][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:14:57,814][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:14:58,314][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:14:58,812][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:14:59,310][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:14:59,810][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:15:00,311][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:15:00,811][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:15:01,310][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:15:01,810][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:15:02,312][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:15:02,812][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:15:03,312][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:15:03,813][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:15:04,315][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10874 tokens. -[2025-11-09 21:15:05,058][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 55.72%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 -[2025-11-09 21:15:05,802][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:15:05,803][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:15:05,805][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:15:07,006][__main__][INFO] - Iteration 228 took 51s (29.17% Gen, 68.48% Train). Generation: 14s, Training: 35s. Estimated remaining time: 39h 19m 32s. Estimated total time: 42h 35m 42s. Time estimates for 10 more iterations: 8m 31s, 100 more iterations: 1h 25m 11s, 500 more iterations: 7h 5m 57s. -[2025-11-09 21:15:07,008][__main__][INFO] - Starting iteration 228. -[2025-11-09 21:15:07,485][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 22 and human policies 1. -[2025-11-09 21:15:07,486][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:15:21,650][__main__][INFO] - Number of regex retries in iteration 228: 0 -[2025-11-09 21:15:21,651][__main__][INFO] - agents played in iteration 228 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:15:22,459][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:15:22,484][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:15:22,509][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:15:22,532][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:15:22,533][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:15:22,534][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:15:23,194][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:15:23,657][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:15:24,161][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:15:24,659][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:15:25,158][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:15:25,656][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:15:26,168][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:15:26,664][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:15:27,167][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:15:27,684][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:15:28,186][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:15:28,688][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:15:29,188][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:15:29,689][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:15:30,198][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:15:30,696][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:15:31,197][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:15:31,694][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:15:32,193][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:15:32,693][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:15:33,191][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:15:33,691][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:15:34,199][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:15:34,698][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:15:35,200][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:15:35,704][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:15:36,211][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:15:36,713][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:15:37,215][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:15:37,714][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:15:38,215][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:15:38,717][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:15:39,227][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:15:39,727][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:15:40,227][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:15:40,730][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:15:41,231][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:15:41,731][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:15:42,235][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:15:42,734][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:15:43,234][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:15:43,739][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:15:44,242][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:15:44,752][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:15:45,252][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:15:45,754][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:15:46,264][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:15:46,765][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:15:47,264][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:15:47,768][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:15:48,270][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:15:48,785][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:15:49,283][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:15:49,782][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:15:50,285][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:15:50,783][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:15:51,290][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:15:51,790][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:15:52,289][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:15:52,797][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:15:53,297][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:15:53,797][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:15:54,297][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:15:54,797][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:15:55,302][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10846 tokens. -[2025-11-09 21:15:56,012][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 55.71%, Block Peak % of device VRAM: 62.41%, ΔTime: 00:00:32 -[2025-11-09 21:15:56,781][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:15:56,783][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:15:56,785][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:15:57,850][__main__][INFO] - Iteration 229 took 50s (28.12% Gen, 69.76% Train). Generation: 14s, Training: 35s. Estimated remaining time: 38h 41m 15s. Estimated total time: 41h 58m 16s. Time estimates for 10 more iterations: 8m 23s, 100 more iterations: 1h 23m 56s, 500 more iterations: 6h 59m 42s. -[2025-11-09 21:15:57,853][__main__][INFO] - Starting iteration 229. -[2025-11-09 21:15:58,316][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 22 and human policies 1. -[2025-11-09 21:15:58,317][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:16:04,011][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 21:16:12,940][__main__][INFO] - Number of regex retries in iteration 229: 1 -[2025-11-09 21:16:12,941][__main__][INFO] - agents played in iteration 229 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:16:13,711][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:16:13,736][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:16:13,761][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:16:13,783][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:16:13,784][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:16:13,785][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:16:14,458][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:16:14,920][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:16:15,428][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:16:15,932][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:16:16,431][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:16:16,930][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:16:17,430][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:16:17,928][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:16:18,430][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:16:18,932][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:16:19,430][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:16:19,930][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:16:20,430][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:16:20,928][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:16:21,429][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:16:21,927][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:16:22,427][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:16:22,925][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:16:23,425][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:16:23,927][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:16:24,427][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:16:24,925][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:16:25,427][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:16:25,927][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:16:26,428][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:16:26,929][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:16:27,430][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:16:27,930][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:16:28,430][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:16:28,931][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:16:29,432][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:16:29,933][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:16:30,431][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:16:30,932][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:16:31,432][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:16:31,939][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:16:32,440][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:16:32,940][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:16:33,443][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:16:33,944][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:16:34,442][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:16:34,944][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:16:35,444][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:16:35,949][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:16:36,449][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:16:36,947][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:16:37,453][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:16:37,953][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:16:38,449][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:16:38,947][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:16:39,445][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:16:39,940][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:16:40,438][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:16:40,937][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:16:41,436][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:16:41,935][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:16:42,434][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:16:42,932][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:16:43,429][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:16:43,927][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:16:44,429][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:16:44,930][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:16:45,429][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:16:45,933][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:16:46,435][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10843 tokens. -[2025-11-09 21:16:47,137][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 55.69%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 -[2025-11-09 21:16:47,916][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:16:47,918][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:16:47,920][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:16:48,943][__main__][INFO] - Iteration 230 took 50s (28.89% Gen, 69.09% Train). Generation: 14s, Training: 34s. Estimated remaining time: 38h 53m 28s. Estimated total time: 42h 11m 20s. Time estimates for 10 more iterations: 8m 26s, 100 more iterations: 1h 24m 22s, 500 more iterations: 7h 1m 53s. -[2025-11-09 21:16:48,945][__main__][INFO] - Starting iteration 230. -[2025-11-09 21:16:49,407][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 22 and human policies 1. -[2025-11-09 21:16:49,408][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:17:04,237][__main__][INFO] - Number of regex retries in iteration 230: 0 -[2025-11-09 21:17:04,237][__main__][INFO] - agents played in iteration 230 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:17:05,008][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:17:05,035][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:17:05,062][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:17:05,085][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:17:05,085][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:17:05,086][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:17:05,765][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:17:06,224][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:17:06,731][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:17:07,235][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:17:07,736][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:17:08,238][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:17:08,737][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:17:09,234][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:17:09,733][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:17:10,231][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:17:10,727][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:17:11,224][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:17:11,722][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:17:12,218][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:17:12,720][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:17:13,217][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:17:13,715][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:17:14,211][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:17:14,707][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:17:15,205][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:17:15,701][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:17:16,200][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:17:16,704][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:17:17,201][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:17:17,696][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:17:18,197][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:17:18,694][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:17:19,192][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:17:19,693][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:17:20,195][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:17:20,711][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:17:21,212][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:17:21,712][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:17:22,219][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:17:22,718][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:17:23,218][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:17:23,718][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:17:24,217][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:17:24,724][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:17:25,226][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:17:25,729][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:17:26,231][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:17:26,738][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:17:27,240][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:17:27,740][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:17:28,242][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:17:28,747][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:17:29,253][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:17:29,755][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:17:30,258][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:17:30,759][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:17:31,263][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:17:31,765][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:17:32,265][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:17:32,770][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:17:33,272][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:17:33,775][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:17:34,276][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:17:34,773][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:17:35,275][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:17:35,777][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:17:36,276][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:17:36,793][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:17:37,292][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:17:37,795][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10842 tokens. -[2025-11-09 21:17:38,523][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 55.69%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 -[2025-11-09 21:17:39,290][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:17:39,292][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:17:39,294][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:17:41,304][__main__][INFO] - Iteration 231 took 51s (28.58% Gen, 67.55% Train). Generation: 14s, Training: 35s. Estimated remaining time: 39h 56m 8s. Estimated total time: 43h 14m 53s. Time estimates for 10 more iterations: 8m 38s, 100 more iterations: 1h 26m 29s, 500 more iterations: 7h 12m 28s. -[2025-11-09 21:17:41,306][__main__][INFO] - Starting iteration 231. -[2025-11-09 21:17:41,772][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 23 and human policies 1. -[2025-11-09 21:17:41,773][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:17:47,315][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 21:17:58,812][__main__][INFO] - Number of regex retries in iteration 231: 1 -[2025-11-09 21:17:58,812][__main__][INFO] - agents played in iteration 231 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:17:59,588][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:17:59,618][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:17:59,645][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:17:59,669][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:17:59,670][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:17:59,670][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:18:00,307][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:18:00,762][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:18:01,265][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:18:01,764][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:18:02,262][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:18:02,763][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:18:03,263][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:18:03,760][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:18:04,258][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:18:04,760][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:18:05,256][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:18:05,758][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:18:06,255][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:18:06,753][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:18:07,252][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:18:07,751][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:18:08,251][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:18:08,750][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:18:09,249][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:18:09,746][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:18:10,246][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:18:10,748][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:18:11,249][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:18:11,751][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:18:12,253][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:18:12,756][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:18:13,257][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:18:13,761][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:18:14,264][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:18:14,764][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:18:15,264][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:18:15,765][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:18:16,269][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:18:16,768][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:18:17,269][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:18:17,773][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:18:18,275][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:18:18,779][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:18:19,280][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:18:19,784][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:18:20,296][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:18:20,800][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:18:21,304][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:18:21,806][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:18:22,308][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:18:22,818][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:18:23,320][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:18:23,821][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:18:24,327][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:18:24,827][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:18:25,325][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:18:25,823][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:18:26,322][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:18:26,828][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:18:27,329][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:18:27,830][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:18:28,325][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:18:28,824][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:18:29,325][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:18:29,826][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:18:30,329][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:18:30,830][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:18:31,331][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:18:31,833][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:18:32,334][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10835 tokens. -[2025-11-09 21:18:33,040][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 55.72%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 -[2025-11-09 21:18:33,800][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:18:33,802][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:18:33,804][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:18:34,787][__main__][INFO] - Iteration 232 took 53s (32.14% Gen, 66.00% Train). Generation: 17s, Training: 34s. Estimated remaining time: 40h 51m 8s. Estimated total time: 44h 10m 46s. Time estimates for 10 more iterations: 8m 50s, 100 more iterations: 1h 28m 21s, 500 more iterations: 7h 21m 47s. -[2025-11-09 21:18:34,790][__main__][INFO] - Starting iteration 232. -[2025-11-09 21:18:35,258][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 23 and human policies 1. -[2025-11-09 21:18:35,259][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:18:49,516][__main__][INFO] - Number of regex retries in iteration 232: 0 -[2025-11-09 21:18:49,517][__main__][INFO] - agents played in iteration 232 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:18:50,288][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:18:50,316][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:18:50,342][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:18:50,365][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:18:50,365][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:18:50,366][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:18:50,990][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:18:51,445][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:18:51,947][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:18:52,446][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:18:52,945][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:18:53,443][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:18:53,943][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:18:54,443][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:18:54,942][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:18:55,440][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:18:55,936][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:18:56,435][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:18:56,934][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:18:57,431][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:18:57,929][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:18:58,426][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:18:58,925][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:18:59,432][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:18:59,944][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:19:00,452][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:19:00,957][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:19:01,464][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:19:01,968][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:19:02,466][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:19:02,966][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:19:03,470][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:19:03,973][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:19:04,473][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:19:04,973][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:19:05,474][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:19:05,972][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:19:06,473][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:19:06,977][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:19:07,478][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:19:07,987][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:19:08,488][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:19:08,994][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:19:09,494][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:19:09,999][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:19:10,499][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:19:11,001][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:19:11,505][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:19:12,005][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:19:12,505][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:19:13,015][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:19:13,514][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:19:14,015][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:19:14,516][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:19:15,018][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:19:15,522][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:19:16,027][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:19:16,530][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:19:17,040][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:19:17,543][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:19:18,044][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:19:18,544][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:19:19,044][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:19:19,547][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:19:20,047][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:19:20,548][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:19:21,052][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:19:21,553][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:19:22,055][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:19:22,557][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:19:23,059][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10820 tokens. -[2025-11-09 21:19:23,796][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 -[2025-11-09 21:19:24,561][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:19:24,562][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:19:24,564][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:19:25,544][__main__][INFO] - Iteration 233 took 50s (28.35% Gen, 69.69% Train). Generation: 14s, Training: 35s. Estimated remaining time: 38h 33m 49s. Estimated total time: 41h 54m 18s. Time estimates for 10 more iterations: 8m 22s, 100 more iterations: 1h 23m 48s, 500 more iterations: 6h 59m 3s. -[2025-11-09 21:19:25,546][__main__][INFO] - Starting iteration 233. -[2025-11-09 21:19:26,010][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 23 and human policies 1. -[2025-11-09 21:19:26,010][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:19:42,760][__main__][INFO] - Number of regex retries in iteration 233: 0 -[2025-11-09 21:19:42,761][__main__][INFO] - agents played in iteration 233 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:19:43,532][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:19:43,560][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:19:43,586][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:19:43,609][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:19:43,609][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:19:43,610][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:19:44,237][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:19:44,703][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:19:45,209][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:19:45,705][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:19:46,204][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:19:46,704][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:19:47,203][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:19:47,703][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:19:48,202][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:19:48,704][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:19:49,202][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:19:49,701][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:19:50,202][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:19:50,700][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:19:51,202][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:19:51,701][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:19:52,201][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:19:52,699][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:19:53,197][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:19:53,696][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:19:54,193][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:19:54,692][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:19:55,193][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:19:55,694][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:19:56,196][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:19:56,699][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:19:57,201][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:19:57,703][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:19:58,205][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:19:58,707][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:19:59,209][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:19:59,709][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:20:00,223][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:20:00,721][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:20:01,219][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:20:01,714][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:20:02,212][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:20:02,723][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:20:03,221][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:20:03,716][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:20:04,214][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:20:04,719][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:20:05,220][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:20:05,725][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:20:06,228][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:20:06,745][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:20:07,244][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:20:07,747][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:20:08,251][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:20:08,753][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:20:09,257][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:20:09,759][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:20:10,261][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:20:10,767][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:20:11,269][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:20:11,771][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:20:12,273][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:20:12,773][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:20:13,278][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:20:13,779][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:20:14,286][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:20:14,789][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:20:15,292][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:20:15,796][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:20:16,296][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10830 tokens. -[2025-11-09 21:20:17,048][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.25%, Current % of VRAM taken: 55.75%, Block Peak % of device VRAM: 62.35%, ΔTime: 00:00:32 -[2025-11-09 21:20:17,772][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:20:17,774][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:20:17,776][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:20:18,761][__main__][INFO] - Iteration 234 took 52s (31.75% Gen, 66.38% Train). Generation: 16s, Training: 35s. Estimated remaining time: 40h 36m 14s. Estimated total time: 43h 57m 36s. Time estimates for 10 more iterations: 8m 47s, 100 more iterations: 1h 27m 55s, 500 more iterations: 7h 19m 36s. -[2025-11-09 21:20:18,763][__main__][INFO] - Starting iteration 234. -[2025-11-09 21:20:19,229][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 23 and human policies 1. -[2025-11-09 21:20:19,229][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:20:23,388][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 21:20:25,349][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 21:20:34,879][__main__][INFO] - Number of regex retries in iteration 234: 2 -[2025-11-09 21:20:34,880][__main__][INFO] - agents played in iteration 234 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:20:35,657][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:20:35,684][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:20:35,710][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:20:35,733][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:20:35,733][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:20:35,734][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:20:36,371][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:20:36,842][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:20:37,347][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:20:37,846][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:20:38,347][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:20:38,845][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:20:39,355][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:20:39,854][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:20:40,353][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:20:40,852][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:20:41,349][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:20:41,846][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:20:42,346][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:20:42,843][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:20:43,356][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:20:43,852][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:20:44,351][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:20:44,857][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:20:45,359][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:20:45,868][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:20:46,375][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:20:46,877][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:20:47,381][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:20:47,885][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:20:48,393][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:20:48,898][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:20:49,400][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:20:49,906][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:20:50,411][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:20:50,911][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:20:51,412][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:20:51,912][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:20:52,413][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:20:52,912][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:20:53,412][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:20:53,912][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:20:54,411][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:20:54,911][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:20:55,411][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:20:55,910][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:20:56,411][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:20:56,908][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:20:57,405][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:20:57,906][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:20:58,405][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:20:58,905][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:20:59,405][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:20:59,905][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:21:00,408][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:21:00,910][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:21:01,411][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:21:01,913][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:21:02,417][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:21:02,921][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:21:03,422][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:21:03,922][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:21:04,423][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:21:04,925][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:21:05,440][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:21:05,941][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:21:06,443][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:21:06,945][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:21:07,444][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:21:07,945][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:21:08,446][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10844 tokens. -[2025-11-09 21:21:09,158][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 55.71%, Block Peak % of device VRAM: 62.41%, ΔTime: 00:00:32 -[2025-11-09 21:21:09,907][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:21:09,909][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:21:09,912][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:21:10,905][__main__][INFO] - Iteration 235 took 51s (30.29% Gen, 67.79% Train). Generation: 15s, Training: 35s. Estimated remaining time: 39h 41m 37s. Estimated total time: 43h 3m 51s. Time estimates for 10 more iterations: 8m 36s, 100 more iterations: 1h 26m 7s, 500 more iterations: 7h 10m 38s. -[2025-11-09 21:21:10,907][__main__][INFO] - Starting iteration 235. -[2025-11-09 21:21:11,370][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 23 and human policies 1. -[2025-11-09 21:21:11,371][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:21:20,709][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 21:21:27,775][__main__][INFO] - Number of regex retries in iteration 235: 1 -[2025-11-09 21:21:27,775][__main__][INFO] - agents played in iteration 235 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:21:28,548][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:21:28,574][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:21:28,598][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:21:28,620][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:21:28,621][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:21:28,622][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:21:29,273][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:21:30,062][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:21:30,567][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:21:31,067][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:21:31,563][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:21:32,062][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:21:32,561][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:21:33,060][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:21:33,558][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:21:34,059][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:21:34,556][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:21:35,056][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:21:35,557][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:21:36,065][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:21:36,567][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:21:37,071][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:21:37,574][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:21:38,077][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:21:38,574][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:21:39,072][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:21:39,579][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:21:40,081][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:21:40,585][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:21:41,086][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:21:41,590][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:21:42,098][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:21:42,600][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:21:43,097][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:21:43,599][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:21:44,099][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:21:44,599][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:21:45,099][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:21:45,598][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:21:46,100][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:21:46,601][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:21:47,100][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:21:47,606][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:21:48,111][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:21:48,611][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:21:49,114][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:21:49,614][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:21:50,115][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:21:50,616][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:21:51,117][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:21:51,617][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:21:52,122][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:21:52,625][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:21:53,127][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:21:53,626][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:21:54,140][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:21:54,642][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:21:55,144][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:21:55,646][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:21:56,149][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:21:56,664][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:21:57,165][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:21:57,668][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:21:58,169][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:21:58,668][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:21:59,173][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:21:59,674][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:22:00,175][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:22:00,676][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:22:01,179][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:22:01,684][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10805 tokens. -[2025-11-09 21:22:02,399][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 55.68%, Block Peak % of device VRAM: 62.53%, ΔTime: 00:00:33 -[2025-11-09 21:22:03,114][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:22:03,115][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:22:03,118][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:22:04,113][__main__][INFO] - Iteration 236 took 52s (31.10% Gen, 67.01% Train). Generation: 16s, Training: 35s. Estimated remaining time: 40h 34m 3s. Estimated total time: 43h 57m 10s. Time estimates for 10 more iterations: 8m 47s, 100 more iterations: 1h 27m 54s, 500 more iterations: 7h 19m 31s. -[2025-11-09 21:22:04,115][__main__][INFO] - Starting iteration 236. -[2025-11-09 21:22:04,585][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 23 and human policies 1. -[2025-11-09 21:22:04,585][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:22:11,487][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 21:22:18,654][__main__][INFO] - Number of regex retries in iteration 236: 1 -[2025-11-09 21:22:18,655][__main__][INFO] - agents played in iteration 236 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:22:19,425][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:22:19,449][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:22:19,473][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:22:19,495][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:22:19,495][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:22:19,496][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:22:20,127][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:22:20,583][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:22:21,089][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:22:21,589][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:22:22,089][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:22:22,588][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:22:23,094][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:22:23,595][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:22:24,094][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:22:24,594][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:22:25,092][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:22:25,594][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:22:26,091][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:22:26,588][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:22:27,088][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:22:27,588][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:22:28,087][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:22:28,589][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:22:29,091][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:22:29,597][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:22:30,102][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:22:30,604][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:22:31,105][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:22:31,607][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:22:32,110][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:22:32,611][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:22:33,110][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:22:33,613][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:22:34,113][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:22:34,612][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:22:35,113][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:22:35,611][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:22:36,109][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:22:36,607][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:22:37,105][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:22:37,605][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:22:38,105][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:22:38,605][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:22:39,102][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:22:39,600][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:22:40,101][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:22:40,600][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:22:41,099][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:22:41,595][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:22:42,094][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:22:42,595][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:22:43,095][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:22:43,595][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:22:44,101][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:22:44,599][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:22:45,101][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:22:45,603][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:22:46,103][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:22:46,609][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:22:47,114][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:22:47,614][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:22:48,118][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:22:48,620][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:22:49,125][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:22:49,625][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:22:50,128][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:22:50,631][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:22:51,131][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:22:51,633][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:22:52,134][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10833 tokens. -[2025-11-09 21:22:52,859][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.26%, Current % of VRAM taken: 55.75%, Block Peak % of device VRAM: 62.42%, ΔTime: 00:00:32 -[2025-11-09 21:22:53,597][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:22:53,599][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:22:53,602][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:22:54,586][__main__][INFO] - Iteration 237 took 50s (28.14% Gen, 69.89% Train). Generation: 14s, Training: 34s. Estimated remaining time: 38h 16m 7s. Estimated total time: 41h 40m 4s. Time estimates for 10 more iterations: 8m 20s, 100 more iterations: 1h 23m 20s, 500 more iterations: 6h 56m 40s. -[2025-11-09 21:22:54,589][__main__][INFO] - Starting iteration 237. -[2025-11-09 21:22:55,065][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 23 and human policies 1. -[2025-11-09 21:22:55,065][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:23:01,448][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 21:23:10,946][__main__][INFO] - Number of regex retries in iteration 237: 1 -[2025-11-09 21:23:10,946][__main__][INFO] - agents played in iteration 237 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:23:11,722][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:23:11,745][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:23:11,767][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:23:11,789][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:23:11,790][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:23:11,790][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:23:12,436][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:23:12,900][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:23:13,400][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:23:13,898][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:23:14,397][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:23:14,897][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:23:15,401][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:23:15,899][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:23:16,397][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:23:16,898][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:23:17,395][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:23:17,900][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:23:18,398][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:23:18,899][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:23:19,406][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:23:19,908][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:23:20,410][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:23:20,909][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:23:21,411][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:23:21,913][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:23:22,413][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:23:22,914][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:23:23,420][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:23:23,919][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:23:24,419][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:23:24,919][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:23:25,420][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:23:25,925][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:23:26,423][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:23:26,926][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:23:27,426][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:23:27,925][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:23:28,425][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:23:28,923][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:23:29,422][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:23:29,925][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:23:30,423][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:23:30,923][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:23:31,422][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:23:31,919][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:23:32,417][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:23:32,919][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:23:33,418][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:23:33,916][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:23:34,417][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:23:34,916][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:23:35,415][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:23:35,915][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:23:36,415][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:23:36,918][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:23:37,416][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:23:37,917][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:23:38,414][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:23:38,918][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:23:39,416][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:23:39,918][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:23:40,415][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:23:40,919][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:23:41,420][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:23:41,926][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:23:42,428][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:23:42,929][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:23:43,436][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:23:43,938][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:23:44,439][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10808 tokens. -[2025-11-09 21:23:45,176][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 55.72%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 -[2025-11-09 21:23:45,913][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:23:45,915][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:23:45,916][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:23:46,879][__main__][INFO] - Iteration 238 took 51s (30.65% Gen, 67.49% Train). Generation: 15s, Training: 34s. Estimated remaining time: 39h 45m 54s. Estimated total time: 43h 10m 44s. Time estimates for 10 more iterations: 8m 38s, 100 more iterations: 1h 26m 21s, 500 more iterations: 7h 11m 47s. -[2025-11-09 21:23:46,881][__main__][INFO] - Starting iteration 238. -[2025-11-09 21:23:47,354][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 23 and human policies 1. -[2025-11-09 21:23:47,355][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:24:03,622][__main__][INFO] - Number of regex retries in iteration 238: 0 -[2025-11-09 21:24:03,623][__main__][INFO] - agents played in iteration 238 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:24:04,401][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:24:04,424][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:24:04,446][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:24:04,468][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:24:04,469][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:24:04,469][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:24:05,093][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:24:05,547][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:24:06,054][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:24:06,552][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:24:07,049][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:24:07,548][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:24:08,045][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:24:08,547][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:24:09,048][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:24:09,548][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:24:10,049][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:24:10,548][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:24:11,046][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:24:11,547][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:24:12,050][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:24:12,552][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:24:13,052][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:24:13,550][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:24:14,048][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:24:14,550][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:24:15,057][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:24:15,558][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:24:16,061][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:24:16,563][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:24:17,068][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:24:17,569][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:24:18,069][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:24:18,568][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:24:19,079][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:24:19,584][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:24:20,084][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:24:20,584][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:24:21,084][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:24:21,595][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:24:22,093][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:24:22,591][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:24:23,096][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:24:23,595][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:24:24,105][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:24:24,603][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:24:25,103][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:24:25,604][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:24:26,105][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:24:26,605][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:24:27,107][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:24:27,607][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:24:28,108][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:24:28,608][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:24:29,109][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:24:29,617][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:24:30,123][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:24:30,627][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:24:31,129][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:24:31,632][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:24:32,136][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:24:32,638][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:24:33,141][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:24:33,646][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:24:34,151][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:24:34,653][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:24:35,154][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:24:35,656][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:24:36,159][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:24:36,663][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:24:37,165][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10822 tokens. -[2025-11-09 21:24:37,904][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.26%, Current % of VRAM taken: 55.76%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 -[2025-11-09 21:24:38,669][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:24:38,670][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:24:38,672][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:24:39,717][__main__][INFO] - Iteration 239 took 52s (31.07% Gen, 66.93% Train). Generation: 16s, Training: 35s. Estimated remaining time: 40h 12m 29s. Estimated total time: 43h 38m 12s. Time estimates for 10 more iterations: 8m 43s, 100 more iterations: 1h 27m 16s, 500 more iterations: 7h 16m 22s. -[2025-11-09 21:24:39,720][__main__][INFO] - Starting iteration 239. -[2025-11-09 21:24:40,182][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 23 and human policies 1. -[2025-11-09 21:24:40,183][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:24:55,125][__main__][INFO] - Number of regex retries in iteration 239: 0 -[2025-11-09 21:24:55,126][__main__][INFO] - agents played in iteration 239 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:24:55,895][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:24:55,918][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:24:55,940][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:24:55,962][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:24:55,963][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:24:55,964][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:24:56,593][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:24:57,055][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:24:57,561][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:24:58,063][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:24:58,564][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:24:59,068][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:24:59,572][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:25:00,073][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:25:00,571][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:25:01,069][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:25:01,568][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:25:02,063][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:25:02,564][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:25:03,064][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:25:03,567][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:25:04,067][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:25:04,570][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:25:05,069][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:25:05,572][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:25:06,079][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:25:06,582][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:25:07,083][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:25:07,589][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:25:08,087][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:25:08,589][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:25:09,101][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:25:09,605][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:25:10,118][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:25:10,621][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:25:11,123][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:25:11,638][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:25:12,139][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:25:12,644][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:25:13,146][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:25:13,647][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:25:14,148][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:25:14,648][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:25:15,147][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:25:15,650][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:25:16,152][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:25:16,649][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:25:17,148][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:25:17,650][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:25:18,153][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:25:18,648][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:25:19,150][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:25:19,650][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:25:20,148][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:25:20,648][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:25:21,146][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:25:21,645][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:25:22,149][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:25:22,648][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:25:23,146][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:25:23,645][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:25:24,147][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:25:24,650][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:25:25,150][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:25:25,650][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:25:26,149][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:25:26,651][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:25:27,150][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:25:27,652][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:25:28,150][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:25:28,654][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10776 tokens. -[2025-11-09 21:25:29,382][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 55.72%, Block Peak % of device VRAM: 62.53%, ΔTime: 00:00:32 -[2025-11-09 21:25:30,122][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:25:30,123][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:25:30,125][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:25:31,081][__main__][INFO] - Iteration 240 took 50s (29.36% Gen, 68.76% Train). Generation: 14s, Training: 34s. Estimated remaining time: 38h 58m 22s. Estimated total time: 42h 24m 56s. Time estimates for 10 more iterations: 8m 28s, 100 more iterations: 1h 24m 49s, 500 more iterations: 7h 4m 9s. -[2025-11-09 21:25:31,083][__main__][INFO] - Starting iteration 240. -[2025-11-09 21:25:31,544][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 23 and human policies 1. -[2025-11-09 21:25:31,544][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:25:35,292][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 21:25:46,574][__main__][INFO] - Number of regex retries in iteration 240: 1 -[2025-11-09 21:25:46,574][__main__][INFO] - agents played in iteration 240 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:25:47,342][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:25:47,367][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:25:47,391][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:25:47,415][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:25:47,415][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:25:47,416][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:25:48,056][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:25:48,511][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:25:49,016][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:25:49,531][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:25:50,031][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:25:50,530][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:25:51,033][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:25:51,531][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:25:52,041][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:25:52,540][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:25:53,038][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:25:53,542][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:25:54,041][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:25:54,538][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:25:55,039][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:25:55,538][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:25:56,051][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:25:56,556][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:25:57,055][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:25:57,560][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:25:58,062][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:25:58,566][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:25:59,071][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:25:59,574][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:26:00,079][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:26:00,581][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:26:01,081][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:26:01,584][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:26:02,086][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:26:02,587][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:26:03,088][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:26:03,589][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:26:04,086][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:26:04,587][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:26:05,087][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:26:05,587][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:26:06,086][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:26:06,587][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:26:07,087][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:26:07,587][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:26:08,089][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:26:08,588][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:26:09,085][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:26:09,585][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:26:10,082][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:26:10,583][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:26:11,084][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:26:11,585][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:26:12,086][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:26:12,588][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:26:13,089][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:26:13,587][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:26:14,091][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:26:14,594][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:26:15,099][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:26:15,605][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:26:16,108][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:26:16,607][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:26:17,134][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:26:17,639][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:26:18,142][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:26:18,644][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:26:19,147][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:26:19,650][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:26:20,155][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10770 tokens. -[2025-11-09 21:26:20,909][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.25%, Current % of VRAM taken: 55.75%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 -[2025-11-09 21:26:21,636][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:26:21,639][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:26:21,640][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:26:23,444][__main__][INFO] - Iteration 241 took 51s (28.96% Gen, 67.56% Train). Generation: 15s, Training: 35s. Estimated remaining time: 39h 47m 37s. Estimated total time: 43h 15m 4s. Time estimates for 10 more iterations: 8m 39s, 100 more iterations: 1h 26m 30s, 500 more iterations: 7h 12m 30s. -[2025-11-09 21:26:23,447][__main__][INFO] - Starting iteration 241. -[2025-11-09 21:26:23,918][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 24 and human policies 1. -[2025-11-09 21:26:23,918][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:26:32,176][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 21:26:41,141][__main__][INFO] - Number of regex retries in iteration 241: 1 -[2025-11-09 21:26:41,142][__main__][INFO] - agents played in iteration 241 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:26:41,913][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:26:41,941][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:26:41,966][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:26:41,989][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:26:41,990][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:26:41,991][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:26:42,634][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:26:43,088][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:26:43,591][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:26:44,091][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:26:44,591][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:26:45,089][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:26:45,586][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:26:46,087][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:26:46,583][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:26:47,082][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:26:47,583][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:26:48,084][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:26:48,581][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:26:49,081][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:26:49,581][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:26:50,085][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:26:50,585][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:26:51,084][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:26:51,585][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:26:52,087][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:26:52,588][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:26:53,090][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:26:53,591][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:26:54,095][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:26:54,596][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:26:55,100][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:26:55,599][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:26:56,098][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:26:56,601][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:26:57,099][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:26:57,598][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:26:58,098][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:26:58,597][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:26:59,095][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:26:59,594][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:27:00,091][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:27:00,590][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:27:01,086][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:27:01,584][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:27:02,079][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:27:02,577][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:27:03,077][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:27:03,574][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:27:04,072][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:27:04,570][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:27:05,069][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:27:05,567][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:27:06,067][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:27:06,569][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:27:07,069][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:27:07,568][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:27:08,069][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:27:08,571][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:27:09,067][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:27:09,566][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:27:10,066][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:27:10,568][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:27:11,073][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:27:11,574][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:27:12,076][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:27:12,580][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:27:13,080][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:27:13,582][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:27:14,085][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:27:14,587][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10789 tokens. -[2025-11-09 21:27:15,338][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.25%, Current % of VRAM taken: 55.75%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 -[2025-11-09 21:27:16,077][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:27:16,079][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:27:16,081][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:27:17,015][__main__][INFO] - Iteration 242 took 53s (32.44% Gen, 65.80% Train). Generation: 17s, Training: 34s. Estimated remaining time: 40h 46m 32s. Estimated total time: 44h 14m 52s. Time estimates for 10 more iterations: 8m 50s, 100 more iterations: 1h 28m 29s, 500 more iterations: 7h 22m 28s. -[2025-11-09 21:27:17,017][__main__][INFO] - Starting iteration 242. -[2025-11-09 21:27:17,495][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 24 and human policies 1. -[2025-11-09 21:27:17,496][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:27:22,834][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 21:27:24,853][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 21:27:33,244][__main__][INFO] - Number of regex retries in iteration 242: 2 -[2025-11-09 21:27:33,244][__main__][INFO] - agents played in iteration 242 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:27:34,022][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:27:34,047][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:27:34,072][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:27:34,094][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:27:34,094][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:27:34,095][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:27:34,734][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:27:35,189][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:27:35,691][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:27:36,195][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:27:36,692][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:27:37,190][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:27:37,688][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:27:38,188][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:27:38,687][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:27:39,189][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:27:39,687][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:27:40,189][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:27:40,690][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:27:41,194][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:27:41,694][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:27:42,199][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:27:42,701][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:27:43,203][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:27:43,702][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:27:44,204][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:27:44,705][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:27:45,206][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:27:45,707][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:27:46,211][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:27:46,713][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:27:47,213][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:27:47,715][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:27:48,217][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:27:48,716][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:27:49,217][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:27:49,716][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:27:50,218][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:27:50,719][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:27:51,219][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:27:51,718][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:27:52,228][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:27:52,727][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:27:53,227][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:27:53,726][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:27:54,227][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:27:54,726][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:27:55,226][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:27:55,724][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:27:56,222][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:27:56,724][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:27:57,244][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:27:57,742][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:27:58,236][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:27:58,738][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:27:59,236][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:27:59,744][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:28:00,246][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:28:00,745][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:28:01,246][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:28:01,743][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:28:02,241][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:28:02,743][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:28:03,240][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:28:03,754][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:28:04,253][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:28:04,752][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:28:05,252][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:28:05,753][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:28:06,265][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:28:06,770][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10799 tokens. -[2025-11-09 21:28:07,522][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 55.66%, Block Peak % of device VRAM: 62.42%, ΔTime: 00:00:32 -[2025-11-09 21:28:08,249][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:28:08,251][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:28:08,252][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:28:09,288][__main__][INFO] - Iteration 243 took 51s (30.41% Gen, 67.59% Train). Generation: 15s, Training: 35s. Estimated remaining time: 39h 40m 27s. Estimated total time: 43h 9m 39s. Time estimates for 10 more iterations: 8m 37s, 100 more iterations: 1h 26m 19s, 500 more iterations: 7h 11m 36s. -[2025-11-09 21:28:09,290][__main__][INFO] - Starting iteration 243. -[2025-11-09 21:28:09,759][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 24 and human policies 1. -[2025-11-09 21:28:09,760][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:28:26,775][__main__][INFO] - Number of regex retries in iteration 243: 0 -[2025-11-09 21:28:26,775][__main__][INFO] - agents played in iteration 243 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:28:27,594][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:28:27,621][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:28:27,648][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:28:27,670][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:28:27,671][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:28:27,672][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:28:28,309][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:28:28,762][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:28:29,265][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:28:29,761][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:28:30,263][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:28:30,757][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:28:31,255][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:28:31,752][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:28:32,247][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:28:32,760][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:28:33,262][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:28:33,762][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:28:34,272][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:28:34,771][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:28:35,273][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:28:35,772][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:28:36,272][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:28:36,778][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:28:37,278][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:28:37,781][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:28:38,286][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:28:38,787][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:28:39,292][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:28:39,797][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:28:40,299][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:28:40,797][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:28:41,296][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:28:41,794][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:28:42,296][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:28:42,794][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:28:43,295][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:28:43,790][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:28:44,289][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:28:44,790][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:28:45,289][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:28:45,789][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:28:46,290][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:28:46,789][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:28:47,286][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:28:47,788][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:28:48,287][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:28:48,787][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:28:49,286][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:28:49,784][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:28:50,286][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:28:50,788][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:28:51,287][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:28:51,788][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:28:52,292][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:28:52,791][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:28:53,292][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:28:53,794][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:28:54,294][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:28:54,792][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:28:55,289][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:28:55,788][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:28:56,286][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:28:56,789][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:28:57,291][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:28:57,793][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:28:58,298][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:28:58,800][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:28:59,297][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:28:59,801][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:29:00,303][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10788 tokens. -[2025-11-09 21:29:01,065][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 55.69%, Block Peak % of device VRAM: 62.38%, ΔTime: 00:00:32 -[2025-11-09 21:29:01,790][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:29:01,792][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:29:01,796][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:29:02,753][__main__][INFO] - Iteration 244 took 52s (32.11% Gen, 66.08% Train). Generation: 17s, Training: 35s. Estimated remaining time: 40h 39m 37s. Estimated total time: 44h 9m 43s. Time estimates for 10 more iterations: 8m 49s, 100 more iterations: 1h 28m 19s, 500 more iterations: 7h 21m 37s. -[2025-11-09 21:29:02,755][__main__][INFO] - Starting iteration 244. -[2025-11-09 21:29:03,231][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 24 and human policies 1. -[2025-11-09 21:29:03,232][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:29:08,713][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 21:29:19,035][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 21:29:20,069][__main__][INFO] - Number of regex retries in iteration 244: 2 -[2025-11-09 21:29:20,070][__main__][INFO] - agents played in iteration 244 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:29:20,855][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:29:20,883][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:29:20,909][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:29:20,932][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:29:20,933][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:29:20,934][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:29:21,579][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:29:22,033][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:29:22,538][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:29:23,045][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:29:23,541][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:29:24,037][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:29:24,547][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:29:25,041][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:29:25,548][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:29:26,047][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:29:26,550][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:29:27,052][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:29:27,552][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:29:28,054][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:29:28,562][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:29:29,063][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:29:29,562][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:29:30,063][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:29:30,560][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:29:31,074][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:29:31,575][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:29:32,075][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:29:32,578][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:29:33,075][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:29:33,580][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:29:34,084][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:29:34,601][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:29:35,100][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:29:35,598][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:29:36,099][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:29:36,598][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:29:37,097][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:29:37,594][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:29:38,094][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:29:38,595][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:29:39,094][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:29:39,594][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:29:40,094][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:29:40,592][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:29:41,092][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:29:41,595][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:29:42,097][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:29:42,596][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:29:43,095][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:29:43,594][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:29:44,097][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:29:44,600][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:29:45,103][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:29:45,604][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:29:46,102][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:29:46,600][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:29:47,099][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:29:47,596][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:29:48,095][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:29:48,596][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:29:49,099][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:29:49,602][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:29:50,102][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:29:50,604][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:29:51,106][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:29:51,611][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:29:52,115][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:29:52,616][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:29:53,118][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:29:53,627][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10723 tokens. -[2025-11-09 21:29:54,350][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.26%, Current % of VRAM taken: 55.75%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:32 -[2025-11-09 21:29:55,090][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:29:55,092][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:29:55,094][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:29:56,045][__main__][INFO] - Iteration 245 took 52s (31.88% Gen, 66.32% Train). Generation: 16s, Training: 35s. Estimated remaining time: 40h 29m 42s. Estimated total time: 44h 0m 41s. Time estimates for 10 more iterations: 8m 48s, 100 more iterations: 1h 28m 1s, 500 more iterations: 7h 20m 6s. -[2025-11-09 21:29:56,047][__main__][INFO] - Starting iteration 245. -[2025-11-09 21:29:56,523][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 24 and human policies 1. -[2025-11-09 21:29:56,523][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:30:13,062][__main__][INFO] - Number of regex retries in iteration 245: 0 -[2025-11-09 21:30:13,063][__main__][INFO] - agents played in iteration 245 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:30:13,891][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:30:13,919][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:30:13,945][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:30:13,968][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:30:13,968][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:30:13,969][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:30:14,610][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:30:15,068][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:30:15,570][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:30:16,068][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:30:16,567][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:30:17,079][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:30:17,582][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:30:18,084][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:30:18,586][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:30:19,094][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:30:19,600][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:30:20,099][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:30:20,602][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:30:21,103][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:30:21,604][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:30:22,107][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:30:22,609][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:30:23,108][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:30:23,613][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:30:24,112][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:30:24,611][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:30:25,111][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:30:25,608][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:30:26,104][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:30:26,603][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:30:27,103][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:30:27,602][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:30:28,101][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:30:28,601][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:30:29,101][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:30:29,601][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:30:30,101][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:30:30,601][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:30:31,100][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:30:31,602][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:30:32,100][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:30:32,599][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:30:33,098][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:30:33,598][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:30:34,098][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:30:34,596][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:30:35,091][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:30:35,601][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:30:36,104][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:30:36,624][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:30:37,124][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:30:37,623][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:30:38,124][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:30:38,624][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:30:39,117][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:30:39,614][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:30:40,111][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:30:40,614][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:30:41,113][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:30:41,613][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:30:42,116][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:30:42,616][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:30:43,118][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:30:43,620][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:30:44,121][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:30:44,658][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:30:45,134][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:30:45,636][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:30:46,139][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:30:46,640][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10744 tokens. -[2025-11-09 21:30:47,398][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.41%, ΔTime: 00:00:32 -[2025-11-09 21:30:48,144][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:30:48,145][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:30:48,147][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:30:49,101][__main__][INFO] - Iteration 246 took 52s (31.46% Gen, 66.73% Train). Generation: 16s, Training: 35s. Estimated remaining time: 40h 17m 6s. Estimated total time: 43h 48m 58s. Time estimates for 10 more iterations: 8m 45s, 100 more iterations: 1h 27m 37s, 500 more iterations: 7h 18m 9s. -[2025-11-09 21:30:49,104][__main__][INFO] - Starting iteration 246. -[2025-11-09 21:30:49,567][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 24 and human policies 1. -[2025-11-09 21:30:49,568][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:31:05,991][__main__][INFO] - Number of regex retries in iteration 246: 0 -[2025-11-09 21:31:05,992][__main__][INFO] - agents played in iteration 246 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:31:06,817][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:31:06,844][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:31:06,870][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:31:06,893][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:31:06,894][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:31:06,895][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:31:07,576][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:31:08,035][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:31:08,541][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:31:09,042][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:31:09,545][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:31:10,044][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:31:10,545][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:31:11,047][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:31:11,546][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:31:12,058][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:31:12,559][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:31:13,062][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:31:13,562][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:31:14,064][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:31:14,566][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:31:15,067][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:31:15,569][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:31:16,088][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:31:16,592][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:31:17,093][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:31:17,595][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:31:18,094][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:31:18,596][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:31:19,097][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:31:19,598][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:31:20,110][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:31:20,611][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:31:21,113][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:31:21,617][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:31:22,117][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:31:22,620][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:31:23,120][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:31:23,619][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:31:24,119][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:31:24,650][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:31:25,155][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:31:25,657][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:31:26,160][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:31:26,679][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:31:27,179][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:31:27,680][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:31:28,188][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:31:28,692][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:31:29,208][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:31:29,707][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:31:30,206][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:31:30,713][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:31:31,210][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:31:31,708][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:31:32,208][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:31:32,709][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:31:33,218][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:31:33,720][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:31:34,223][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:31:34,725][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:31:35,226][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:31:35,728][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:31:36,229][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:31:36,730][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:31:37,233][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:31:37,734][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:31:38,235][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:31:38,734][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:31:39,234][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:31:39,735][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10808 tokens. -[2025-11-09 21:31:40,455][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 55.71%, Block Peak % of device VRAM: 62.42%, ΔTime: 00:00:32 -[2025-11-09 21:31:41,226][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:31:41,228][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:31:41,230][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:31:42,249][__main__][INFO] - Iteration 247 took 52s (31.17% Gen, 66.89% Train). Generation: 16s, Training: 35s. Estimated remaining time: 40h 21m 22s. Estimated total time: 43h 54m 7s. Time estimates for 10 more iterations: 8m 46s, 100 more iterations: 1h 27m 48s, 500 more iterations: 7h 19m 1s. -[2025-11-09 21:31:42,251][__main__][INFO] - Starting iteration 247. -[2025-11-09 21:31:42,718][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 24 and human policies 1. -[2025-11-09 21:31:42,718][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:31:51,681][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 21:31:59,235][__main__][INFO] - Number of regex retries in iteration 247: 1 -[2025-11-09 21:31:59,235][__main__][INFO] - agents played in iteration 247 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:32:00,012][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:32:00,038][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:32:00,063][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:32:00,086][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:32:00,086][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:32:00,087][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:32:00,799][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:32:01,260][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:32:01,768][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:32:02,289][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:32:02,794][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:32:03,302][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:32:03,806][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:32:04,309][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:32:04,811][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:32:05,315][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:32:05,816][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:32:06,321][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:32:06,821][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:32:07,322][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:32:07,824][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:32:08,325][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:32:08,826][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:32:09,324][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:32:09,822][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:32:10,320][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:32:10,816][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:32:11,319][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:32:11,818][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:32:12,316][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:32:12,814][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:32:13,316][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:32:13,817][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:32:14,315][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:32:14,815][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:32:15,320][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:32:15,820][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:32:16,322][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:32:16,821][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:32:17,321][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:32:17,823][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:32:18,322][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:32:18,821][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:32:19,329][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:32:19,831][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:32:20,331][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:32:20,828][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:32:21,327][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:32:21,829][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:32:22,327][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:32:22,827][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:32:23,329][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:32:23,826][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:32:24,326][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:32:24,824][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:32:25,330][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:32:25,831][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:32:26,335][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:32:26,868][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:32:27,375][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:32:27,874][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:32:28,380][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:32:28,884][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:32:29,386][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:32:29,887][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:32:30,387][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:32:30,903][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:32:31,403][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:32:31,902][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:32:32,399][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:32:32,897][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10672 tokens. -[2025-11-09 21:32:33,607][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.50%, ΔTime: 00:00:32 -[2025-11-09 21:32:34,342][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:32:34,344][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:32:34,345][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:32:35,380][__main__][INFO] - Iteration 248 took 52s (31.36% Gen, 66.67% Train). Generation: 16s, Training: 35s. Estimated remaining time: 40h 19m 30s. Estimated total time: 43h 53m 9s. Time estimates for 10 more iterations: 8m 46s, 100 more iterations: 1h 27m 46s, 500 more iterations: 7h 18m 51s. -[2025-11-09 21:32:35,382][__main__][INFO] - Starting iteration 248. -[2025-11-09 21:32:35,889][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 24 and human policies 1. -[2025-11-09 21:32:35,889][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:32:52,531][__main__][INFO] - Number of regex retries in iteration 248: 0 -[2025-11-09 21:32:52,531][__main__][INFO] - agents played in iteration 248 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:32:53,360][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:32:53,387][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:32:53,415][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:32:53,438][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:32:53,438][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:32:53,439][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:32:54,164][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:32:54,622][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:32:55,128][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:32:55,636][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:32:56,140][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:32:56,641][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:32:57,141][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:32:57,642][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:32:58,148][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:32:58,649][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:32:59,150][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:32:59,650][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:33:00,149][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:33:00,649][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:33:01,147][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:33:01,646][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:33:02,146][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:33:02,645][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:33:03,146][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:33:03,642][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:33:04,140][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:33:04,639][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:33:05,139][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:33:05,638][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:33:06,144][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:33:06,643][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:33:07,142][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:33:07,640][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:33:08,139][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:33:08,641][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:33:09,142][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:33:09,637][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:33:10,138][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:33:10,639][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:33:11,141][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:33:11,643][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:33:12,145][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:33:12,646][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:33:13,144][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:33:13,640][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:33:14,139][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:33:14,639][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:33:15,140][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:33:15,639][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:33:16,146][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:33:16,648][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:33:17,148][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:33:17,651][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:33:18,152][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:33:18,651][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:33:19,153][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:33:19,654][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:33:20,158][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:33:20,660][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:33:21,162][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:33:21,664][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:33:22,165][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:33:22,665][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:33:23,167][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:33:23,669][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:33:24,170][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:33:24,670][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:33:25,171][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:33:25,673][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:33:26,178][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10751 tokens. -[2025-11-09 21:33:26,938][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.27%, Current % of VRAM taken: 55.76%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 -[2025-11-09 21:33:27,690][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:33:27,691][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:33:27,693][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:33:28,686][__main__][INFO] - Iteration 249 took 52s (31.52% Gen, 66.60% Train). Generation: 16s, Training: 35s. Estimated remaining time: 40h 25m 22s. Estimated total time: 43h 59m 54s. Time estimates for 10 more iterations: 8m 47s, 100 more iterations: 1h 27m 59s, 500 more iterations: 7h 19m 59s. -[2025-11-09 21:33:28,689][__main__][INFO] - Starting iteration 249. -[2025-11-09 21:33:29,160][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 24 and human policies 1. -[2025-11-09 21:33:29,161][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:33:44,038][__main__][INFO] - Number of regex retries in iteration 249: 0 -[2025-11-09 21:33:44,038][__main__][INFO] - agents played in iteration 249 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:33:44,828][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:33:44,856][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:33:44,882][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:33:44,905][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:33:44,906][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:33:44,907][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:33:45,600][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:33:46,056][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:33:46,562][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:33:47,065][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:33:47,569][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:33:48,068][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:33:48,573][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:33:49,073][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:33:49,575][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:33:50,077][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:33:50,580][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:33:51,084][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:33:51,586][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:33:52,087][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:33:52,587][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:33:53,086][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:33:53,587][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:33:54,086][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:33:54,587][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:33:55,086][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:33:55,585][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:33:56,084][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:33:56,585][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:33:57,086][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:33:57,587][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:33:58,092][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:33:58,616][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:33:59,118][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:33:59,619][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:34:00,120][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:34:00,620][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:34:01,121][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:34:01,629][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:34:02,126][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:34:02,636][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:34:03,132][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:34:03,632][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:34:04,135][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:34:04,636][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:34:05,133][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:34:05,631][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:34:06,131][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:34:06,644][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:34:07,142][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:34:07,637][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:34:08,140][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:34:08,643][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:34:09,148][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:34:09,650][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:34:10,151][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:34:10,648][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:34:11,150][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:34:11,654][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:34:12,154][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:34:12,654][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:34:13,159][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:34:13,659][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:34:14,156][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:34:14,657][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:34:15,158][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:34:15,659][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:34:16,158][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:34:16,656][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:34:17,158][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:34:17,656][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10814 tokens. -[2025-11-09 21:34:18,351][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 55.70%, Block Peak % of device VRAM: 62.50%, ΔTime: 00:00:32 -[2025-11-09 21:34:19,083][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:34:19,084][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:34:19,086][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:34:20,138][__main__][INFO] - Iteration 250 took 50s (29.18% Gen, 68.75% Train). Generation: 14s, Training: 35s. Estimated remaining time: 38h 53m 32s. Estimated total time: 42h 28m 55s. Time estimates for 10 more iterations: 8m 29s, 100 more iterations: 1h 24m 57s, 500 more iterations: 7h 4m 49s. -[2025-11-09 21:34:20,140][__main__][INFO] - Starting iteration 250. -[2025-11-09 21:34:20,601][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 24 and human policies 1. -[2025-11-09 21:34:20,602][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:34:25,815][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 21:34:25,886][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 21:34:38,070][__main__][INFO] - Number of regex retries in iteration 250: 2 -[2025-11-09 21:34:38,070][__main__][INFO] - agents played in iteration 250 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:34:38,858][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:34:38,886][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:34:38,912][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:34:38,935][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:34:38,936][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:34:38,937][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:34:39,617][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:34:40,079][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:34:40,587][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:34:41,092][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:34:41,595][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:34:42,112][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:34:42,614][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:34:43,114][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:34:43,613][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:34:44,111][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:34:44,624][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:34:45,124][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:34:45,624][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:34:46,126][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:34:46,626][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:34:47,127][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:34:47,627][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:34:48,127][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:34:48,644][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:34:49,139][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:34:49,638][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:34:50,139][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:34:50,635][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:34:51,134][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:34:51,634][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:34:52,134][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:34:52,643][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:34:53,142][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:34:53,647][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:34:54,147][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:34:54,644][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:34:55,146][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:34:55,648][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:34:56,149][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:34:56,648][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:34:57,146][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:34:57,644][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:34:58,141][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:34:58,638][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:34:59,135][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:34:59,633][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:35:00,129][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:35:00,624][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:35:01,119][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:35:01,624][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:35:02,121][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:35:02,624][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:35:03,125][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:35:03,622][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:35:04,122][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:35:04,639][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:35:05,147][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:35:05,647][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:35:06,145][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:35:06,647][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:35:07,146][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:35:07,646][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:35:08,147][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:35:08,647][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:35:09,144][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:35:09,652][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:35:10,151][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:35:10,652][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:35:11,148][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:35:11,647][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10713 tokens. -[2025-11-09 21:35:12,356][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 55.69%, Block Peak % of device VRAM: 62.28%, ΔTime: 00:00:32 -[2025-11-09 21:35:13,102][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:35:13,104][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:35:13,105][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:35:15,117][__main__][INFO] - Iteration 251 took 54s (32.04% Gen, 64.26% Train). Generation: 17s, Training: 35s. Estimated remaining time: 41h 49m 32s. Estimated total time: 45h 25m 50s. Time estimates for 10 more iterations: 9m 5s, 100 more iterations: 1h 30m 51s, 500 more iterations: 7h 34m 18s. -[2025-11-09 21:35:15,120][__main__][INFO] - Starting iteration 251. -[2025-11-09 21:35:15,582][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 25 and human policies 1. -[2025-11-09 21:35:15,582][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:35:32,070][__main__][INFO] - Number of regex retries in iteration 251: 0 -[2025-11-09 21:35:32,071][__main__][INFO] - agents played in iteration 251 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:35:32,862][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:35:32,885][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:35:32,907][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:35:32,929][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:35:32,930][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:35:32,931][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:35:33,678][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:35:34,137][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:35:34,654][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:35:35,157][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:35:35,660][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:35:36,166][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:35:36,671][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:35:37,175][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:35:37,677][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:35:38,183][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:35:38,688][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:35:39,191][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:35:39,693][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:35:40,196][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:35:40,700][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:35:41,209][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:35:41,712][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:35:42,212][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:35:42,713][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:35:43,213][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:35:43,714][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:35:44,223][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:35:44,721][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:35:45,238][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:35:45,738][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:35:46,243][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:35:46,747][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:35:47,247][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:35:47,755][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:35:48,257][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:35:48,761][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:35:49,266][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:35:49,768][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:35:50,270][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:35:50,770][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:35:51,271][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:35:51,773][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:35:52,273][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:35:52,771][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:35:53,270][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:35:53,770][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:35:54,270][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:35:54,769][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:35:55,270][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:35:55,770][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:35:56,273][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:35:56,775][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:35:57,276][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:35:57,780][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:35:58,282][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:35:58,779][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:35:59,281][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:35:59,785][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:36:00,286][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:36:00,788][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:36:01,292][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:36:01,790][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:36:02,296][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:36:02,798][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:36:03,300][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:36:03,799][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:36:04,301][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:36:04,805][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:36:05,304][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:36:05,805][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10797 tokens. -[2025-11-09 21:36:06,498][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.27%, Current % of VRAM taken: 55.76%, Block Peak % of device VRAM: 62.54%, ΔTime: 00:00:32 -[2025-11-09 21:36:07,243][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:36:07,245][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:36:07,247][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:36:08,264][__main__][INFO] - Iteration 252 took 52s (31.30% Gen, 66.77% Train). Generation: 16s, Training: 35s. Estimated remaining time: 40h 16m 55s. Estimated total time: 43h 54m 7s. Time estimates for 10 more iterations: 8m 46s, 100 more iterations: 1h 27m 48s, 500 more iterations: 7h 19m 1s. -[2025-11-09 21:36:08,266][__main__][INFO] - Starting iteration 252. -[2025-11-09 21:36:08,750][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 25 and human policies 1. -[2025-11-09 21:36:08,751][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:36:24,119][__main__][INFO] - Number of regex retries in iteration 252: 0 -[2025-11-09 21:36:24,120][__main__][INFO] - agents played in iteration 252 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:36:24,925][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:36:24,949][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:36:24,971][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:36:24,994][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:36:24,994][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:36:24,995][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:36:25,680][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:36:26,145][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:36:26,653][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:36:27,158][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:36:27,662][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:36:28,163][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:36:28,668][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:36:29,171][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:36:29,676][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:36:30,180][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:36:30,681][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:36:31,180][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:36:31,682][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:36:32,183][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:36:32,685][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:36:33,185][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:36:33,686][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:36:34,188][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:36:34,688][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:36:35,187][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:36:35,689][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:36:36,189][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:36:36,693][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:36:37,195][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:36:37,694][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:36:38,194][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:36:38,693][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:36:39,193][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:36:39,694][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:36:40,196][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:36:40,697][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:36:41,195][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:36:41,693][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:36:42,193][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:36:42,690][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:36:43,190][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:36:43,689][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:36:44,187][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:36:44,689][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:36:45,187][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:36:45,687][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:36:46,184][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:36:46,680][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:36:47,177][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:36:47,675][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:36:48,176][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:36:48,679][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:36:49,178][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:36:49,679][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:36:50,187][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:36:50,689][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:36:51,192][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:36:51,697][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:36:52,198][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:36:52,700][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:36:53,202][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:36:53,706][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:36:54,209][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:36:54,710][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:36:55,210][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:36:55,712][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:36:56,211][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:36:56,710][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:36:57,209][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:36:57,710][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10819 tokens. -[2025-11-09 21:36:58,391][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 55.70%, Block Peak % of device VRAM: 62.43%, ΔTime: 00:00:32 -[2025-11-09 21:36:59,149][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:36:59,151][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:36:59,153][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:37:00,166][__main__][INFO] - Iteration 253 took 51s (29.89% Gen, 68.13% Train). Generation: 15s, Training: 35s. Estimated remaining time: 39h 12m 47s. Estimated total time: 42h 50m 50s. Time estimates for 10 more iterations: 8m 34s, 100 more iterations: 1h 25m 41s, 500 more iterations: 7h 8m 28s. -[2025-11-09 21:37:00,170][__main__][INFO] - Starting iteration 253. -[2025-11-09 21:37:00,639][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 25 and human policies 1. -[2025-11-09 21:37:00,639][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:37:05,021][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 21:37:07,465][mllm.models.large_language_model_local][WARNING] - Response Proposal: 2 hats, 7 books, 11 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 21:37:09,135][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 21:37:12,674][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 21:37:17,682][__main__][INFO] - Number of regex retries in iteration 253: 4 -[2025-11-09 21:37:17,682][__main__][INFO] - agents played in iteration 253 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:37:18,502][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:37:18,527][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:37:18,554][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:37:18,577][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:37:18,578][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:37:18,578][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:37:19,267][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:37:19,722][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:37:20,240][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:37:20,740][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:37:21,241][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:37:21,740][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:37:22,239][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:37:22,752][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:37:23,252][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:37:23,752][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:37:24,257][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:37:24,759][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:37:25,268][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:37:25,768][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:37:26,270][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:37:26,777][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:37:27,277][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:37:27,775][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:37:28,272][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:37:28,774][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:37:29,277][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:37:29,779][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:37:30,279][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:37:30,775][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:37:31,275][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:37:31,777][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:37:32,280][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:37:32,779][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:37:33,277][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:37:33,775][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:37:34,270][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:37:34,769][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:37:35,268][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:37:35,769][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:37:36,266][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:37:36,763][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:37:37,272][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:37:37,768][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:37:38,266][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:37:38,770][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:37:39,268][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:37:39,763][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:37:40,264][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:37:40,764][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:37:41,274][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:37:41,773][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:37:42,274][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:37:42,770][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:37:43,271][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:37:43,775][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:37:44,275][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:37:44,773][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:37:45,275][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:37:45,776][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:37:46,281][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:37:46,781][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:37:47,281][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:37:47,786][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:37:48,285][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:37:48,785][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:37:49,285][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:37:49,782][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:37:50,285][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:37:50,785][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:37:51,285][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10751 tokens. -[2025-11-09 21:37:51,970][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.26%, Current % of VRAM taken: 55.76%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:32 -[2025-11-09 21:37:52,715][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:37:52,717][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:37:52,719][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:37:53,758][__main__][INFO] - Iteration 254 took 53s (32.08% Gen, 65.96% Train). Generation: 17s, Training: 35s. Estimated remaining time: 40h 37m 1s. Estimated total time: 44h 15m 58s. Time estimates for 10 more iterations: 8m 51s, 100 more iterations: 1h 28m 31s, 500 more iterations: 7h 22m 39s. -[2025-11-09 21:37:53,760][__main__][INFO] - Starting iteration 254. -[2025-11-09 21:37:54,246][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 25 and human policies 1. -[2025-11-09 21:37:54,247][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:38:11,347][__main__][INFO] - Number of regex retries in iteration 254: 0 -[2025-11-09 21:38:11,348][__main__][INFO] - agents played in iteration 254 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:38:12,141][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:38:12,168][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:38:12,194][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:38:12,217][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:38:12,218][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:38:12,219][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:38:12,890][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:38:13,352][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:38:13,859][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:38:14,367][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:38:14,868][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:38:15,372][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:38:15,877][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:38:16,380][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:38:16,882][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:38:17,384][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:38:17,888][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:38:18,396][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:38:18,896][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:38:19,396][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:38:19,896][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:38:20,396][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:38:20,899][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:38:21,401][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:38:21,896][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:38:22,396][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:38:22,897][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:38:23,398][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:38:23,899][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:38:24,402][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:38:24,903][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:38:25,402][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:38:25,899][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:38:26,400][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:38:26,898][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:38:27,397][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:38:27,895][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:38:28,393][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:38:28,891][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:38:29,391][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:38:29,888][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:38:30,390][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:38:30,887][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:38:31,388][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:38:31,881][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:38:32,380][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:38:32,887][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:38:33,390][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:38:33,892][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:38:34,400][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:38:34,905][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:38:35,408][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:38:35,940][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:38:36,447][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:38:36,953][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:38:37,454][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:38:37,953][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:38:38,453][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:38:38,951][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:38:39,449][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:38:39,949][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:38:40,449][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:38:40,949][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:38:41,444][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:38:41,942][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:38:42,441][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:38:42,938][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:38:43,435][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:38:43,936][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:38:44,434][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:38:44,944][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10727 tokens. -[2025-11-09 21:38:45,614][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.28%, Current % of VRAM taken: 55.78%, Block Peak % of device VRAM: 62.53%, ΔTime: 00:00:32 -[2025-11-09 21:38:46,384][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:38:46,385][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:38:46,387][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:38:47,406][__main__][INFO] - Iteration 255 took 53s (32.17% Gen, 65.91% Train). Generation: 17s, Training: 35s. Estimated remaining time: 40h 38m 11s. Estimated total time: 44h 18m 2s. Time estimates for 10 more iterations: 8m 51s, 100 more iterations: 1h 28m 36s, 500 more iterations: 7h 23m 0s. -[2025-11-09 21:38:47,408][__main__][INFO] - Starting iteration 255. -[2025-11-09 21:38:47,873][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 25 and human policies 1. -[2025-11-09 21:38:47,874][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:38:59,920][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 21:39:04,659][__main__][INFO] - Number of regex retries in iteration 255: 1 -[2025-11-09 21:39:04,659][__main__][INFO] - agents played in iteration 255 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:39:05,445][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.67%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:39:05,473][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.67%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:39:05,499][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.67%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:39:05,522][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.67%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:39:05,522][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:39:05,524][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:39:06,198][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:39:06,656][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:39:07,178][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:39:07,682][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:39:08,189][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:39:08,694][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:39:09,199][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:39:09,703][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:39:10,205][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:39:10,706][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:39:11,210][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:39:11,712][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:39:12,215][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:39:12,721][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:39:13,222][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:39:13,726][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:39:14,229][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:39:14,728][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:39:15,229][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:39:15,730][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:39:16,232][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:39:16,735][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:39:17,233][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:39:17,733][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:39:18,232][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:39:18,733][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:39:19,232][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:39:19,731][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:39:20,234][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:39:20,732][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:39:21,229][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:39:21,730][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:39:22,230][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:39:22,731][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:39:23,230][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:39:23,729][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:39:24,226][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:39:24,725][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:39:25,224][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:39:25,725][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:39:26,220][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:39:26,721][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:39:27,224][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:39:27,726][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:39:28,231][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:39:28,730][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:39:29,231][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:39:29,731][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:39:30,233][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:39:30,732][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:39:31,232][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:39:31,733][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:39:32,237][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:39:32,739][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:39:33,241][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:39:33,743][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:39:34,245][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:39:34,743][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:39:35,241][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:39:35,740][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:39:36,243][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:39:36,745][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:39:37,243][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:39:37,748][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:39:38,245][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10697 tokens. -[2025-11-09 21:39:38,941][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 55.71%, Block Peak % of device VRAM: 62.38%, ΔTime: 00:00:32 -[2025-11-09 21:39:39,689][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:39:39,690][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:39:39,692][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:39:40,714][__main__][INFO] - Iteration 256 took 52s (31.77% Gen, 66.30% Train). Generation: 16s, Training: 35s. Estimated remaining time: 40h 21m 19s. Estimated total time: 44h 2m 3s. Time estimates for 10 more iterations: 8m 48s, 100 more iterations: 1h 28m 4s, 500 more iterations: 7h 20m 20s. -[2025-11-09 21:39:40,716][__main__][INFO] - Starting iteration 256. -[2025-11-09 21:39:41,179][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 25 and human policies 1. -[2025-11-09 21:39:41,180][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:39:56,110][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 21:39:57,559][__main__][INFO] - Number of regex retries in iteration 256: 1 -[2025-11-09 21:39:57,559][__main__][INFO] - agents played in iteration 256 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:39:58,377][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:39:58,405][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:39:58,432][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:39:58,455][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:39:58,456][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:39:58,456][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:39:59,129][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:39:59,584][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:40:00,105][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:40:00,602][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:40:01,104][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:40:01,604][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:40:02,103][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:40:02,606][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:40:03,104][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:40:03,609][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:40:04,113][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:40:04,612][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:40:05,112][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:40:05,612][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:40:06,117][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:40:06,624][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:40:07,127][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:40:07,632][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:40:08,152][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:40:08,651][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:40:09,154][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:40:09,656][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:40:10,156][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:40:10,666][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:40:11,165][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:40:11,662][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:40:12,161][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:40:12,658][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:40:13,169][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:40:13,664][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:40:14,157][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:40:14,655][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:40:15,151][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:40:15,654][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:40:16,152][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:40:16,651][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:40:17,170][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:40:17,668][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:40:18,168][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:40:18,667][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:40:19,167][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:40:19,684][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:40:20,188][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:40:20,688][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:40:21,190][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:40:21,693][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:40:22,189][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:40:22,692][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:40:23,193][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:40:23,694][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:40:24,197][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:40:24,698][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:40:25,201][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:40:25,702][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:40:26,202][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:40:26,697][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:40:27,195][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:40:27,696][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:40:28,196][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:40:28,701][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:40:29,200][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:40:29,695][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:40:30,202][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:40:30,701][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:40:31,200][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10673 tokens. -[2025-11-09 21:40:31,898][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 55.71%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 -[2025-11-09 21:40:32,685][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:40:32,687][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:40:32,689][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:40:33,717][__main__][INFO] - Iteration 257 took 52s (31.18% Gen, 66.86% Train). Generation: 16s, Training: 35s. Estimated remaining time: 40h 5m 19s. Estimated total time: 43h 46m 56s. Time estimates for 10 more iterations: 8m 45s, 100 more iterations: 1h 27m 33s, 500 more iterations: 7h 17m 49s. -[2025-11-09 21:40:33,719][__main__][INFO] - Starting iteration 257. -[2025-11-09 21:40:34,185][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 25 and human policies 1. -[2025-11-09 21:40:34,187][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:40:40,291][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 21:40:52,548][__main__][INFO] - Number of regex retries in iteration 257: 1 -[2025-11-09 21:40:52,549][__main__][INFO] - agents played in iteration 257 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:40:53,367][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:40:53,398][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:40:53,425][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:40:53,449][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:40:53,450][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:40:53,451][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:40:54,113][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:40:54,568][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:40:55,070][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:40:55,574][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:40:56,075][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:40:56,575][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:40:57,074][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:40:57,573][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:40:58,073][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:40:58,579][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:40:59,079][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:40:59,580][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:41:00,079][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:41:00,579][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:41:01,080][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:41:01,577][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:41:02,079][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:41:02,582][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:41:03,085][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:41:03,586][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:41:04,086][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:41:04,587][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:41:05,092][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:41:05,591][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:41:06,099][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:41:06,598][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:41:07,110][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:41:07,608][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:41:08,103][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:41:08,613][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:41:09,112][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:41:09,622][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:41:10,116][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:41:10,614][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:41:11,141][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:41:11,640][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:41:12,143][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:41:12,642][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:41:13,144][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:41:13,650][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:41:14,154][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:41:14,656][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:41:15,161][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:41:15,664][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:41:16,168][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:41:16,671][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:41:17,173][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:41:17,675][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:41:18,182][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:41:18,681][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:41:19,182][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:41:19,681][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:41:20,181][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:41:20,679][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:41:21,178][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:41:21,677][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:41:22,174][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:41:22,670][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:41:23,169][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:41:23,668][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:41:24,169][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:41:24,668][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:41:25,171][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:41:25,672][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:41:26,172][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10729 tokens. -[2025-11-09 21:41:26,896][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.29%, Current % of VRAM taken: 55.79%, Block Peak % of device VRAM: 62.49%, ΔTime: 00:00:32 -[2025-11-09 21:41:27,652][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:41:27,654][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:41:27,656][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:41:28,682][__main__][INFO] - Iteration 258 took 54s (33.69% Gen, 64.42% Train). Generation: 18s, Training: 35s. Estimated remaining time: 41h 42m 21s. Estimated total time: 45h 24m 53s. Time estimates for 10 more iterations: 9m 4s, 100 more iterations: 1h 30m 49s, 500 more iterations: 7h 34m 8s. -[2025-11-09 21:41:28,685][__main__][INFO] - Starting iteration 258. -[2025-11-09 21:41:29,158][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 25 and human policies 1. -[2025-11-09 21:41:29,159][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:41:45,047][__main__][INFO] - Number of regex retries in iteration 258: 0 -[2025-11-09 21:41:45,048][__main__][INFO] - agents played in iteration 258 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:41:45,889][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.67%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:41:45,916][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.67%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:41:45,942][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.67%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:41:45,966][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.67%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:41:45,966][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:41:45,967][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:41:46,635][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:41:47,091][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:41:47,600][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:41:48,105][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:41:48,612][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:41:49,118][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:41:49,619][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:41:50,122][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:41:50,624][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:41:51,122][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:41:51,622][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:41:52,122][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:41:52,623][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:41:53,125][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:41:53,627][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:41:54,127][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:41:54,623][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:41:55,120][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:41:55,620][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:41:56,120][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:41:56,621][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:41:57,124][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:41:57,625][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:41:58,126][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:41:58,627][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:41:59,125][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:41:59,635][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:42:00,132][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:42:00,628][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:42:01,126][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:42:01,623][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:42:02,124][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:42:02,619][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:42:03,116][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:42:03,637][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:42:04,132][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:42:04,631][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:42:05,132][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:42:05,632][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:42:06,141][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:42:06,642][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:42:07,147][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:42:07,652][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:42:08,152][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:42:08,654][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:42:09,161][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:42:09,662][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:42:10,166][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:42:10,670][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:42:11,173][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:42:11,678][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:42:12,178][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:42:12,676][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:42:13,172][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:42:13,670][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:42:14,169][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:42:14,668][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:42:15,169][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:42:15,673][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:42:16,174][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:42:16,674][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:42:17,175][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:42:17,675][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:42:18,175][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:42:18,674][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10730 tokens. -[2025-11-09 21:42:19,353][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 -[2025-11-09 21:42:20,096][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:42:20,098][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:42:20,100][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:42:21,166][__main__][INFO] - Iteration 259 took 52s (30.55% Gen, 67.40% Train). Generation: 15s, Training: 35s. Estimated remaining time: 39h 37m 2s. Estimated total time: 43h 20m 26s. Time estimates for 10 more iterations: 8m 40s, 100 more iterations: 1h 26m 40s, 500 more iterations: 7h 13m 24s. -[2025-11-09 21:42:21,169][__main__][INFO] - Starting iteration 259. -[2025-11-09 21:42:21,650][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 25 and human policies 1. -[2025-11-09 21:42:21,651][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:42:38,475][__main__][INFO] - Number of regex retries in iteration 259: 0 -[2025-11-09 21:42:38,476][__main__][INFO] - agents played in iteration 259 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:42:39,336][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:42:39,361][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:42:39,384][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:42:39,406][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:42:39,407][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:42:39,408][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:42:40,145][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:42:40,617][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:42:41,124][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:42:41,629][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:42:42,135][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:42:42,634][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:42:43,144][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:42:43,647][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:42:44,148][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:42:44,657][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:42:45,154][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:42:45,652][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:42:46,157][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:42:46,651][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:42:47,153][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:42:47,653][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:42:48,151][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:42:48,659][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:42:49,161][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:42:49,659][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:42:50,161][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:42:50,660][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:42:51,165][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:42:51,665][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:42:52,168][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:42:52,670][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:42:53,170][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:42:53,673][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:42:54,169][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:42:54,669][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:42:55,173][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:42:55,673][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:42:56,172][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:42:56,675][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:42:57,176][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:42:57,698][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:42:58,205][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:42:58,708][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:42:59,208][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:42:59,717][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:43:00,230][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:43:00,732][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:43:01,233][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:43:01,738][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:43:02,241][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:43:02,745][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:43:03,250][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:43:03,753][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:43:04,263][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:43:04,767][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:43:05,269][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:43:05,767][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:43:06,273][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:43:06,772][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:43:07,271][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:43:07,768][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:43:08,269][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:43:08,767][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:43:09,266][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:43:09,761][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:43:10,257][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:43:10,756][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:43:11,253][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:43:11,752][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:43:12,251][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10652 tokens. -[2025-11-09 21:43:12,934][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 55.66%, Block Peak % of device VRAM: 62.20%, ΔTime: 00:00:32 -[2025-11-09 21:43:13,696][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:43:13,698][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:43:13,700][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:43:14,715][__main__][INFO] - Iteration 260 took 53s (31.71% Gen, 66.38% Train). Generation: 16s, Training: 35s. Estimated remaining time: 40h 28m 58s. Estimated total time: 44h 13m 16s. Time estimates for 10 more iterations: 8m 50s, 100 more iterations: 1h 28m 26s, 500 more iterations: 7h 22m 12s. -[2025-11-09 21:43:14,717][__main__][INFO] - Starting iteration 260. -[2025-11-09 21:43:15,193][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 25 and human policies 1. -[2025-11-09 21:43:15,194][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:43:23,889][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 21:43:35,850][__main__][INFO] - Number of regex retries in iteration 260: 1 -[2025-11-09 21:43:35,851][__main__][INFO] - agents played in iteration 260 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:43:36,669][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:43:36,692][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:43:36,718][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:43:36,740][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:43:36,741][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:43:36,741][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:43:37,502][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:43:37,959][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:43:38,473][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:43:38,977][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:43:39,477][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:43:39,978][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:43:40,478][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:43:40,980][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:43:41,486][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:43:41,987][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:43:42,488][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:43:42,990][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:43:43,490][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:43:43,990][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:43:44,490][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:43:44,993][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:43:45,492][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:43:45,991][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:43:46,491][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:43:46,989][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:43:47,488][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:43:47,991][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:43:48,491][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:43:48,992][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:43:49,491][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:43:49,991][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:43:50,493][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:43:50,993][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:43:51,492][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:43:51,992][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:43:52,492][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:43:52,998][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:43:53,504][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:43:54,010][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:43:54,513][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:43:55,018][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:43:55,527][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:43:56,030][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:43:56,537][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:43:57,035][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:43:57,535][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:43:58,036][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:43:58,538][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:43:59,038][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:43:59,539][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:44:00,041][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:44:00,564][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:44:01,067][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:44:01,568][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:44:02,071][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:44:02,575][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:44:03,079][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:44:03,575][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:44:04,076][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:44:04,584][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:44:05,082][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:44:05,584][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:44:06,086][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:44:06,588][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:44:07,095][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:44:07,596][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:44:08,098][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:44:08,601][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:44:09,101][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:44:09,602][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10629 tokens. -[2025-11-09 21:44:10,326][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 -[2025-11-09 21:44:11,085][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:44:11,086][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:44:11,088][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:44:14,434][__main__][INFO] - Iteration 261 took 59s (34.87% Gen, 59.48% Train). Generation: 20s, Training: 35s. Estimated remaining time: 45h 36m 47s. Estimated total time: 49h 22m 5s. Time estimates for 10 more iterations: 9m 52s, 100 more iterations: 1h 38m 44s, 500 more iterations: 8h 13m 40s. -[2025-11-09 21:44:14,438][__main__][INFO] - Starting iteration 261. -[2025-11-09 21:44:14,960][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 26 and human policies 1. -[2025-11-09 21:44:14,961][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:44:44,173][__main__][INFO] - Number of regex retries in iteration 261: 0 -[2025-11-09 21:44:44,174][__main__][INFO] - agents played in iteration 261 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:44:45,000][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:44:45,023][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:44:45,047][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:44:45,070][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:44:45,071][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:44:45,072][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:44:45,808][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:44:46,274][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:44:46,784][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:44:47,287][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:44:47,789][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:44:48,288][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:44:48,790][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:44:49,291][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:44:49,791][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:44:50,289][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:44:50,800][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:44:51,299][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:44:51,803][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:44:52,308][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:44:52,812][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:44:53,313][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:44:53,817][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:44:54,327][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:44:54,828][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:44:55,331][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:44:55,842][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:44:56,345][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:44:56,848][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:44:57,351][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:44:57,862][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:44:58,393][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:44:58,904][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:44:59,410][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:44:59,923][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:45:00,435][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:45:00,949][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:45:01,453][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:45:01,959][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:45:02,467][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:45:02,972][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:45:03,476][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:45:03,983][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:45:04,491][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:45:04,998][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:45:05,502][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:45:06,009][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:45:06,513][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:45:07,015][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:45:07,517][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:45:08,021][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:45:08,524][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:45:09,026][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:45:09,526][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:45:10,029][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:45:10,528][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:45:11,027][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:45:11,528][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:45:12,025][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:45:12,524][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:45:13,024][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:45:13,524][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:45:14,024][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:45:14,523][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:45:15,020][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:45:15,518][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:45:16,016][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:45:16,512][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:45:17,010][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:45:17,510][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:45:18,012][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10698 tokens. -[2025-11-09 21:45:18,694][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 55.68%, Block Peak % of device VRAM: 62.58%, ΔTime: 00:00:32 -[2025-11-09 21:45:19,395][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:45:19,396][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:45:19,398][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:45:20,479][__main__][INFO] - Iteration 262 took 1m 5s (44.59% Gen, 53.76% Train). Generation: 29s, Training: 35s. Estimated remaining time: 50h 49m 35s. Estimated total time: 54h 35m 59s. Time estimates for 10 more iterations: 10m 55s, 100 more iterations: 1h 49m 11s, 500 more iterations: 9h 5m 59s. -[2025-11-09 21:45:20,481][__main__][INFO] - Starting iteration 262. -[2025-11-09 21:45:20,947][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 26 and human policies 1. -[2025-11-09 21:45:20,948][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:45:45,250][__main__][INFO] - Number of regex retries in iteration 262: 0 -[2025-11-09 21:45:45,251][__main__][INFO] - agents played in iteration 262 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:45:46,039][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:45:46,067][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:45:46,093][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:45:46,117][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:45:46,117][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:45:46,118][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:45:46,788][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:45:47,247][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:45:47,753][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:45:48,252][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:45:48,751][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:45:49,268][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:45:49,771][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:45:50,270][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:45:50,773][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:45:51,278][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:45:51,777][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:45:52,276][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:45:52,777][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:45:53,291][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:45:53,792][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:45:54,299][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:45:54,805][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:45:55,307][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:45:55,807][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:45:56,307][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:45:56,811][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:45:57,314][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:45:57,817][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:45:58,318][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:45:58,819][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:45:59,324][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:45:59,827][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:46:00,331][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:46:00,835][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:46:01,340][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:46:01,839][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:46:02,344][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:46:02,844][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:46:03,351][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:46:03,889][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:46:04,391][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:46:04,899][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:46:05,401][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:46:05,902][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:46:06,405][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:46:06,904][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:46:07,405][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:46:07,904][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:46:08,400][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:46:08,901][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:46:09,399][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:46:09,897][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:46:10,400][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:46:10,899][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:46:11,397][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:46:11,897][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:46:12,395][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:46:12,894][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:46:13,392][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:46:13,889][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:46:14,400][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:46:14,899][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:46:15,399][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:46:15,902][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:46:16,402][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:46:16,905][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:46:17,404][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:46:17,906][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:46:18,410][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:46:18,915][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10684 tokens. -[2025-11-09 21:46:19,628][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.25%, Current % of VRAM taken: 55.75%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:32 -[2025-11-09 21:46:20,350][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:46:20,352][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:46:20,353][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:46:21,359][__main__][INFO] - Iteration 263 took 1m 0s (40.23% Gen, 58.10% Train). Generation: 24s, Training: 35s. Estimated remaining time: 46h 33m 12s. Estimated total time: 50h 20m 37s. Time estimates for 10 more iterations: 10m 4s, 100 more iterations: 1h 40m 41s, 500 more iterations: 8h 23m 26s. -[2025-11-09 21:46:21,361][__main__][INFO] - Starting iteration 263. -[2025-11-09 21:46:21,839][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 26 and human policies 1. -[2025-11-09 21:46:21,839][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:46:41,256][__main__][INFO] - Number of regex retries in iteration 263: 0 -[2025-11-09 21:46:41,256][__main__][INFO] - agents played in iteration 263 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:46:42,064][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:46:42,091][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:46:42,118][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:46:42,141][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:46:42,142][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:46:42,143][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:46:42,780][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:46:43,240][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:46:43,741][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:46:44,243][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:46:44,743][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:46:45,241][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:46:45,741][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:46:46,238][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:46:46,740][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:46:47,244][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:46:47,745][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:46:48,244][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:46:48,742][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:46:49,242][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:46:49,742][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:46:50,243][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:46:50,743][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:46:51,243][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:46:51,744][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:46:52,245][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:46:52,748][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:46:53,249][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:46:53,750][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:46:54,257][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:46:54,757][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:46:55,257][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:46:55,756][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:46:56,255][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:46:56,771][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:46:57,274][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:46:57,780][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:46:58,280][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:46:58,780][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:46:59,303][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:46:59,986][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:47:01,243][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:47:01,936][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:47:02,440][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:47:02,936][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:47:03,437][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:47:03,936][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:47:04,433][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:47:04,930][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:47:05,426][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:47:05,921][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:47:06,420][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:47:06,918][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:47:07,414][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:47:07,910][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:47:08,408][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:47:08,908][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:47:09,408][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:47:09,916][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:47:10,411][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:47:10,910][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:47:11,408][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:47:11,912][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:47:12,415][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:47:12,915][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:47:13,414][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:47:13,916][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:47:14,418][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:47:14,919][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:47:15,427][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:47:15,930][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10501 tokens. -[2025-11-09 21:47:16,639][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 55.72%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:33 -[2025-11-09 21:47:17,271][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:47:17,273][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:47:17,274][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:47:18,244][__main__][INFO] - Iteration 264 took 56s (34.42% Gen, 63.86% Train). Generation: 19s, Training: 36s. Estimated remaining time: 43h 11m 55s. Estimated total time: 47h 0m 17s. Time estimates for 10 more iterations: 9m 24s, 100 more iterations: 1h 34m 0s, 500 more iterations: 7h 50m 2s. -[2025-11-09 21:47:18,246][__main__][INFO] - Starting iteration 264. -[2025-11-09 21:47:18,729][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 26 and human policies 1. -[2025-11-09 21:47:18,729][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:47:36,354][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 21:47:49,070][__main__][INFO] - Number of regex retries in iteration 264: 1 -[2025-11-09 21:47:49,070][__main__][INFO] - agents played in iteration 264 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:47:49,869][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:47:49,892][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:47:49,917][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:47:49,940][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:47:49,941][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:47:49,942][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:47:50,674][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:47:51,146][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:47:51,653][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:47:52,155][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:47:52,655][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:47:53,156][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:47:53,665][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:47:54,167][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:47:54,670][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:47:55,191][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:47:55,693][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:47:56,207][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:47:56,709][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:47:57,214][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:47:57,720][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:47:58,223][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:47:58,726][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:47:59,226][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:47:59,728][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:48:00,229][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:48:00,731][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:48:01,230][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:48:01,733][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:48:02,232][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:48:02,732][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:48:03,236][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:48:03,738][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:48:04,237][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:48:04,743][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:48:05,244][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:48:05,748][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:48:06,251][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:48:06,755][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:48:07,255][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:48:07,755][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:48:08,257][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:48:08,756][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:48:09,259][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:48:09,763][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:48:10,267][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:48:10,772][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:48:11,272][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:48:11,773][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:48:12,272][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:48:12,772][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:48:13,280][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:48:13,777][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:48:14,277][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:48:14,792][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:48:15,292][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:48:15,806][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:48:16,309][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:48:16,830][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:48:17,337][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:48:17,840][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:48:18,344][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:48:18,845][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:48:19,341][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:48:19,845][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:48:20,345][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:48:20,843][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:48:21,344][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:48:21,847][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:48:22,345][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:48:22,845][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10578 tokens. -[2025-11-09 21:48:23,551][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 55.66%, Block Peak % of device VRAM: 62.34%, ΔTime: 00:00:32 -[2025-11-09 21:48:24,294][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:48:24,296][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:48:24,298][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:48:25,285][__main__][INFO] - Iteration 265 took 1m 6s (45.59% Gen, 52.93% Train). Generation: 30s, Training: 35s. Estimated remaining time: 51h 38m 23s. Estimated total time: 55h 27m 51s. Time estimates for 10 more iterations: 11m 5s, 100 more iterations: 1h 50m 55s, 500 more iterations: 9h 14m 38s. -[2025-11-09 21:48:25,288][__main__][INFO] - Starting iteration 265. -[2025-11-09 21:48:25,758][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 26 and human policies 1. -[2025-11-09 21:48:25,758][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:48:47,042][__main__][INFO] - Number of regex retries in iteration 265: 0 -[2025-11-09 21:48:47,042][__main__][INFO] - agents played in iteration 265 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:48:47,853][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:48:47,877][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:48:47,900][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:48:47,923][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:48:47,924][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:48:47,924][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:48:48,620][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:48:49,077][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:48:49,586][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:48:50,090][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:48:50,589][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:48:51,085][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:48:51,596][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:48:52,092][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:48:52,596][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:48:53,093][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:48:53,589][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:48:54,099][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:48:54,599][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:48:55,096][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:48:55,608][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:48:56,109][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:48:56,629][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:48:57,131][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:48:57,631][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:48:58,138][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:48:58,640][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:48:59,141][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:48:59,644][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:49:00,146][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:49:00,651][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:49:01,154][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:49:01,655][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:49:02,153][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:49:02,656][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:49:03,158][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:49:03,659][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:49:04,159][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:49:04,661][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:49:05,184][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:49:05,688][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:49:06,192][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:49:06,695][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:49:07,196][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:49:07,697][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:49:08,197][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:49:08,695][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:49:09,196][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:49:09,711][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:49:10,210][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:49:10,709][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:49:11,210][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:49:11,711][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:49:12,209][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:49:12,707][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:49:13,207][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:49:13,719][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:49:14,221][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:49:14,718][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:49:15,223][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:49:15,723][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:49:16,232][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:49:16,734][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:49:17,234][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:49:17,740][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:49:18,237][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:49:18,741][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:49:19,242][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:49:19,742][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:49:20,244][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:49:20,743][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10585 tokens. -[2025-11-09 21:49:21,467][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 55.65%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:32 -[2025-11-09 21:49:22,182][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:49:22,184][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:49:22,185][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:49:23,188][__main__][INFO] - Iteration 266 took 57s (37.06% Gen, 61.19% Train). Generation: 21s, Training: 35s. Estimated remaining time: 44h 1m 8s. Estimated total time: 47h 51m 34s. Time estimates for 10 more iterations: 9m 34s, 100 more iterations: 1h 35m 43s, 500 more iterations: 7h 58m 35s. -[2025-11-09 21:49:23,190][__main__][INFO] - Starting iteration 266. -[2025-11-09 21:49:23,654][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 26 and human policies 1. -[2025-11-09 21:49:23,655][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:49:30,075][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 21:49:43,090][__main__][INFO] - Number of regex retries in iteration 266: 1 -[2025-11-09 21:49:43,091][__main__][INFO] - agents played in iteration 266 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:49:43,897][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:49:43,920][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:49:43,942][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:49:43,964][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:49:43,965][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:49:43,965][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:49:44,653][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:49:45,111][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:49:45,621][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:49:46,124][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:49:46,619][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:49:47,116][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:49:47,620][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:49:48,138][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:49:48,637][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:49:49,138][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:49:49,639][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:49:50,139][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:49:50,652][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:49:51,153][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:49:51,657][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:49:52,163][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:49:52,666][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:49:53,168][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:49:53,668][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:49:54,169][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:49:54,673][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:49:55,175][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:49:55,676][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:49:56,177][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:49:56,676][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:49:57,179][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:49:57,679][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:49:58,178][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:49:58,681][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:49:59,180][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:49:59,681][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:50:00,181][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:50:00,680][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:50:01,184][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:50:01,687][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:50:02,207][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:50:02,727][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:50:03,227][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:50:03,739][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:50:04,239][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:50:04,740][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:50:05,250][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:50:05,746][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:50:06,243][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:50:06,738][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:50:07,234][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:50:07,741][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:50:08,239][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:50:08,737][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:50:09,237][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:50:09,735][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:50:10,234][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:50:10,737][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:50:11,237][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:50:11,737][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:50:12,990][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:50:13,995][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:50:14,502][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:50:15,007][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:50:15,510][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:50:16,013][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:50:16,515][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:50:17,033][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:50:17,541][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:50:18,043][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10524 tokens. -[2025-11-09 21:50:18,839][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.33%, ΔTime: 00:00:34 -[2025-11-09 21:50:19,495][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:50:19,497][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:50:19,499][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:50:20,421][__main__][INFO] - Iteration 267 took 56s (34.24% Gen, 64.14% Train). Generation: 19s, Training: 36s. Estimated remaining time: 43h 26m 56s. Estimated total time: 47h 18m 20s. Time estimates for 10 more iterations: 9m 27s, 100 more iterations: 1h 34m 36s, 500 more iterations: 7h 53m 3s. -[2025-11-09 21:50:20,425][__main__][INFO] - Starting iteration 267. -[2025-11-09 21:50:20,899][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 26 and human policies 1. -[2025-11-09 21:50:20,900][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:50:43,697][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 21:50:54,921][__main__][INFO] - Number of regex retries in iteration 267: 1 -[2025-11-09 21:50:54,921][__main__][INFO] - agents played in iteration 267 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:50:55,725][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:50:55,754][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:50:55,780][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:50:55,804][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:50:55,804][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:50:55,805][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:50:56,557][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:50:57,026][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:50:57,532][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:50:58,045][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:50:58,546][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:50:59,050][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:50:59,553][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:51:00,054][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:51:00,556][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:51:01,055][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:51:01,557][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:51:02,062][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:51:02,565][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:51:03,069][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:51:03,572][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:51:04,074][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:51:04,575][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:51:05,082][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:51:05,583][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:51:06,080][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:51:06,582][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:51:07,079][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:51:07,580][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:51:08,078][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:51:08,581][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:51:09,080][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:51:09,585][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:51:10,087][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:51:10,587][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:51:11,105][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:51:11,625][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:51:12,127][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:51:12,643][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:51:13,143][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:51:13,640][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:51:14,144][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:51:14,647][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:51:15,149][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:51:15,653][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:51:16,155][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:51:16,665][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:51:17,169][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:51:17,671][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:51:18,173][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:51:18,676][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:51:19,180][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:51:19,682][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:51:20,184][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:51:20,689][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:51:21,190][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:51:21,696][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:51:22,195][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:51:22,700][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:51:23,204][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:51:23,709][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:51:24,209][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:51:24,709][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:51:25,206][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:51:25,707][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:51:26,205][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:51:26,701][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:51:27,202][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:51:27,700][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:51:28,199][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:51:28,697][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10478 tokens. -[2025-11-09 21:51:29,396][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 55.68%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 -[2025-11-09 21:51:30,146][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:51:30,148][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:51:30,149][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:51:31,147][__main__][INFO] - Iteration 268 took 1m 10s (48.43% Gen, 50.15% Train). Generation: 34s, Training: 35s. Estimated remaining time: 54h 39m 51s. Estimated total time: 58h 32m 26s. Time estimates for 10 more iterations: 11m 42s, 100 more iterations: 1h 57m 4s, 500 more iterations: 9h 45m 24s. -[2025-11-09 21:51:31,149][__main__][INFO] - Starting iteration 268. -[2025-11-09 21:51:31,611][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 26 and human policies 1. -[2025-11-09 21:51:31,611][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:51:49,164][__main__][INFO] - Number of regex retries in iteration 268: 0 -[2025-11-09 21:51:49,164][__main__][INFO] - agents played in iteration 268 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:51:49,958][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:51:49,981][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:51:50,004][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:51:50,026][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:51:50,027][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:51:50,028][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:51:50,683][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:51:51,143][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:51:51,644][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:51:52,147][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:51:52,646][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:51:53,149][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:51:53,662][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:51:54,166][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:51:54,667][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:51:55,176][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:51:55,678][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:51:56,183][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:51:56,685][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:51:57,188][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:51:57,691][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:51:58,196][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:51:58,698][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:51:59,202][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:51:59,701][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:52:00,207][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:52:00,704][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:52:01,205][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:52:01,703][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:52:02,204][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:52:02,701][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:52:03,201][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:52:03,702][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:52:04,211][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:52:04,715][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:52:05,219][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:52:05,728][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:52:06,225][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:52:06,725][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:52:07,228][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:52:07,727][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:52:08,230][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:52:08,730][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:52:09,235][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:52:09,737][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:52:10,238][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:52:10,755][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:52:11,257][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:52:11,758][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:52:12,262][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:52:12,759][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:52:13,260][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:52:13,758][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:52:14,261][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:52:14,776][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:52:15,279][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:52:15,781][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:52:16,283][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:52:16,782][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:52:17,282][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:52:17,778][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:52:18,278][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:52:18,779][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:52:19,279][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:52:19,777][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:52:20,276][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:52:20,772][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:52:21,272][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:52:21,773][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:52:22,270][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:52:22,772][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10491 tokens. -[2025-11-09 21:52:23,467][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 55.69%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:32 -[2025-11-09 21:52:24,229][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:52:24,230][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:52:24,232][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:52:25,256][__main__][INFO] - Iteration 269 took 53s (32.72% Gen, 65.37% Train). Generation: 17s, Training: 35s. Estimated remaining time: 40h 48m 48s. Estimated total time: 44h 42m 16s. Time estimates for 10 more iterations: 8m 56s, 100 more iterations: 1h 29m 24s, 500 more iterations: 7h 27m 2s. -[2025-11-09 21:52:25,258][__main__][INFO] - Starting iteration 269. -[2025-11-09 21:52:25,722][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 26 and human policies 1. -[2025-11-09 21:52:25,723][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:52:45,504][__main__][INFO] - Number of regex retries in iteration 269: 0 -[2025-11-09 21:52:45,504][__main__][INFO] - agents played in iteration 269 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:52:46,304][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:52:46,330][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:52:46,355][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:52:46,381][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:52:46,381][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:52:46,383][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:52:47,076][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:52:47,545][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:52:48,054][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:52:48,553][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:52:49,056][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:52:49,554][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:52:50,056][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:52:50,555][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:52:51,058][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:52:51,573][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:52:52,074][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:52:52,575][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:52:53,073][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:52:53,574][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:52:54,083][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:52:54,716][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:52:56,367][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:52:56,874][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:52:57,376][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:52:57,877][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:52:58,378][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:52:58,881][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:52:59,396][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:52:59,900][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:53:00,409][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:53:00,911][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:53:01,412][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:53:01,922][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:53:02,431][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:53:02,936][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:53:03,451][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:53:03,950][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:53:04,453][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:53:04,955][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:53:05,480][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:53:06,000][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:53:06,508][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:53:07,016][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:53:07,516][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:53:08,018][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:53:08,524][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:53:09,027][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:53:09,534][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:53:10,038][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:53:10,544][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:53:11,048][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:53:11,550][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:53:12,054][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:53:12,556][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:53:13,056][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:53:13,560][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:53:14,061][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:53:14,564][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:53:15,065][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:53:15,566][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:53:16,066][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:53:16,568][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:53:17,071][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:53:17,594][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:53:18,094][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:53:18,599][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:53:19,100][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:53:19,604][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:53:20,104][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:53:20,606][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10486 tokens. -[2025-11-09 21:53:21,355][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 55.70%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:34 -[2025-11-09 21:53:21,993][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:53:21,995][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:53:21,997][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:53:22,902][__main__][INFO] - Iteration 270 took 57s (34.59% Gen, 63.82% Train). Generation: 19s, Training: 36s. Estimated remaining time: 43h 44m 35s. Estimated total time: 47h 39m 1s. Time estimates for 10 more iterations: 9m 31s, 100 more iterations: 1h 35m 18s, 500 more iterations: 7h 56m 30s. -[2025-11-09 21:53:22,905][__main__][INFO] - Starting iteration 270. -[2025-11-09 21:53:23,370][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 26 and human policies 1. -[2025-11-09 21:53:23,370][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:53:55,341][__main__][INFO] - Number of regex retries in iteration 270: 0 -[2025-11-09 21:53:55,341][__main__][INFO] - agents played in iteration 270 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:53:56,148][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:53:56,171][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:53:56,194][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:53:56,216][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:53:56,218][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:53:56,219][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:53:56,923][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:53:57,383][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:53:57,891][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:53:58,393][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:53:58,896][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:53:59,402][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:53:59,899][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:54:00,401][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:54:00,906][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:54:01,409][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:54:01,913][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:54:02,417][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:54:02,924][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:54:03,433][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:54:03,933][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:54:04,432][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:54:04,933][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:54:05,432][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:54:05,938][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:54:06,439][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:54:06,939][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:54:07,440][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:54:07,937][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:54:08,439][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:54:08,941][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:54:09,441][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:54:09,943][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:54:10,439][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:54:10,940][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:54:11,445][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:54:11,944][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:54:12,452][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:54:12,952][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:54:13,453][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:54:13,956][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:54:14,454][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:54:14,957][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:54:15,459][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:54:15,957][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:54:16,471][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:54:16,974][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:54:17,472][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:54:17,982][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:54:18,484][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:54:18,989][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:54:19,492][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:54:19,992][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:54:20,504][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:54:21,002][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:54:21,498][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:54:21,995][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:54:22,493][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:54:22,998][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:54:23,498][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:54:23,997][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:54:24,508][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:54:25,007][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:54:25,510][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:54:26,005][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:54:26,504][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:54:27,009][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:54:27,510][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:54:28,012][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:54:28,514][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:54:29,014][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10473 tokens. -[2025-11-09 21:54:29,711][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 55.69%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 -[2025-11-09 21:54:30,438][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:54:30,441][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:54:30,442][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:54:32,411][__main__][INFO] - Iteration 271 took 1m 9s (46.31% Gen, 50.84% Train). Generation: 31s, Training: 35s. Estimated remaining time: 53h 36m 29s. Estimated total time: 57h 32m 5s. Time estimates for 10 more iterations: 11m 30s, 100 more iterations: 1h 55m 4s, 500 more iterations: 9h 35m 20s. -[2025-11-09 21:54:32,413][__main__][INFO] - Starting iteration 271. -[2025-11-09 21:54:32,878][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 27 and human policies 1. -[2025-11-09 21:54:32,878][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:54:51,864][__main__][INFO] - Number of regex retries in iteration 271: 0 -[2025-11-09 21:54:51,865][__main__][INFO] - agents played in iteration 271 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:54:52,666][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:54:52,693][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:54:52,715][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:54:52,738][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:54:52,739][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:54:52,740][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:54:53,470][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:54:53,930][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:54:54,435][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:54:54,938][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:54:55,439][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:54:55,946][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:54:56,452][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:54:56,954][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:54:57,457][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:54:57,961][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:54:58,463][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:54:58,962][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:54:59,465][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:54:59,970][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:55:00,474][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:55:00,973][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:55:01,486][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:55:01,986][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:55:02,503][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:55:03,011][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:55:03,516][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:55:04,026][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:55:04,527][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:55:05,032][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:55:05,539][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:55:06,044][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:55:06,549][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:55:07,051][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:55:07,554][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:55:08,056][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:55:08,557][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:55:09,066][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:55:09,568][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:55:10,071][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:55:10,576][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:55:11,077][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:55:11,578][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:55:12,080][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:55:12,581][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:55:13,085][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:55:13,585][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:55:14,087][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:55:14,588][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:55:15,090][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:55:15,605][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:55:16,103][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:55:16,602][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:55:17,102][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:55:17,604][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:55:18,108][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:55:18,606][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:55:19,108][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:55:19,611][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:55:20,109][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:55:20,609][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:55:21,109][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:55:21,616][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:55:22,121][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:55:22,621][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:55:23,120][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:55:23,618][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:55:24,118][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:55:24,624][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:55:25,125][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:55:25,625][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10478 tokens. -[2025-11-09 21:55:26,320][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 55.67%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 -[2025-11-09 21:55:27,041][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:55:27,043][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:55:27,045][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:55:28,153][__main__][INFO] - Iteration 272 took 55s (34.35% Gen, 63.65% Train). Generation: 18s, Training: 35s. Estimated remaining time: 42h 7m 15s. Estimated total time: 46h 3m 46s. Time estimates for 10 more iterations: 9m 12s, 100 more iterations: 1h 32m 7s, 500 more iterations: 7h 40m 37s. -[2025-11-09 21:55:28,155][__main__][INFO] - Starting iteration 272. -[2025-11-09 21:55:28,619][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 27 and human policies 1. -[2025-11-09 21:55:28,619][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:55:40,178][mllm.models.large_language_model_local][WARNING] - Response Propposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 21:55:50,027][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 1 b, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 21:55:53,880][__main__][INFO] - Number of regex retries in iteration 272: 2 -[2025-11-09 21:55:53,881][__main__][INFO] - agents played in iteration 272 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:55:54,734][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:55:54,769][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:55:54,794][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:55:54,816][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:55:54,817][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:55:54,817][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:55:55,565][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:55:56,029][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:55:56,538][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:55:57,040][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:55:57,543][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:55:58,049][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:55:58,553][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:55:59,057][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:55:59,562][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:56:00,068][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:56:00,569][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:56:01,076][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:56:01,579][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:56:02,084][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:56:02,590][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:56:03,095][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:56:03,600][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:56:04,107][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:56:04,611][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:56:05,115][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:56:05,617][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:56:06,120][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:56:06,624][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:56:07,128][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:56:07,637][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:56:08,139][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:56:08,637][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:56:09,139][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:56:09,638][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:56:10,138][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:56:10,635][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:56:11,135][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:56:11,635][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:56:12,138][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:56:12,638][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:56:13,142][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:56:13,642][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:56:14,146][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:56:14,650][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:56:15,156][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:56:15,657][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:56:16,157][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:56:16,656][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:56:17,153][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:56:17,656][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:56:18,155][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:56:18,653][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:56:19,153][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:56:19,651][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:56:20,153][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:56:20,654][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:56:21,154][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:56:21,654][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:56:22,160][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:56:22,658][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:56:23,155][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:56:23,654][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:56:24,152][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:56:24,649][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:56:25,150][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:56:25,651][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:56:26,147][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:56:26,647][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:56:27,147][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:56:27,651][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10383 tokens. -[2025-11-09 21:56:28,342][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 55.66%, Block Peak % of device VRAM: 62.38%, ΔTime: 00:00:32 -[2025-11-09 21:56:28,979][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:56:28,980][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:56:28,982][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:56:29,867][__main__][INFO] - Iteration 273 took 1m 1s (41.25% Gen, 57.31% Train). Generation: 25s, Training: 35s. Estimated remaining time: 47h 4m 53s. Estimated total time: 51h 2m 26s. Time estimates for 10 more iterations: 10m 12s, 100 more iterations: 1h 42m 4s, 500 more iterations: 8h 30m 24s. -[2025-11-09 21:56:29,869][__main__][INFO] - Starting iteration 273. -[2025-11-09 21:56:30,355][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 27 and human policies 1. -[2025-11-09 21:56:30,356][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:56:59,801][__main__][INFO] - Number of regex retries in iteration 273: 0 -[2025-11-09 21:56:59,801][__main__][INFO] - agents played in iteration 273 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:57:00,616][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:57:00,640][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:57:00,663][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:57:00,690][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:57:00,691][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:57:00,692][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:57:01,451][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:57:01,912][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:57:02,420][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:57:02,926][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:57:03,428][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:57:03,931][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:57:04,433][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:57:04,947][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:57:05,448][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:57:05,952][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:57:06,458][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:57:06,965][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:57:07,468][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:57:07,971][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:57:08,474][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:57:08,981][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:57:09,484][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:57:09,988][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:57:10,489][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:57:10,989][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:57:11,493][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:57:11,993][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:57:12,494][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:57:12,992][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:57:13,492][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:57:13,992][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:57:14,490][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:57:14,989][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:57:15,491][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:57:15,989][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:57:16,487][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:57:16,988][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:57:17,494][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:57:18,000][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:57:18,505][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:57:19,007][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:57:19,521][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:57:20,020][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:57:20,530][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:57:21,029][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:57:21,526][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:57:22,028][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:57:22,529][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:57:23,028][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:57:23,528][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:57:24,029][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:57:24,544][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:57:25,044][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:57:25,542][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:57:26,038][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:57:26,533][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:57:27,030][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:57:27,530][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:57:28,031][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:57:28,529][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:57:29,026][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:57:29,521][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:57:30,019][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:57:30,517][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:57:31,014][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:57:31,515][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:57:32,012][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:57:32,510][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:57:33,011][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:57:33,509][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10503 tokens. -[2025-11-09 21:57:34,175][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.26%, Current % of VRAM taken: 55.76%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 -[2025-11-09 21:57:34,888][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:57:34,890][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:57:34,891][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:57:35,870][__main__][INFO] - Iteration 274 took 1m 5s (44.94% Gen, 53.56% Train). Generation: 29s, Training: 35s. Estimated remaining time: 50h 37m 6s. Estimated total time: 54h 35m 45s. Time estimates for 10 more iterations: 10m 55s, 100 more iterations: 1h 49m 11s, 500 more iterations: 9h 5m 57s. -[2025-11-09 21:57:35,872][__main__][INFO] - Starting iteration 274. -[2025-11-09 21:57:36,349][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 27 and human policies 1. -[2025-11-09 21:57:36,350][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:57:56,480][__main__][INFO] - Number of regex retries in iteration 274: 0 -[2025-11-09 21:57:56,481][__main__][INFO] - agents played in iteration 274 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:57:57,682][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:57:58,515][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:57:58,864][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:57:58,895][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:57:58,897][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:57:58,898][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:57:59,687][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:58:00,169][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:58:00,679][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:58:01,181][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:58:01,685][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:58:02,186][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:58:02,688][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:58:03,190][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:58:03,692][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:58:04,194][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:58:04,695][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:58:05,196][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:58:05,701][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:58:06,205][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:58:06,707][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:58:07,210][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:58:07,711][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:58:08,229][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:58:08,734][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:58:09,237][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:58:09,741][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:58:10,246][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:58:10,752][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:58:11,255][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:58:11,760][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:58:12,261][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:58:12,764][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:58:13,273][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:58:13,777][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:58:14,280][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:58:14,782][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:58:15,286][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:58:15,792][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:58:16,295][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:58:16,799][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:58:17,308][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:58:17,813][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:58:18,314][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:58:18,816][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:58:19,315][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:58:19,816][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:58:20,314][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:58:20,812][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:58:21,315][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:58:21,813][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:58:22,313][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:58:22,812][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:58:23,310][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:58:23,809][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:58:24,332][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:58:24,832][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:58:25,331][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:58:25,831][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:58:26,346][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:58:26,845][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:58:27,345][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:58:27,843][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:58:28,340][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:58:28,840][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:58:29,338][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:58:29,837][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:58:30,332][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:58:30,831][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:58:31,336][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:58:31,841][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10521 tokens. -[2025-11-09 21:58:32,507][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 55.72%, Block Peak % of device VRAM: 62.35%, ΔTime: 00:00:32 -[2025-11-09 21:58:33,128][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:58:33,130][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:58:33,131][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:58:34,039][__main__][INFO] - Iteration 275 took 57s (34.90% Gen, 63.53% Train). Generation: 20s, Training: 36s. Estimated remaining time: 44h 4m 53s. Estimated total time: 48h 4m 30s. Time estimates for 10 more iterations: 9m 36s, 100 more iterations: 1h 36m 9s, 500 more iterations: 8h 0m 45s. -[2025-11-09 21:58:34,041][__main__][INFO] - Starting iteration 275. -[2025-11-09 21:58:34,509][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 27 and human policies 1. -[2025-11-09 21:58:34,509][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:59:02,117][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 21:59:04,711][__main__][INFO] - Number of regex retries in iteration 275: 1 -[2025-11-09 21:59:04,712][__main__][INFO] - agents played in iteration 275 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 21:59:05,520][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:59:05,549][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:59:05,577][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:59:05,600][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 21:59:05,601][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 21:59:05,602][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 21:59:06,296][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 21:59:06,754][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 21:59:07,263][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 21:59:07,764][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 21:59:08,270][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 21:59:08,778][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 21:59:09,280][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 21:59:09,790][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 21:59:10,295][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 21:59:10,803][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 21:59:11,305][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 21:59:11,805][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 21:59:12,312][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 21:59:12,819][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 21:59:13,321][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 21:59:13,838][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 21:59:14,345][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 21:59:14,847][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 21:59:15,354][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 21:59:15,857][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 21:59:16,366][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 21:59:16,869][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 21:59:17,370][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 21:59:17,873][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 21:59:18,370][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 21:59:18,871][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 21:59:19,371][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 21:59:19,873][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 21:59:20,384][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 21:59:20,887][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 21:59:21,390][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 21:59:21,893][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 21:59:22,394][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 21:59:22,891][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 21:59:23,390][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 21:59:23,887][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 21:59:24,388][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 21:59:24,888][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 21:59:25,383][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 21:59:25,885][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 21:59:26,385][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 21:59:26,883][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 21:59:27,383][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 21:59:27,881][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 21:59:28,381][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 21:59:28,884][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 21:59:29,379][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 21:59:29,876][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 21:59:30,374][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 21:59:30,874][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 21:59:31,372][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 21:59:31,870][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 21:59:32,368][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 21:59:32,867][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 21:59:33,364][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 21:59:33,865][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 21:59:34,362][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 21:59:34,860][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 21:59:35,356][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 21:59:35,851][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 21:59:36,347][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 21:59:36,841][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 21:59:37,338][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 21:59:37,844][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 21:59:38,343][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10577 tokens. -[2025-11-09 21:59:39,017][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.54%, ΔTime: 00:00:32 -[2025-11-09 21:59:39,741][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 21:59:39,743][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 21:59:39,745][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 21:59:40,745][__main__][INFO] - Iteration 276 took 1m 6s (45.60% Gen, 52.89% Train). Generation: 30s, Training: 35s. Estimated remaining time: 51h 11m 8s. Estimated total time: 55h 11m 52s. Time estimates for 10 more iterations: 11m 2s, 100 more iterations: 1h 50m 23s, 500 more iterations: 9h 11m 58s. -[2025-11-09 21:59:40,747][__main__][INFO] - Starting iteration 276. -[2025-11-09 21:59:41,211][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 27 and human policies 1. -[2025-11-09 21:59:41,212][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 21:59:51,206][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 21:59:59,860][__main__][INFO] - Number of regex retries in iteration 276: 1 -[2025-11-09 21:59:59,861][__main__][INFO] - agents played in iteration 276 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:00:00,664][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:00:00,688][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:00:00,711][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:00:00,734][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:00:00,734][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:00:00,735][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:00:01,495][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:00:01,961][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:00:02,479][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:00:02,985][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:00:03,487][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:00:04,000][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:00:04,500][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:00:05,002][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:00:05,504][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:00:06,007][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:00:06,514][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:00:07,015][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:00:07,516][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:00:08,020][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:00:08,522][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:00:09,025][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:00:09,527][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:00:10,026][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:00:10,524][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:00:11,023][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:00:11,519][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:00:12,020][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:00:12,519][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:00:13,013][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:00:13,514][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:00:14,017][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:00:14,522][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:00:15,021][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:00:15,521][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:00:16,027][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:00:16,527][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:00:17,028][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:00:17,527][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:00:18,030][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:00:18,534][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:00:19,035][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:00:19,536][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:00:20,037][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:00:20,536][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:00:21,037][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:00:21,538][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:00:22,035][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:00:22,540][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:00:23,041][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:00:23,541][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:00:24,044][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:00:24,542][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:00:25,038][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:00:25,534][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:00:26,033][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:00:26,536][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:00:27,032][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:00:27,532][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:00:28,031][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:00:28,529][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:00:29,031][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:00:29,528][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:00:30,028][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:00:30,526][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:00:31,023][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:00:31,524][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:00:32,022][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:00:32,522][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:00:33,022][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:00:33,521][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10572 tokens. -[2025-11-09 22:00:34,212][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 55.72%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 -[2025-11-09 22:00:34,947][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:00:34,949][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:00:34,950][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:00:35,953][__main__][INFO] - Iteration 277 took 54s (34.07% Gen, 64.10% Train). Generation: 18s, Training: 35s. Estimated remaining time: 41h 35m 28s. Estimated total time: 45h 37m 7s. Time estimates for 10 more iterations: 9m 7s, 100 more iterations: 1h 31m 14s, 500 more iterations: 7h 36m 11s. -[2025-11-09 22:00:35,955][__main__][INFO] - Starting iteration 277. -[2025-11-09 22:00:36,414][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 27 and human policies 1. -[2025-11-09 22:00:36,414][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:00:56,606][__main__][INFO] - Number of regex retries in iteration 277: 0 -[2025-11-09 22:00:56,607][__main__][INFO] - agents played in iteration 277 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:00:57,507][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:00:57,541][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:00:57,571][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:00:57,596][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:00:57,597][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:00:57,597][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:00:58,363][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:00:58,826][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:00:59,345][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:00:59,847][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:01:00,350][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:01:00,856][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:01:01,359][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:01:01,865][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:01:02,371][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:01:02,876][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:01:03,382][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:01:03,887][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:01:04,395][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:01:04,899][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:01:05,403][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:01:05,912][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:01:06,416][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:01:06,918][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:01:07,419][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:01:07,918][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:01:08,419][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:01:08,920][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:01:09,422][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:01:09,919][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:01:10,418][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:01:10,918][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:01:11,417][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:01:11,916][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:01:12,420][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:01:12,921][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:01:13,417][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:01:13,922][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:01:14,426][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:01:14,925][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:01:15,427][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:01:15,923][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:01:16,425][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:01:16,925][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:01:17,427][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:01:17,933][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:01:18,433][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:01:18,930][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:01:19,429][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:01:19,930][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:01:20,434][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:01:20,934][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:01:21,435][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:01:21,939][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:01:22,441][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:01:22,942][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:01:23,441][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:01:23,940][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:01:24,446][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:01:24,950][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:01:25,454][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:01:25,959][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:01:26,462][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:01:26,961][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:01:27,461][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:01:27,958][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:01:28,458][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:01:28,956][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:01:29,455][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:01:29,953][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:01:30,451][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10577 tokens. -[2025-11-09 22:01:31,131][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.26%, Current % of VRAM taken: 55.76%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:32 -[2025-11-09 22:01:31,762][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:01:31,764][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:01:31,766][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:01:32,683][__main__][INFO] - Iteration 278 took 56s (35.89% Gen, 62.48% Train). Generation: 20s, Training: 35s. Estimated remaining time: 42h 50m 53s. Estimated total time: 46h 53m 28s. Time estimates for 10 more iterations: 9m 22s, 100 more iterations: 1h 33m 46s, 500 more iterations: 7h 48m 54s. -[2025-11-09 22:01:32,685][__main__][INFO] - Starting iteration 278. -[2025-11-09 22:01:33,161][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 27 and human policies 1. -[2025-11-09 22:01:33,162][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:02:01,918][__main__][INFO] - Number of regex retries in iteration 278: 0 -[2025-11-09 22:02:01,918][__main__][INFO] - agents played in iteration 278 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:02:02,739][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:02:02,762][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:02:02,785][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:02:02,808][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:02:02,809][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:02:02,810][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:02:03,587][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:02:04,046][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:02:04,557][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:02:05,065][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:02:05,570][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:02:06,072][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:02:06,578][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:02:07,084][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:02:07,591][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:02:08,092][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:02:08,593][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:02:09,109][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:02:09,617][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:02:10,126][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:02:10,626][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:02:11,124][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:02:11,642][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:02:12,142][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:02:12,639][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:02:13,138][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:02:13,637][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:02:14,134][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:02:14,632][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:02:15,132][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:02:15,631][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:02:16,131][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:02:16,633][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:02:17,136][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:02:17,636][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:02:18,135][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:02:18,634][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:02:19,136][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:02:19,639][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:02:20,142][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:02:20,645][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:02:21,145][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:02:21,648][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:02:22,153][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:02:22,654][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:02:23,155][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:02:23,658][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:02:24,160][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:02:24,658][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:02:25,157][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:02:25,653][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:02:26,152][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:02:26,650][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:02:27,152][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:02:27,650][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:02:28,144][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:02:28,647][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:02:29,146][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:02:29,641][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:02:30,136][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:02:30,631][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:02:31,130][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:02:31,626][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:02:32,124][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:02:32,634][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:02:33,134][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:02:33,632][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:02:34,128][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:02:34,625][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:02:35,122][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:02:35,619][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10597 tokens. -[2025-11-09 22:02:36,292][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 55.70%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:32 -[2025-11-09 22:02:37,048][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:02:37,050][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:02:37,052][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:02:38,026][__main__][INFO] - Iteration 279 took 1m 4s (44.33% Gen, 54.16% Train). Generation: 28s, Training: 35s. Estimated remaining time: 49h 59m 34s. Estimated total time: 54h 3m 15s. Time estimates for 10 more iterations: 10m 48s, 100 more iterations: 1h 48m 6s, 500 more iterations: 9h 0m 32s. -[2025-11-09 22:02:38,028][__main__][INFO] - Starting iteration 279. -[2025-11-09 22:02:38,495][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 27 and human policies 1. -[2025-11-09 22:02:38,495][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:02:59,070][__main__][INFO] - Number of regex retries in iteration 279: 0 -[2025-11-09 22:02:59,070][__main__][INFO] - agents played in iteration 279 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:02:59,871][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:02:59,896][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:02:59,921][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:02:59,945][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:02:59,945][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:02:59,946][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:03:01,827][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:03:02,409][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:03:03,013][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:03:03,520][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:03:04,027][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:03:04,535][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:03:05,035][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:03:05,558][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:03:06,060][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:03:06,565][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:03:07,068][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:03:07,575][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:03:08,077][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:03:08,579][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:03:09,084][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:03:09,585][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:03:10,090][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:03:10,594][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:03:11,099][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:03:11,603][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:03:12,109][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:03:12,614][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:03:13,118][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:03:13,633][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:03:14,138][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:03:14,651][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:03:15,151][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:03:15,656][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:03:16,154][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:03:16,655][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:03:17,157][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:03:17,659][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:03:18,158][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:03:18,657][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:03:19,159][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:03:19,661][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:03:20,158][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:03:20,656][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:03:21,157][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:03:21,655][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:03:22,157][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:03:22,664][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:03:23,159][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:03:23,656][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:03:24,153][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:03:24,652][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:03:25,162][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:03:25,662][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:03:26,161][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:03:26,657][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:03:27,155][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:03:27,657][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:03:28,157][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:03:28,656][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:03:29,164][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:03:29,662][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:03:30,161][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:03:30,661][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:03:31,164][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:03:31,669][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:03:32,170][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:03:32,664][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:03:33,163][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:03:33,660][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:03:34,160][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10537 tokens. -[2025-11-09 22:03:34,856][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 55.69%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:33 -[2025-11-09 22:03:35,509][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:03:35,510][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:03:35,512][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:03:36,420][__main__][INFO] - Iteration 280 took 57s (35.52% Gen, 62.91% Train). Generation: 20s, Training: 36s. Estimated remaining time: 44h 11m 38s. Estimated total time: 48h 16m 17s. Time estimates for 10 more iterations: 9m 39s, 100 more iterations: 1h 36m 32s, 500 more iterations: 8h 2m 42s. -[2025-11-09 22:03:36,423][__main__][INFO] - Starting iteration 280. -[2025-11-09 22:03:36,902][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 27 and human policies 1. -[2025-11-09 22:03:36,902][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:04:09,932][__main__][INFO] - Number of regex retries in iteration 280: 0 -[2025-11-09 22:04:09,933][__main__][INFO] - agents played in iteration 280 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:04:10,740][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:04:10,766][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:04:10,791][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:04:10,816][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:04:10,817][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:04:10,818][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:04:11,571][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:04:12,032][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:04:12,540][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:04:13,044][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:04:13,548][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:04:14,064][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:04:14,571][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:04:15,075][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:04:15,581][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:04:16,082][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:04:16,588][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:04:17,092][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:04:17,593][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:04:18,093][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:04:18,596][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:04:19,104][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:04:19,608][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:04:20,106][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:04:20,609][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:04:21,112][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:04:21,615][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:04:22,117][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:04:22,621][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:04:23,134][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:04:23,630][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:04:24,130][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:04:24,637][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:04:25,135][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:04:25,639][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:04:26,133][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:04:26,633][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:04:27,146][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:04:27,640][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:04:28,143][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:04:28,643][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:04:29,145][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:04:29,665][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:04:30,165][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:04:30,666][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:04:31,168][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:04:31,671][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:04:32,169][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:04:32,668][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:04:33,170][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:04:33,672][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:04:34,170][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:04:34,665][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:04:35,162][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:04:35,658][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:04:36,154][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:04:36,654][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:04:37,153][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:04:37,668][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:04:38,168][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:04:38,668][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:04:39,163][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:04:39,659][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:04:40,163][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:04:40,660][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:04:41,155][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:04:41,664][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:04:42,158][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:04:42,654][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:04:43,152][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:04:43,648][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10482 tokens. -[2025-11-09 22:04:44,346][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 55.65%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 -[2025-11-09 22:04:45,077][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:04:45,078][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:04:45,080][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:04:47,066][__main__][INFO] - Iteration 281 took 1m 10s (47.07% Gen, 50.09% Train). Generation: 33s, Training: 35s. Estimated remaining time: 54h 22m 24s. Estimated total time: 58h 28m 14s. Time estimates for 10 more iterations: 11m 41s, 100 more iterations: 1h 56m 56s, 500 more iterations: 9h 44m 42s. -[2025-11-09 22:04:47,068][__main__][INFO] - Starting iteration 281. -[2025-11-09 22:04:47,557][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 28 and human policies 1. -[2025-11-09 22:04:47,558][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:05:10,134][__main__][INFO] - Number of regex retries in iteration 281: 0 -[2025-11-09 22:05:10,136][__main__][INFO] - agents played in iteration 281 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:05:10,985][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:05:11,018][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:05:11,045][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:05:11,068][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:05:11,069][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:05:11,069][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:05:11,852][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:05:12,310][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:05:12,825][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:05:13,336][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:05:13,839][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:05:14,347][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:05:14,850][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:05:15,352][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:05:15,857][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:05:16,360][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:05:16,863][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:05:17,368][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:05:17,872][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:05:18,381][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:05:18,884][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:05:19,388][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:05:19,891][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:05:20,394][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:05:20,896][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:05:21,394][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:05:21,894][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:05:22,394][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:05:22,894][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:05:23,397][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:05:23,896][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:05:24,400][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:05:24,910][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:05:25,415][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:05:25,919][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:05:26,422][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:05:26,925][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:05:27,432][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:05:27,936][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:05:28,437][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:05:28,938][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:05:29,442][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:05:29,942][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:05:30,441][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:05:30,942][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:05:31,443][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:05:31,942][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:05:32,442][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:05:32,938][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:05:33,437][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:05:33,936][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:05:34,431][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:05:34,934][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:05:35,433][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:05:35,931][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:05:36,432][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:05:36,932][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:05:37,430][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:05:37,930][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:05:38,429][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:05:38,930][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:05:39,429][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:05:39,930][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:05:40,429][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:05:40,927][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:05:41,423][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:05:41,920][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:05:42,417][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:05:42,917][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:05:43,421][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:05:43,926][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10474 tokens. -[2025-11-09 22:05:44,627][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 55.67%, Block Peak % of device VRAM: 62.42%, ΔTime: 00:00:32 -[2025-11-09 22:05:45,257][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:05:45,258][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:05:45,260][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:05:46,156][__main__][INFO] - Iteration 282 took 58s (38.53% Gen, 59.94% Train). Generation: 22s, Training: 35s. Estimated remaining time: 44h 43m 9s. Estimated total time: 48h 49m 58s. Time estimates for 10 more iterations: 9m 45s, 100 more iterations: 1h 37m 39s, 500 more iterations: 8h 8m 19s. -[2025-11-09 22:05:46,158][__main__][INFO] - Starting iteration 282. -[2025-11-09 22:05:46,622][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 28 and human policies 1. -[2025-11-09 22:05:46,623][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:06:15,175][__main__][INFO] - Number of regex retries in iteration 282: 0 -[2025-11-09 22:06:15,176][__main__][INFO] - agents played in iteration 282 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:06:15,989][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:06:16,012][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:06:16,036][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:06:16,058][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:06:16,060][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:06:16,061][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:06:16,805][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:06:17,269][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:06:17,775][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:06:18,282][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:06:18,785][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:06:19,292][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:06:19,796][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:06:20,302][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:06:20,804][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:06:21,311][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:06:21,814][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:06:22,321][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:06:22,824][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:06:23,326][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:06:23,833][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:06:24,335][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:06:24,836][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:06:25,343][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:06:25,846][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:06:26,350][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:06:26,854][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:06:27,355][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:06:27,857][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:06:28,362][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:06:28,859][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:06:29,363][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:06:29,860][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:06:30,373][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:06:30,874][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:06:31,375][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:06:31,881][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:06:32,381][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:06:32,893][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:06:33,391][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:06:33,890][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:06:34,390][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:06:34,886][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:06:35,381][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:06:35,882][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:06:36,380][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:06:36,888][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:06:37,389][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:06:37,884][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:06:38,381][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:06:38,876][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:06:39,376][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:06:39,879][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:06:40,379][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:06:40,891][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:06:41,391][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:06:41,890][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:06:42,409][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:06:42,913][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:06:43,418][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:06:43,920][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:06:44,425][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:06:44,925][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:06:45,422][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:06:45,921][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:06:46,416][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:06:46,914][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:06:47,415][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:06:47,912][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:06:48,409][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:06:48,912][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10587 tokens. -[2025-11-09 22:06:49,577][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 55.68%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:32 -[2025-11-09 22:06:50,332][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:06:50,334][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:06:50,335][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:06:51,351][__main__][INFO] - Iteration 283 took 1m 4s (44.11% Gen, 54.32% Train). Generation: 28s, Training: 35s. Estimated remaining time: 49h 48m 34s. Estimated total time: 53h 56m 29s. Time estimates for 10 more iterations: 10m 47s, 100 more iterations: 1h 47m 52s, 500 more iterations: 8h 59m 24s. -[2025-11-09 22:06:51,353][__main__][INFO] - Starting iteration 283. -[2025-11-09 22:06:51,812][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 28 and human policies 1. -[2025-11-09 22:06:51,813][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:07:02,314][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 22:07:15,721][__main__][INFO] - Number of regex retries in iteration 283: 1 -[2025-11-09 22:07:15,723][__main__][INFO] - agents played in iteration 283 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:07:16,572][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:07:16,614][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:07:16,645][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:07:16,673][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:07:16,673][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:07:16,675][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:07:17,441][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:07:17,911][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:07:18,423][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:07:18,934][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:07:19,438][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:07:19,942][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:07:20,446][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:07:20,954][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:07:21,459][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:07:21,965][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:07:22,465][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:07:22,969][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:07:23,474][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:07:23,977][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:07:24,497][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:07:25,000][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:07:25,502][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:07:26,004][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:07:26,505][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:07:27,011][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:07:27,514][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:07:28,015][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:07:28,521][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:07:29,022][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:07:29,526][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:07:30,031][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:07:30,536][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:07:31,040][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:07:31,545][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:07:32,051][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:07:32,552][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:07:33,055][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:07:33,559][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:07:34,061][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:07:34,563][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:07:35,073][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:07:35,577][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:07:36,085][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:07:36,590][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:07:37,092][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:07:37,610][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:07:38,113][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:07:38,613][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:07:39,112][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:07:39,611][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:07:40,115][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:07:40,616][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:07:41,114][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:07:41,613][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:07:42,109][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:07:42,608][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:07:43,110][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:07:43,609][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:07:44,113][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:07:44,612][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:07:45,111][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:07:45,610][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:07:46,106][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:07:46,605][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:07:47,105][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:07:47,601][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:07:48,109][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:07:48,606][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:07:49,106][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:07:49,602][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10554 tokens. -[2025-11-09 22:07:50,294][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 55.66%, Block Peak % of device VRAM: 62.38%, ΔTime: 00:00:32 -[2025-11-09 22:07:50,938][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:07:50,939][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:07:50,941][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:07:51,854][__main__][INFO] - Iteration 284 took 1m 0s (39.82% Gen, 58.66% Train). Generation: 23s, Training: 35s. Estimated remaining time: 45h 53m 12s. Estimated total time: 50h 2m 7s. Time estimates for 10 more iterations: 10m 0s, 100 more iterations: 1h 40m 4s, 500 more iterations: 8h 20m 21s. -[2025-11-09 22:07:51,856][__main__][INFO] - Starting iteration 284. -[2025-11-09 22:07:52,325][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 28 and human policies 1. -[2025-11-09 22:07:52,326][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:08:19,721][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 22:08:25,243][__main__][INFO] - Number of regex retries in iteration 284: 1 -[2025-11-09 22:08:25,244][__main__][INFO] - agents played in iteration 284 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:08:26,068][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:08:26,093][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:08:26,118][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:08:26,140][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:08:26,141][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:08:26,142][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:08:26,875][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:08:27,331][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:08:27,844][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:08:28,348][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:08:28,851][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:08:29,354][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:08:29,857][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:08:30,358][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:08:30,863][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:08:31,368][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:08:31,875][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:08:32,379][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:08:32,884][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:08:33,387][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:08:33,905][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:08:34,412][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:08:34,938][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:08:35,437][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:08:35,936][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:08:36,446][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:08:36,949][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:08:37,452][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:08:37,957][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:08:38,456][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:08:38,955][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:08:39,455][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:08:39,957][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:08:40,462][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:08:40,961][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:08:41,460][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:08:41,965][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:08:42,466][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:08:42,965][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:08:43,464][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:08:43,959][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:08:44,458][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:08:44,954][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:08:45,454][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:08:45,956][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:08:46,453][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:08:46,949][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:08:47,450][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:08:47,948][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:08:48,447][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:08:48,949][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:08:49,449][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:08:49,948][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:08:50,446][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:08:50,950][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:08:51,448][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:08:51,942][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:08:52,444][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:08:52,945][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:08:53,450][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:08:53,953][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:08:54,451][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:08:54,951][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:08:55,451][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:08:55,946][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:08:56,442][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:08:56,942][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:08:57,444][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:08:57,944][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:08:58,437][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:08:58,933][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10533 tokens. -[2025-11-09 22:08:59,608][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 55.67%, Block Peak % of device VRAM: 62.49%, ΔTime: 00:00:32 -[2025-11-09 22:09:00,337][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:09:00,339][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:09:00,341][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:09:01,341][__main__][INFO] - Iteration 285 took 1m 9s (47.70% Gen, 50.85% Train). Generation: 32s, Training: 35s. Estimated remaining time: 53h 20m 45s. Estimated total time: 57h 30m 50s. Time estimates for 10 more iterations: 11m 30s, 100 more iterations: 1h 55m 1s, 500 more iterations: 9h 35m 8s. -[2025-11-09 22:09:01,343][__main__][INFO] - Starting iteration 285. -[2025-11-09 22:09:01,808][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 28 and human policies 1. -[2025-11-09 22:09:01,809][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:09:23,118][__main__][INFO] - Number of regex retries in iteration 285: 0 -[2025-11-09 22:09:23,119][__main__][INFO] - agents played in iteration 285 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:09:23,941][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:09:23,979][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:09:24,010][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:09:24,037][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:09:24,038][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:09:24,039][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:09:24,797][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:09:25,258][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:09:25,763][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:09:26,269][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:09:26,771][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:09:27,278][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:09:27,779][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:09:28,288][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:09:28,793][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:09:29,294][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:09:29,796][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:09:30,298][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:09:30,801][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:09:31,321][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:09:31,824][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:09:32,327][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:09:32,830][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:09:33,337][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:09:33,842][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:09:34,346][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:09:34,848][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:09:35,356][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:09:35,858][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:09:36,365][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:09:36,871][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:09:37,377][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:09:37,884][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:09:38,391][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:09:38,892][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:09:39,405][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:09:39,902][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:09:40,419][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:09:40,920][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:09:41,423][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:09:41,927][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:09:42,430][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:09:42,938][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:09:43,439][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:09:43,942][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:09:44,443][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:09:44,942][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:09:45,446][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:09:45,946][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:09:46,448][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:09:46,953][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:09:47,456][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:09:47,965][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:09:48,474][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:09:48,980][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:09:49,480][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:09:49,983][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:09:50,480][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:09:50,982][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:09:51,482][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:09:51,986][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:09:52,494][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:09:52,990][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:09:53,489][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:09:53,985][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:09:54,481][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:09:54,984][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:09:55,484][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:09:55,984][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:09:56,481][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:09:56,981][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10625 tokens. -[2025-11-09 22:09:57,678][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 55.69%, Block Peak % of device VRAM: 62.50%, ΔTime: 00:00:32 -[2025-11-09 22:09:58,302][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:09:58,304][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:09:58,306][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:09:59,208][__main__][INFO] - Iteration 286 took 57s (37.13% Gen, 61.30% Train). Generation: 21s, Training: 35s. Estimated remaining time: 43h 38m 59s. Estimated total time: 47h 50m 1s. Time estimates for 10 more iterations: 9m 34s, 100 more iterations: 1h 35m 40s, 500 more iterations: 7h 58m 20s. -[2025-11-09 22:09:59,210][__main__][INFO] - Starting iteration 286. -[2025-11-09 22:09:59,672][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 28 and human policies 1. -[2025-11-09 22:09:59,673][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:10:30,139][__main__][INFO] - Number of regex retries in iteration 286: 0 -[2025-11-09 22:10:30,140][__main__][INFO] - agents played in iteration 286 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:10:30,917][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:10:30,944][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:10:30,971][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:10:30,994][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:10:30,994][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:10:30,995][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:10:31,684][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:10:32,139][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:10:32,643][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:10:33,143][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:10:33,642][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:10:34,140][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:10:34,640][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:10:35,140][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:10:35,638][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:10:36,141][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:10:36,642][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:10:37,151][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:10:37,653][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:10:38,154][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:10:38,655][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:10:39,158][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:10:39,661][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:10:40,167][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:10:40,670][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:10:41,175][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:10:41,674][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:10:42,180][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:10:42,682][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:10:43,185][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:10:43,690][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:10:44,212][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:10:44,716][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:10:45,221][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:10:45,723][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:10:46,243][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:10:46,744][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:10:47,243][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:10:47,742][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:10:48,241][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:10:48,738][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:10:49,236][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:10:49,733][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:10:50,230][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:10:50,724][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:10:51,222][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:10:51,723][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:10:52,225][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:10:52,739][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:10:53,236][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:10:53,737][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:10:54,236][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:10:54,734][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:10:55,234][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:10:55,736][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:10:56,234][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:10:56,749][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:10:57,244][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:10:57,742][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:10:58,241][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:10:58,739][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:10:59,245][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:10:59,742][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:11:00,241][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:11:00,753][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:11:01,251][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:11:01,751][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:11:02,252][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:11:02,754][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:11:03,252][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:11:03,749][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10735 tokens. -[2025-11-09 22:11:04,430][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 55.71%, Block Peak % of device VRAM: 62.41%, ΔTime: 00:00:32 -[2025-11-09 22:11:05,147][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:11:05,149][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:11:05,150][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:11:06,127][__main__][INFO] - Iteration 287 took 1m 6s (45.85% Gen, 52.68% Train). Generation: 30s, Training: 35s. Estimated remaining time: 51h 10m 40s. Estimated total time: 55h 22m 49s. Time estimates for 10 more iterations: 11m 4s, 100 more iterations: 1h 50m 45s, 500 more iterations: 9h 13m 48s. -[2025-11-09 22:11:06,129][__main__][INFO] - Starting iteration 287. -[2025-11-09 22:11:06,590][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 28 and human policies 1. -[2025-11-09 22:11:06,591][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:11:28,989][__main__][INFO] - Number of regex retries in iteration 287: 0 -[2025-11-09 22:11:28,991][__main__][INFO] - agents played in iteration 287 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:11:29,849][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:11:29,903][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:11:29,941][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:11:29,971][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:11:29,972][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:11:29,972][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:11:30,848][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:11:31,310][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:11:31,821][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:11:32,324][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:11:32,827][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:11:33,332][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:11:33,836][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:11:34,337][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:11:34,843][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:11:35,340][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:11:35,854][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:11:36,355][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:11:36,856][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:11:37,372][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:11:37,873][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:11:38,375][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:11:38,876][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:11:39,385][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:11:39,892][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:11:40,398][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:11:40,902][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:11:41,407][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:11:41,911][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:11:42,415][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:11:42,917][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:11:43,418][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:11:43,919][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:11:44,419][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:11:44,917][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:11:45,416][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:11:45,916][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:11:46,417][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:11:46,917][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:11:47,416][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:11:47,924][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:11:48,426][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:11:48,927][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:11:49,425][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:11:49,923][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:11:50,422][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:11:50,923][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:11:51,424][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:11:51,935][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:11:52,435][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:11:52,960][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:11:53,463][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:11:53,966][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:11:54,476][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:11:54,981][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:11:55,483][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:11:55,983][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:11:56,477][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:11:56,977][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:11:57,479][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:11:57,977][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:11:58,478][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:11:58,975][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:11:59,475][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:11:59,978][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:12:00,479][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:12:00,980][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:12:01,480][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:12:01,977][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:12:02,477][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:12:02,979][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10446 tokens. -[2025-11-09 22:12:03,702][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.35%, ΔTime: 00:00:32 -[2025-11-09 22:12:04,327][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:12:04,329][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:12:04,330][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:12:05,247][__main__][INFO] - Iteration 288 took 58s (38.19% Gen, 60.25% Train). Generation: 22s, Training: 35s. Estimated remaining time: 44h 39m 43s. Estimated total time: 48h 52m 51s. Time estimates for 10 more iterations: 9m 46s, 100 more iterations: 1h 37m 45s, 500 more iterations: 8h 8m 48s. -[2025-11-09 22:12:05,249][__main__][INFO] - Starting iteration 288. -[2025-11-09 22:12:05,710][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 28 and human policies 1. -[2025-11-09 22:12:05,711][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:12:36,793][__main__][INFO] - Number of regex retries in iteration 288: 0 -[2025-11-09 22:12:36,793][__main__][INFO] - agents played in iteration 288 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:12:37,613][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:12:37,638][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:12:37,663][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:12:37,685][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:12:37,686][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:12:37,686][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:12:38,352][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:12:38,809][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:12:39,317][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:12:39,818][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:12:40,332][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:12:40,834][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:12:41,330][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:12:41,838][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:12:42,334][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:12:42,858][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:12:43,361][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:12:43,863][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:12:44,360][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:12:44,862][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:12:45,363][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:12:45,867][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:12:46,369][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:12:46,876][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:12:47,381][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:12:47,888][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:12:48,391][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:12:48,894][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:12:49,398][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:12:49,902][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:12:50,410][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:12:50,915][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:12:51,418][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:12:51,922][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:12:52,420][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:12:52,921][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:12:53,429][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:12:53,930][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:12:54,432][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:12:54,932][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:12:55,430][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:12:55,925][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:12:56,440][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:12:56,936][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:12:57,459][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:12:57,956][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:12:58,455][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:12:58,954][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:12:59,452][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:12:59,950][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:13:00,449][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:13:00,945][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:13:01,447][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:13:01,949][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:13:02,446][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:13:02,946][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:13:03,441][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:13:03,953][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:13:04,448][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:13:04,944][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:13:05,447][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:13:05,941][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:13:06,438][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:13:06,940][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:13:07,441][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:13:07,950][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:13:08,448][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:13:08,949][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:13:09,453][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:13:09,955][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:13:10,466][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10524 tokens. -[2025-11-09 22:13:11,154][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 55.70%, Block Peak % of device VRAM: 62.41%, ΔTime: 00:00:32 -[2025-11-09 22:13:11,890][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:13:11,892][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:13:11,894][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:13:12,908][__main__][INFO] - Iteration 289 took 1m 7s (46.25% Gen, 52.23% Train). Generation: 31s, Training: 35s. Estimated remaining time: 51h 45m 39s. Estimated total time: 55h 59m 55s. Time estimates for 10 more iterations: 11m 11s, 100 more iterations: 1h 51m 59s, 500 more iterations: 9h 19m 59s. -[2025-11-09 22:13:12,910][__main__][INFO] - Starting iteration 289. -[2025-11-09 22:13:13,378][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 28 and human policies 1. -[2025-11-09 22:13:13,378][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:13:29,319][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 22:13:42,134][__main__][INFO] - Number of regex retries in iteration 289: 1 -[2025-11-09 22:13:42,135][__main__][INFO] - agents played in iteration 289 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:13:42,979][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:13:43,022][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:13:43,057][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:13:43,086][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:13:43,088][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:13:43,092][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:13:43,888][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:13:44,351][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:13:44,862][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:13:45,368][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:13:45,876][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:13:46,377][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:13:46,881][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:13:47,385][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:13:47,891][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:13:48,398][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:13:48,902][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:13:49,405][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:13:49,912][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:13:50,413][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:13:50,917][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:13:51,425][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:13:51,929][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:13:52,438][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:13:52,957][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:13:53,459][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:13:53,964][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:13:54,471][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:13:54,976][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:13:55,480][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:13:55,984][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:13:56,487][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:13:56,991][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:13:57,496][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:13:58,000][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:13:58,505][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:13:59,010][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:13:59,513][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:14:00,016][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:14:00,522][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:14:01,027][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:14:01,529][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:14:02,030][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:14:02,532][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:14:03,041][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:14:03,538][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:14:04,035][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:14:04,542][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:14:05,043][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:14:05,543][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:14:06,043][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:14:06,542][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:14:07,048][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:14:07,550][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:14:08,048][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:14:08,561][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:14:09,058][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:14:09,563][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:14:10,067][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:14:10,567][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:14:11,071][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:14:11,570][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:14:12,071][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:14:12,570][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:14:13,071][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:14:13,586][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:14:14,082][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:14:14,578][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:14:15,079][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:14:15,578][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:14:16,075][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10582 tokens. -[2025-11-09 22:14:16,781][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 55.63%, Block Peak % of device VRAM: 62.39%, ΔTime: 00:00:32 -[2025-11-09 22:14:17,426][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:14:17,427][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:14:17,429][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:14:18,343][__main__][INFO] - Iteration 290 took 1m 4s (44.26% Gen, 54.33% Train). Generation: 28s, Training: 35s. Estimated remaining time: 49h 52m 56s. Estimated total time: 54h 8m 17s. Time estimates for 10 more iterations: 10m 49s, 100 more iterations: 1h 48m 16s, 500 more iterations: 9h 1m 22s. -[2025-11-09 22:14:18,345][__main__][INFO] - Starting iteration 290. -[2025-11-09 22:14:18,827][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 28 and human policies 1. -[2025-11-09 22:14:18,828][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:14:42,749][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 22:14:47,484][__main__][INFO] - Number of regex retries in iteration 290: 1 -[2025-11-09 22:14:47,485][__main__][INFO] - agents played in iteration 290 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:14:48,269][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:14:48,298][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:14:48,325][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:14:48,349][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:14:48,349][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:14:48,350][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:14:49,009][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:14:49,463][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:14:49,968][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:14:50,480][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:14:50,979][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:14:51,486][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:14:51,992][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:14:52,495][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:14:53,006][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:14:53,508][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:14:54,006][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:14:54,505][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:14:55,006][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:14:55,512][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:14:56,013][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:14:56,513][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:14:57,023][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:14:57,525][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:14:58,031][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:14:58,534][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:14:59,033][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:14:59,534][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:15:00,034][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:15:00,534][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:15:01,040][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:15:01,541][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:15:02,045][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:15:02,548][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:15:03,053][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:15:03,557][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:15:04,056][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:15:04,556][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:15:05,055][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:15:05,554][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:15:06,056][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:15:06,554][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:15:07,052][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:15:07,554][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:15:08,053][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:15:08,554][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:15:09,052][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:15:09,549][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:15:10,048][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:15:10,545][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:15:11,041][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:15:11,540][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:15:12,039][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:15:12,537][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:15:13,036][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:15:13,535][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:15:14,060][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:15:14,559][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:15:15,058][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:15:15,558][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:15:16,062][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:15:16,565][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:15:17,064][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:15:17,567][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:15:18,067][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:15:18,569][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:15:19,069][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:15:19,573][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:15:20,078][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:15:20,580][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:15:21,079][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10456 tokens. -[2025-11-09 22:15:21,790][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 55.64%, Block Peak % of device VRAM: 62.36%, ΔTime: 00:00:32 -[2025-11-09 22:15:22,484][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:15:22,488][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:15:22,490][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:15:24,396][__main__][INFO] - Iteration 291 took 1m 5s (43.70% Gen, 53.39% Train). Generation: 28s, Training: 35s. Estimated remaining time: 50h 22m 2s. Estimated total time: 54h 38m 29s. Time estimates for 10 more iterations: 10m 55s, 100 more iterations: 1h 49m 16s, 500 more iterations: 9h 6m 24s. -[2025-11-09 22:15:24,398][__main__][INFO] - Starting iteration 291. -[2025-11-09 22:15:24,871][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 29 and human policies 1. -[2025-11-09 22:15:24,872][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:15:45,393][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 22:15:46,696][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 22:15:48,252][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 22:15:57,324][__main__][INFO] - Number of regex retries in iteration 291: 3 -[2025-11-09 22:15:57,325][__main__][INFO] - agents played in iteration 291 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:15:58,155][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:15:58,179][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:15:58,204][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:15:58,227][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:15:58,228][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:15:58,228][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:15:58,967][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:15:59,429][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:15:59,937][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:16:00,443][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:16:00,946][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:16:01,451][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:16:01,952][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:16:02,454][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:16:02,961][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:16:03,460][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:16:03,961][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:16:04,463][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:16:04,966][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:16:05,472][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:16:05,977][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:16:06,476][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:16:06,984][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:16:07,490][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:16:07,990][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:16:08,496][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:16:08,997][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:16:09,502][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:16:10,004][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:16:10,506][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:16:11,007][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:16:11,510][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:16:12,027][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:16:12,526][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:16:13,023][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:16:13,522][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:16:14,020][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:16:14,522][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:16:15,019][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:16:15,515][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:16:16,022][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:16:16,521][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:16:17,016][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:16:17,515][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:16:18,012][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:16:18,520][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:16:19,017][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:16:19,517][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:16:20,020][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:16:20,521][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:16:21,022][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:16:21,524][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:16:22,024][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:16:22,531][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:16:23,034][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:16:23,532][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:16:24,032][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:16:24,537][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:16:25,034][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:16:25,534][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:16:26,034][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:16:26,539][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:16:27,042][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:16:27,552][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:16:28,056][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:16:28,560][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:16:29,064][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:16:29,568][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:16:30,070][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:16:30,573][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:16:31,077][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10414 tokens. -[2025-11-09 22:16:31,848][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.28%, Current % of VRAM taken: 55.78%, Block Peak % of device VRAM: 62.36%, ΔTime: 00:00:32 -[2025-11-09 22:16:32,476][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:16:32,478][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:16:32,480][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:16:33,410][__main__][INFO] - Iteration 292 took 1m 8s (47.35% Gen, 51.29% Train). Generation: 32s, Training: 35s. Estimated remaining time: 52h 49m 20s. Estimated total time: 57h 6m 57s. Time estimates for 10 more iterations: 11m 25s, 100 more iterations: 1h 54m 13s, 500 more iterations: 9h 31m 9s. -[2025-11-09 22:16:33,412][__main__][INFO] - Starting iteration 292. -[2025-11-09 22:16:33,880][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 29 and human policies 1. -[2025-11-09 22:16:33,880][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:16:42,606][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 22:16:54,191][__main__][INFO] - Number of regex retries in iteration 292: 1 -[2025-11-09 22:16:54,191][__main__][INFO] - agents played in iteration 292 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:16:54,978][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:16:55,001][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:16:55,026][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:16:55,048][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:16:55,049][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:16:55,049][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:16:55,726][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:16:56,182][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:16:56,693][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:16:57,193][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:16:57,695][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:16:58,207][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:16:58,706][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:16:59,204][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:16:59,704][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:17:00,204][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:17:00,735][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:17:01,237][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:17:01,743][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:17:02,244][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:17:02,745][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:17:03,252][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:17:03,754][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:17:04,252][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:17:04,758][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:17:05,259][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:17:05,757][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:17:06,256][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:17:06,756][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:17:07,260][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:17:07,762][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:17:08,263][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:17:08,767][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:17:09,269][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:17:09,771][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:17:10,272][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:17:10,772][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:17:11,277][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:17:11,779][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:17:12,281][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:17:12,785][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:17:13,284][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:17:13,782][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:17:14,282][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:17:14,784][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:17:15,286][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:17:15,787][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:17:16,282][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:17:16,781][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:17:17,278][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:17:17,777][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:17:18,273][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:17:18,770][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:17:19,267][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:17:19,763][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:17:20,263][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:17:20,762][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:17:21,261][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:17:21,762][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:17:22,258][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:17:22,754][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:17:23,261][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:17:23,764][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:17:24,266][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:17:24,769][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:17:25,273][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:17:25,777][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:17:26,281][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:17:26,784][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:17:27,288][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:17:27,793][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10588 tokens. -[2025-11-09 22:17:28,566][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 55.74%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:32 -[2025-11-09 22:17:29,257][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:17:29,262][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:17:29,263][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:17:32,379][__main__][INFO] - Iteration 293 took 58s (34.72% Gen, 59.95% Train). Generation: 20s, Training: 35s. Estimated remaining time: 44h 26m 27s. Estimated total time: 48h 45m 3s. Time estimates for 10 more iterations: 9m 45s, 100 more iterations: 1h 37m 30s, 500 more iterations: 8h 7m 30s. -[2025-11-09 22:17:32,385][__main__][INFO] - Starting iteration 293. -[2025-11-09 22:17:32,861][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 29 and human policies 1. -[2025-11-09 22:17:32,862][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:18:00,262][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 22:18:07,928][__main__][INFO] - Number of regex retries in iteration 293: 1 -[2025-11-09 22:18:07,929][__main__][INFO] - agents played in iteration 293 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:18:08,748][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:18:08,772][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:18:08,796][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:18:08,818][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:18:08,819][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:18:08,820][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:18:09,558][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:18:10,019][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:18:10,530][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:18:11,030][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:18:11,533][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:18:12,055][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:18:12,555][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:18:13,059][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:18:13,561][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:18:14,059][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:18:14,566][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:18:15,065][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:18:15,569][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:18:16,067][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:18:16,566][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:18:17,070][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:18:17,574][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:18:18,075][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:18:18,579][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:18:19,081][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:18:19,585][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:18:20,090][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:18:20,593][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:18:21,100][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:18:21,604][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:18:22,107][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:18:22,611][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:18:23,112][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:18:23,614][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:18:24,114][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:18:24,615][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:18:25,127][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:18:25,628][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:18:26,133][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:18:26,641][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:18:27,143][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:18:27,652][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:18:28,151][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:18:28,650][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:18:29,154][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:18:29,657][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:18:30,159][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:18:30,659][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:18:31,161][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:18:31,669][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:18:32,172][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:18:32,678][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:18:33,180][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:18:33,683][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:18:34,187][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:18:34,690][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:18:35,191][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:18:35,696][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:18:36,200][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:18:36,703][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:18:37,207][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:18:37,710][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:18:38,214][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:18:38,713][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:18:39,216][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:18:39,716][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:18:40,216][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:18:40,732][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:18:41,232][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:18:41,734][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10532 tokens. -[2025-11-09 22:18:42,510][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 55.64%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 -[2025-11-09 22:18:43,158][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:18:43,160][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:18:43,163][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:18:44,375][__main__][INFO] - Iteration 294 took 1m 11s (49.03% Gen, 49.27% Train). Generation: 35s, Training: 35s. Estimated remaining time: 55h 15m 55s. Estimated total time: 59h 35m 43s. Time estimates for 10 more iterations: 11m 55s, 100 more iterations: 1h 59m 11s, 500 more iterations: 9h 55m 57s. -[2025-11-09 22:18:44,377][__main__][INFO] - Starting iteration 294. -[2025-11-09 22:18:44,842][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 29 and human policies 1. -[2025-11-09 22:18:44,843][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:18:54,608][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 22:19:07,978][__main__][INFO] - Number of regex retries in iteration 294: 1 -[2025-11-09 22:19:07,978][__main__][INFO] - agents played in iteration 294 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:19:08,773][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:19:08,801][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:19:08,828][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:19:08,852][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:19:08,853][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:19:08,854][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:19:09,570][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:19:10,028][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:19:10,531][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:19:11,038][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:19:11,537][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:19:12,040][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:19:12,542][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:19:13,042][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:19:13,544][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:19:14,052][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:19:14,560][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:19:15,072][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:19:15,577][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:19:16,078][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:19:16,577][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:19:17,079][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:19:17,590][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:19:18,094][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:19:18,613][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:19:19,151][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:19:21,133][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:19:21,659][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:19:22,165][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:19:22,673][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:19:23,180][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:19:23,685][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:19:24,185][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:19:24,683][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:19:25,181][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:19:25,678][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:19:26,178][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:19:26,676][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:19:27,174][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:19:27,673][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:19:28,173][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:19:28,674][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:19:29,176][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:19:29,677][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:19:30,179][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:19:30,682][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:19:31,185][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:19:31,687][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:19:32,193][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:19:32,696][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:19:33,205][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:19:33,710][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:19:34,211][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:19:34,717][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:19:35,219][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:19:35,723][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:19:36,238][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:19:36,739][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:19:37,252][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:19:37,752][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:19:38,259][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:19:38,766][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:19:39,273][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:19:39,774][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:19:40,274][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:19:40,774][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:19:41,283][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:19:41,785][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:19:42,285][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:19:42,792][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:19:43,293][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10404 tokens. -[2025-11-09 22:19:44,045][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 55.65%, Block Peak % of device VRAM: 62.38%, ΔTime: 00:00:34 -[2025-11-09 22:19:44,683][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:19:44,685][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:19:44,686][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:19:45,636][__main__][INFO] - Iteration 295 took 1m 0s (38.06% Gen, 60.38% Train). Generation: 23s, Training: 36s. Estimated remaining time: 46h 18m 54s. Estimated total time: 50h 39m 42s. Time estimates for 10 more iterations: 10m 7s, 100 more iterations: 1h 41m 19s, 500 more iterations: 8h 26m 37s. -[2025-11-09 22:19:45,638][__main__][INFO] - Starting iteration 295. -[2025-11-09 22:19:46,100][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 29 and human policies 1. -[2025-11-09 22:19:46,101][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:20:05,953][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 22:20:19,370][__main__][INFO] - Number of regex retries in iteration 295: 1 -[2025-11-09 22:20:19,370][__main__][INFO] - agents played in iteration 295 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:20:20,209][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:20:20,232][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:20:20,254][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:20:20,278][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:20:20,278][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:20:20,280][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:20:21,026][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:20:21,488][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:20:21,999][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:20:22,502][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:20:23,005][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:20:23,512][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:20:24,016][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:20:24,524][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:20:25,025][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:20:25,528][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:20:26,033][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:20:26,537][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:20:27,035][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:20:27,536][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:20:28,042][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:20:28,560][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:20:29,063][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:20:29,565][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:20:30,065][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:20:30,564][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:20:31,076][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:20:31,573][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:20:32,073][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:20:32,582][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:20:33,079][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:20:33,575][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:20:34,081][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:20:34,578][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:20:35,085][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:20:35,582][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:20:36,079][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:20:36,588][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:20:37,090][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:20:37,597][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:20:38,103][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:20:38,605][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:20:39,113][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:20:39,618][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:20:40,122][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:20:40,631][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:20:41,134][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:20:41,636][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:20:42,138][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:20:42,641][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:20:43,153][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:20:43,655][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:20:44,157][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:20:44,656][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:20:45,161][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:20:45,659][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:20:46,160][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:20:46,660][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:20:47,156][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:20:47,656][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:20:48,154][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:20:48,652][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:20:49,151][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:20:49,666][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:20:50,165][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:20:50,666][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:20:51,166][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:20:51,666][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:20:52,183][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:20:52,680][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:20:53,177][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10468 tokens. -[2025-11-09 22:20:53,887][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 55.66%, Block Peak % of device VRAM: 62.49%, ΔTime: 00:00:32 -[2025-11-09 22:20:54,631][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:20:54,633][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:20:54,635][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:20:55,642][__main__][INFO] - Iteration 296 took 1m 9s (47.84% Gen, 50.71% Train). Generation: 33s, Training: 35s. Estimated remaining time: 53h 35m 10s. Estimated total time: 57h 57m 9s. Time estimates for 10 more iterations: 11m 35s, 100 more iterations: 1h 55m 54s, 500 more iterations: 9h 39m 31s. -[2025-11-09 22:20:55,644][__main__][INFO] - Starting iteration 296. -[2025-11-09 22:20:56,119][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 29 and human policies 1. -[2025-11-09 22:20:56,119][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:21:03,248][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 22:21:04,319][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 22:21:14,939][__main__][INFO] - Number of regex retries in iteration 296: 2 -[2025-11-09 22:21:14,939][__main__][INFO] - agents played in iteration 296 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:21:15,760][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:21:15,787][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:21:15,814][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:21:15,837][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:21:15,838][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:21:15,839][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:21:16,509][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:21:16,973][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:21:17,475][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:21:17,973][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:21:18,481][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:21:18,984][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:21:19,488][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:21:19,994][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:21:20,497][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:21:21,002][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:21:21,500][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:21:22,002][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:21:22,506][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:21:23,008][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:21:23,511][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:21:24,013][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:21:24,515][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:21:25,027][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:21:25,533][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:21:26,037][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:21:26,540][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:21:27,043][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:21:27,560][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:21:28,061][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:21:28,561][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:21:29,064][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:21:29,562][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:21:30,071][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:21:30,572][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:21:31,071][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:21:31,582][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:21:32,077][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:21:32,579][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:21:33,086][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:21:33,589][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:21:34,092][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:21:34,594][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:21:35,095][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:21:35,605][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:21:36,110][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:21:36,616][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:21:37,117][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:21:37,621][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:21:38,129][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:21:38,629][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:21:39,130][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:21:39,633][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:21:40,136][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:21:40,635][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:21:41,135][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:21:41,636][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:21:42,136][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:21:42,638][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:21:43,137][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:21:43,641][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:21:44,138][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:21:44,636][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:21:45,135][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:21:45,636][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:21:46,140][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:21:46,643][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:21:47,143][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:21:47,652][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:21:48,149][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:21:48,651][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10492 tokens. -[2025-11-09 22:21:49,426][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 55.72%, Block Peak % of device VRAM: 62.34%, ΔTime: 00:00:32 -[2025-11-09 22:21:50,158][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:21:50,160][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:21:50,162][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:21:51,140][__main__][INFO] - Iteration 297 took 55s (34.20% Gen, 64.02% Train). Generation: 18s, Training: 35s. Estimated remaining time: 41h 28m 10s. Estimated total time: 45h 51m 4s. Time estimates for 10 more iterations: 9m 10s, 100 more iterations: 1h 31m 42s, 500 more iterations: 7h 38m 30s. -[2025-11-09 22:21:51,142][__main__][INFO] - Starting iteration 297. -[2025-11-09 22:21:51,612][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 29 and human policies 1. -[2025-11-09 22:21:51,613][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:22:12,442][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 22:22:16,667][__main__][INFO] - Number of regex retries in iteration 297: 1 -[2025-11-09 22:22:16,667][__main__][INFO] - agents played in iteration 297 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:22:17,517][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:22:17,552][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:22:17,582][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:22:17,607][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:22:17,608][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:22:17,609][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:22:18,362][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:22:18,823][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:22:19,328][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:22:19,829][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:22:20,338][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:22:20,847][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:22:21,348][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:22:21,850][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:22:22,354][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:22:22,869][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:22:23,373][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:22:23,880][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:22:24,381][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:22:24,888][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:22:25,393][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:22:25,897][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:22:26,398][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:22:26,905][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:22:27,412][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:22:27,918][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:22:28,423][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:22:28,926][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:22:29,432][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:22:29,935][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:22:30,440][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:22:30,945][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:22:31,449][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:22:31,952][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:22:32,456][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:22:32,959][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:22:33,469][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:22:33,974][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:22:34,477][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:22:34,999][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:22:35,504][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:22:36,007][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:22:36,511][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:22:37,013][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:22:37,522][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:22:38,024][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:22:38,525][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:22:39,033][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:22:39,535][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:22:40,036][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:22:40,537][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:22:41,038][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:22:41,539][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:22:42,039][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:22:42,542][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:22:43,046][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:22:43,548][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:22:44,051][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:22:44,549][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:22:45,050][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:22:45,556][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:22:46,058][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:22:46,558][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:22:47,057][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:22:47,556][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:22:48,058][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:22:48,560][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:22:49,060][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:22:49,561][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:22:50,060][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:22:50,556][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10495 tokens. -[2025-11-09 22:22:51,251][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 55.59%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 -[2025-11-09 22:22:51,888][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:22:51,890][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:22:51,891][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:22:52,785][__main__][INFO] - Iteration 298 took 1m 1s (40.96% Gen, 57.58% Train). Generation: 25s, Training: 35s. Estimated remaining time: 46h 34m 44s. Estimated total time: 50h 58m 40s. Time estimates for 10 more iterations: 10m 11s, 100 more iterations: 1h 41m 57s, 500 more iterations: 8h 29m 46s. -[2025-11-09 22:22:52,787][__main__][INFO] - Starting iteration 298. -[2025-11-09 22:22:53,255][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 29 and human policies 1. -[2025-11-09 22:22:53,256][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:23:08,582][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 22:23:21,365][__main__][INFO] - Number of regex retries in iteration 298: 1 -[2025-11-09 22:23:21,366][__main__][INFO] - agents played in iteration 298 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:23:22,169][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:23:22,197][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:23:22,223][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:23:22,247][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:23:22,247][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:23:22,249][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:23:22,949][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:23:23,415][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:23:23,920][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:23:24,425][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:23:24,928][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:23:25,436][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:23:25,939][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:23:26,440][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:23:26,944][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:23:27,452][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:23:27,957][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:23:28,458][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:23:28,963][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:23:29,462][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:23:29,970][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:23:30,471][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:23:30,973][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:23:31,475][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:23:31,979][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:23:32,495][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:23:32,998][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:23:33,500][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:23:34,004][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:23:34,510][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:23:35,018][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:23:35,519][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:23:36,020][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:23:36,530][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:23:37,029][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:23:37,531][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:23:38,027][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:23:38,527][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:23:39,027][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:23:39,551][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:23:40,055][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:23:40,559][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:23:41,062][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:23:41,565][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:23:42,066][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:23:42,635][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:23:44,511][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:23:45,014][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:23:45,520][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:23:46,021][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:23:46,524][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:23:47,032][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:23:47,534][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:23:48,038][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:23:48,539][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:23:49,038][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:23:49,545][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:23:50,045][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:23:50,547][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:23:51,054][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:23:51,554][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:23:52,052][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:23:52,554][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:23:53,055][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:23:53,559][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:23:54,061][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:23:54,568][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:23:55,065][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:23:55,566][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:23:56,065][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:23:56,563][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10412 tokens. -[2025-11-09 22:23:57,357][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 55.64%, Block Peak % of device VRAM: 62.27%, ΔTime: 00:00:34 -[2025-11-09 22:23:58,008][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:23:58,009][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:23:58,011][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:23:58,948][__main__][INFO] - Iteration 299 took 1m 5s (42.79% Gen, 55.78% Train). Generation: 28s, Training: 36s. Estimated remaining time: 50h 19m 39s. Estimated total time: 54h 44m 41s. Time estimates for 10 more iterations: 10m 56s, 100 more iterations: 1h 49m 29s, 500 more iterations: 9h 7m 26s. -[2025-11-09 22:23:58,953][__main__][INFO] - Starting iteration 299. -[2025-11-09 22:23:59,436][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 29 and human policies 1. -[2025-11-09 22:23:59,437][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:24:31,927][__main__][INFO] - Number of regex retries in iteration 299: 0 -[2025-11-09 22:24:31,927][__main__][INFO] - agents played in iteration 299 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:24:32,740][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:24:32,765][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:24:32,790][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:24:32,814][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:24:32,814][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:24:32,815][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:24:33,584][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:24:34,042][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:24:34,551][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:24:35,060][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:24:35,564][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:24:36,073][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:24:36,575][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:24:37,089][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:24:37,597][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:24:38,098][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:24:38,603][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:24:39,108][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:24:39,623][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:24:40,127][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:24:40,630][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:24:41,146][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:24:41,646][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:24:42,148][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:24:42,656][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:24:43,161][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:24:43,666][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:24:44,173][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:24:44,677][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:24:45,181][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:24:45,684][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:24:46,186][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:24:46,688][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:24:47,191][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:24:47,699][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:24:48,201][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:24:48,707][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:24:49,207][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:24:49,708][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:24:50,212][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:24:50,708][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:24:51,210][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:24:51,718][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:24:52,216][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:24:52,714][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:24:53,211][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:24:53,712][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:24:54,211][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:24:54,708][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:24:55,205][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:24:55,706][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:24:56,202][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:24:56,701][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:24:57,203][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:24:57,703][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:24:58,214][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:24:58,717][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:24:59,218][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:24:59,719][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:25:00,219][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:25:00,735][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:25:01,235][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:25:01,738][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:25:02,241][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:25:02,796][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:25:03,300][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:25:03,798][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:25:04,296][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:25:04,797][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:25:05,295][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:25:05,793][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10507 tokens. -[2025-11-09 22:25:06,474][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 55.67%, Block Peak % of device VRAM: 62.23%, ΔTime: 00:00:32 -[2025-11-09 22:25:07,238][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:25:07,239][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:25:07,241][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:25:08,248][__main__][INFO] - Iteration 300 took 1m 8s (47.22% Gen, 51.32% Train). Generation: 32s, Training: 35s. Estimated remaining time: 52h 54m 26s. Estimated total time: 57h 20m 38s. Time estimates for 10 more iterations: 11m 28s, 100 more iterations: 1h 54m 41s, 500 more iterations: 9h 33m 26s. -[2025-11-09 22:25:08,250][__main__][INFO] - Starting iteration 300. -[2025-11-09 22:25:08,718][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 29 and human policies 1. -[2025-11-09 22:25:08,719][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:25:19,389][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 22:25:29,265][__main__][INFO] - Number of regex retries in iteration 300: 1 -[2025-11-09 22:25:29,266][__main__][INFO] - agents played in iteration 300 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:25:30,321][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:25:30,347][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:25:30,373][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:25:30,396][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:25:30,396][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:25:30,397][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:25:31,068][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:25:31,526][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:25:32,031][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:25:32,534][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:25:33,036][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:25:33,547][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:25:34,049][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:25:34,552][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:25:35,061][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:25:35,560][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:25:36,063][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:25:36,566][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:25:37,068][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:25:37,572][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:25:38,074][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:25:38,579][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:25:39,083][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:25:39,588][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:25:40,089][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:25:40,592][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:25:41,094][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:25:41,602][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:25:42,105][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:25:42,609][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:25:43,113][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:25:44,826][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:25:45,350][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:25:45,854][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:25:46,355][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:25:46,856][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:25:47,358][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:25:47,863][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:25:48,363][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:25:48,864][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:25:49,365][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:25:49,866][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:25:50,367][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:25:50,868][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:25:51,367][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:25:51,872][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:25:52,374][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:25:52,872][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:25:53,371][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:25:53,869][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:25:54,365][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:25:54,862][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:25:55,360][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:25:55,867][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:25:56,366][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:25:56,869][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:25:57,367][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:25:57,868][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:25:58,371][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:25:58,872][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:25:59,374][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:25:59,877][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:26:00,379][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:26:00,883][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:26:01,381][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:26:01,879][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:26:02,382][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:26:02,885][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:26:03,385][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:26:03,889][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:26:04,390][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10474 tokens. -[2025-11-09 22:26:05,106][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 55.67%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:34 -[2025-11-09 22:26:05,733][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:26:05,735][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:26:05,737][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:26:07,638][__main__][INFO] - Iteration 301 took 58s (34.87% Gen, 61.90% Train). Generation: 20s, Training: 36s. Estimated remaining time: 44h 38m 52s. Estimated total time: 49h 6m 2s. Time estimates for 10 more iterations: 9m 49s, 100 more iterations: 1h 38m 12s, 500 more iterations: 8h 11m 0s. -[2025-11-09 22:26:07,641][__main__][INFO] - Starting iteration 301. -[2025-11-09 22:26:08,103][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 30 and human policies 1. -[2025-11-09 22:26:08,104][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:26:27,088][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 22:26:39,560][__main__][INFO] - Number of regex retries in iteration 301: 1 -[2025-11-09 22:26:39,561][__main__][INFO] - agents played in iteration 301 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:26:40,377][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:26:40,403][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:26:40,427][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:26:40,452][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:26:40,453][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:26:40,454][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:26:41,230][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:26:41,693][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:26:42,201][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:26:42,703][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:26:43,207][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:26:43,710][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:26:44,212][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:26:44,717][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:26:45,219][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:26:45,721][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:26:46,222][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:26:46,727][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:26:47,227][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:26:47,730][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:26:48,233][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:26:48,757][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:26:49,261][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:26:49,764][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:26:50,270][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:26:50,774][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:26:51,284][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:26:51,787][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:26:52,288][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:26:52,793][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:26:53,296][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:26:53,797][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:26:54,300][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:26:54,803][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:26:55,307][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:26:55,811][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:26:56,313][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:26:56,814][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:26:57,319][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:26:57,824][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:26:58,325][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:26:58,824][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:26:59,329][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:26:59,828][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:27:00,326][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:27:00,825][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:27:01,324][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:27:01,844][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:27:02,343][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:27:02,925][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:27:03,430][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:27:03,930][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:27:04,441][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:27:04,939][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:27:05,438][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:27:05,944][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:27:06,444][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:27:06,947][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:27:07,445][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:27:07,943][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:27:08,441][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:27:08,938][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:27:09,438][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:27:09,935][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:27:10,433][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:27:10,946][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:27:11,443][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:27:11,941][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:27:12,455][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:27:12,956][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:27:13,469][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10420 tokens. -[2025-11-09 22:27:14,151][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.27%, Current % of VRAM taken: 55.77%, Block Peak % of device VRAM: 62.35%, ΔTime: 00:00:32 -[2025-11-09 22:27:14,897][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:27:14,898][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:27:14,900][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:27:15,892][__main__][INFO] - Iteration 302 took 1m 7s (46.40% Gen, 52.13% Train). Generation: 31s, Training: 35s. Estimated remaining time: 52h 1m 7s. Estimated total time: 56h 29m 26s. Time estimates for 10 more iterations: 11m 17s, 100 more iterations: 1h 52m 58s, 500 more iterations: 9h 24m 54s. -[2025-11-09 22:27:15,895][__main__][INFO] - Starting iteration 302. -[2025-11-09 22:27:16,373][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 30 and human policies 1. -[2025-11-09 22:27:16,373][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:27:39,533][__main__][INFO] - Number of regex retries in iteration 302: 0 -[2025-11-09 22:27:39,533][__main__][INFO] - agents played in iteration 302 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:27:40,339][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:27:40,367][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:27:40,393][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:27:40,415][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:27:40,415][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:27:40,416][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:27:41,154][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:27:41,614][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:27:42,923][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:27:43,898][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:27:44,402][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:27:44,907][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:27:45,413][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:27:45,916][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:27:46,422][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:27:46,928][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:27:47,432][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:27:47,935][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:27:48,441][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:27:48,944][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:27:49,444][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:27:49,944][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:27:50,450][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:27:50,954][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:27:51,456][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:27:51,958][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:27:52,462][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:27:52,978][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:27:53,480][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:27:53,982][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:27:54,485][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:27:54,989][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:27:55,495][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:27:56,001][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:27:56,507][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:27:57,013][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:27:57,517][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:27:58,021][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:27:58,526][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:27:59,027][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:27:59,531][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:28:00,038][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:28:00,541][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:28:01,051][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:28:01,557][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:28:02,074][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:28:02,578][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:28:03,081][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:28:03,588][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:28:04,091][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:28:04,592][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:28:05,093][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:28:05,600][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:28:06,104][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:28:06,607][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:28:07,105][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:28:07,606][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:28:08,103][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:28:08,600][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:28:09,099][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:28:09,596][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:28:10,091][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:28:10,588][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:28:11,088][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:28:11,589][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:28:12,089][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:28:12,593][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:28:13,091][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:28:13,591][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:28:14,090][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:28:14,586][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10370 tokens. -[2025-11-09 22:28:15,291][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 55.63%, Block Peak % of device VRAM: 62.34%, ΔTime: 00:00:34 -[2025-11-09 22:28:15,921][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:28:15,923][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:28:15,925][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:28:16,885][__main__][INFO] - Iteration 303 took 1m 0s (38.27% Gen, 60.14% Train). Generation: 23s, Training: 36s. Estimated remaining time: 45h 56m 18s. Estimated total time: 50h 25m 38s. Time estimates for 10 more iterations: 10m 5s, 100 more iterations: 1h 40m 51s, 500 more iterations: 8h 24m 16s. -[2025-11-09 22:28:16,887][__main__][INFO] - Starting iteration 303. -[2025-11-09 22:28:17,350][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 30 and human policies 1. -[2025-11-09 22:28:17,350][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:28:47,875][__main__][INFO] - Number of regex retries in iteration 303: 0 -[2025-11-09 22:28:47,876][__main__][INFO] - agents played in iteration 303 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:28:48,717][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:28:48,741][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:28:48,766][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:28:48,789][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:28:48,789][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:28:48,791][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:28:49,529][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:28:49,989][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:28:50,504][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:28:51,009][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:28:51,515][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:28:52,022][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:28:52,525][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:28:53,026][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:28:53,531][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:28:54,036][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:28:54,543][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:28:55,045][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:28:55,548][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:28:56,058][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:28:56,558][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:28:57,064][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:28:57,570][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:28:58,070][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:28:58,576][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:28:59,076][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:28:59,579][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:29:00,084][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:29:00,588][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:29:01,099][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:29:01,601][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:29:02,100][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:29:02,601][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:29:03,101][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:29:03,604][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:29:04,104][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:29:04,610][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:29:05,110][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:29:05,611][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:29:06,115][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:29:06,618][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:29:07,121][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:29:07,625][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:29:08,129][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:29:08,630][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:29:09,133][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:29:09,638][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:29:10,141][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:29:10,645][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:29:11,141][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:29:11,641][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:29:12,141][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:29:12,640][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:29:13,141][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:29:13,641][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:29:14,140][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:29:14,643][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:29:15,141][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:29:15,643][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:29:16,143][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:29:16,643][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:29:17,145][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:29:17,645][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:29:18,145][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:29:18,642][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:29:19,140][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:29:19,642][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:29:20,140][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:29:20,641][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:29:21,141][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:29:21,645][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10533 tokens. -[2025-11-09 22:29:22,412][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 55.72%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 -[2025-11-09 22:29:23,127][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:29:23,129][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:29:23,130][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:29:24,170][__main__][INFO] - Iteration 304 took 1m 6s (45.68% Gen, 52.76% Train). Generation: 30s, Training: 35s. Estimated remaining time: 51h 10m 35s. Estimated total time: 55h 41m 2s. Time estimates for 10 more iterations: 11m 8s, 100 more iterations: 1h 51m 22s, 500 more iterations: 9h 16m 50s. -[2025-11-09 22:29:24,172][__main__][INFO] - Starting iteration 304. -[2025-11-09 22:29:24,637][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 30 and human policies 1. -[2025-11-09 22:29:24,638][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:29:46,528][__main__][INFO] - Number of regex retries in iteration 304: 0 -[2025-11-09 22:29:46,530][__main__][INFO] - agents played in iteration 304 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:29:47,480][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:29:47,522][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:29:47,554][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:29:47,582][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:29:47,583][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:29:47,584][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:29:48,342][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:29:48,806][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:29:49,317][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:29:49,823][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:29:50,323][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:29:50,824][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:29:51,335][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:29:51,834][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:29:52,340][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:29:52,840][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:29:53,346][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:29:53,849][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:29:54,350][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:29:54,851][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:29:55,352][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:29:55,858][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:29:56,366][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:29:56,871][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:29:57,375][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:29:57,892][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:29:58,392][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:29:58,906][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:29:59,411][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:29:59,914][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:30:00,423][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:30:00,924][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:30:01,422][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:30:01,921][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:30:02,421][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:30:02,926][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:30:03,426][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:30:03,926][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:30:04,429][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:30:04,931][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:30:05,434][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:30:05,938][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:30:06,445][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:30:06,946][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:30:07,446][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:30:07,949][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:30:08,449][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:30:08,953][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:30:09,456][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:30:09,957][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:30:10,462][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:30:10,966][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:30:11,467][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:30:11,976][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:30:12,475][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:30:12,985][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:30:13,509][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:30:14,013][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:30:14,519][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:30:15,021][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:30:15,524][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:30:16,033][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:30:16,537][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:30:17,035][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:30:17,536][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:30:18,033][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:30:18,534][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:30:19,035][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:30:19,533][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:30:20,038][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:30:20,536][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10414 tokens. -[2025-11-09 22:30:21,255][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 55.70%, Block Peak % of device VRAM: 62.52%, ΔTime: 00:00:32 -[2025-11-09 22:30:21,880][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:30:21,882][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:30:21,883][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:30:22,798][__main__][INFO] - Iteration 305 took 58s (37.64% Gen, 60.78% Train). Generation: 21s, Training: 35s. Estimated remaining time: 43h 56m 39s. Estimated total time: 48h 28m 5s. Time estimates for 10 more iterations: 9m 41s, 100 more iterations: 1h 36m 56s, 500 more iterations: 8h 4m 40s. -[2025-11-09 22:30:22,800][__main__][INFO] - Starting iteration 305. -[2025-11-09 22:30:23,264][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 30 and human policies 1. -[2025-11-09 22:30:23,265][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:30:53,488][__main__][INFO] - Number of regex retries in iteration 305: 0 -[2025-11-09 22:30:53,489][__main__][INFO] - agents played in iteration 305 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:30:54,287][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:30:54,310][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:30:54,334][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:30:54,356][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:30:54,357][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:30:54,358][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:30:55,058][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:30:55,519][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:30:56,026][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:30:56,540][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:30:57,038][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:30:57,538][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:30:58,039][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:30:58,544][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:30:59,044][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:30:59,543][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:31:00,044][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:31:00,548][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:31:01,047][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:31:01,549][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:31:02,049][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:31:02,551][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:31:03,054][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:31:03,554][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:31:04,058][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:31:04,556][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:31:05,059][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:31:05,558][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:31:06,058][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:31:06,560][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:31:07,065][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:31:07,565][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:31:08,064][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:31:08,567][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:31:09,070][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:31:09,572][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:31:10,075][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:31:10,585][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:31:11,089][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:31:11,589][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:31:12,090][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:31:12,590][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:31:13,091][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:31:13,603][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:31:14,107][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:31:14,610][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:31:15,113][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:31:15,616][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:31:16,125][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:31:16,624][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:31:17,127][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:31:17,629][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:31:18,128][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:31:18,625][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:31:19,130][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:31:19,627][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:31:20,133][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:31:20,631][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:31:21,129][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:31:21,635][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:31:22,134][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:31:22,634][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:31:23,133][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:31:23,632][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:31:24,132][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:31:24,630][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:31:25,130][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:31:25,631][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:31:26,131][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:31:26,631][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:31:27,132][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10268 tokens. -[2025-11-09 22:31:27,897][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.27%, Current % of VRAM taken: 55.76%, Block Peak % of device VRAM: 62.24%, ΔTime: 00:00:32 -[2025-11-09 22:31:28,613][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:31:28,615][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:31:28,617][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:31:29,608][__main__][INFO] - Iteration 306 took 1m 6s (45.56% Gen, 52.95% Train). Generation: 30s, Training: 35s. Estimated remaining time: 50h 44m 41s. Estimated total time: 55h 17m 14s. Time estimates for 10 more iterations: 11m 3s, 100 more iterations: 1h 50m 34s, 500 more iterations: 9h 12m 52s. -[2025-11-09 22:31:29,610][__main__][INFO] - Starting iteration 306. -[2025-11-09 22:31:30,160][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 30 and human policies 1. -[2025-11-09 22:31:30,160][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:31:54,771][__main__][INFO] - Number of regex retries in iteration 306: 0 -[2025-11-09 22:31:54,773][__main__][INFO] - agents played in iteration 306 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:31:55,643][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:31:55,686][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:31:55,714][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:31:55,740][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:31:55,741][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:31:55,743][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:31:56,508][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:31:56,966][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:31:57,477][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:31:57,979][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:31:58,482][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:31:58,993][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:31:59,497][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:32:00,014][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:32:00,516][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:32:01,030][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:32:01,530][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:32:02,032][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:32:02,539][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:32:03,039][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:32:03,542][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:32:04,048][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:32:04,547][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:32:05,055][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:32:05,557][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:32:06,058][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:32:06,564][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:32:07,066][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:32:07,570][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:32:08,073][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:32:08,573][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:32:09,076][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:32:09,580][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:32:10,079][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:32:10,583][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:32:11,087][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:32:11,593][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:32:12,093][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:32:12,596][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:32:13,096][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:32:13,597][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:32:14,114][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:32:14,615][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:32:15,114][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:32:15,618][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:32:16,121][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:32:16,636][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:32:17,139][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:32:17,639][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:32:18,150][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:32:18,652][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:32:19,152][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:32:19,654][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:32:20,158][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:32:20,662][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:32:21,163][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:32:21,664][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:32:22,167][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:32:22,669][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:32:23,165][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:32:23,665][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:32:24,159][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:32:24,658][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:32:25,154][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:32:25,652][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:32:26,151][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:32:26,645][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:32:27,141][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:32:27,635][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:32:28,134][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:32:28,645][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10416 tokens. -[2025-11-09 22:32:29,350][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 55.70%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:32 -[2025-11-09 22:32:29,995][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:32:29,997][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:32:29,999][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:32:30,892][__main__][INFO] - Iteration 307 took 1m 0s (40.52% Gen, 58.00% Train). Generation: 24s, Training: 35s. Estimated remaining time: 46h 3m 2s. Estimated total time: 50h 36m 36s. Time estimates for 10 more iterations: 10m 7s, 100 more iterations: 1h 41m 13s, 500 more iterations: 8h 26m 6s. -[2025-11-09 22:32:30,894][__main__][INFO] - Starting iteration 307. -[2025-11-09 22:32:31,384][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 30 and human policies 1. -[2025-11-09 22:32:31,385][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:33:00,174][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 22:33:05,576][__main__][INFO] - Number of regex retries in iteration 307: 1 -[2025-11-09 22:33:05,576][__main__][INFO] - agents played in iteration 307 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:33:06,365][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:33:06,391][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:33:06,416][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:33:06,439][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:33:06,439][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:33:06,440][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:33:07,133][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:33:07,589][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:33:08,093][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:33:08,594][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:33:09,094][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:33:09,596][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:33:10,098][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:33:10,598][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:33:11,105][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:33:11,609][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:33:12,114][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:33:12,621][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:33:13,125][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:33:13,641][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:33:14,142][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:33:14,646][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:33:15,150][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:33:15,653][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:33:16,181][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:33:16,688][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:33:17,192][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:33:17,701][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:33:18,208][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:33:18,717][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:33:19,220][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:33:19,721][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:33:20,233][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:33:20,734][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:33:21,239][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:33:21,737][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:33:22,237][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:33:22,735][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:33:23,233][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:33:23,729][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:33:24,228][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:33:24,722][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:33:25,220][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:33:25,717][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:33:26,215][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:33:26,717][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:33:27,215][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:33:27,716][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:33:28,215][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:33:28,711][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:33:29,215][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:33:29,714][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:33:30,213][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:33:30,717][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:33:31,215][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:33:31,714][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:33:32,213][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:33:32,714][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:33:33,213][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:33:33,711][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:33:34,212][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:33:34,715][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:33:35,213][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:33:35,715][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:33:36,218][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:33:36,720][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:33:37,233][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:33:37,732][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:33:38,231][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:33:38,728][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:33:39,224][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10348 tokens. -[2025-11-09 22:33:39,916][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 55.66%, Block Peak % of device VRAM: 62.50%, ΔTime: 00:00:32 -[2025-11-09 22:33:40,639][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:33:40,641][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:33:40,643][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:33:41,667][__main__][INFO] - Iteration 308 took 1m 10s (48.65% Gen, 49.89% Train). Generation: 34s, Training: 35s. Estimated remaining time: 53h 59m 26s. Estimated total time: 58h 34m 11s. Time estimates for 10 more iterations: 11m 42s, 100 more iterations: 1h 57m 8s, 500 more iterations: 9h 45m 41s. -[2025-11-09 22:33:41,669][__main__][INFO] - Starting iteration 308. -[2025-11-09 22:33:42,148][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 30 and human policies 1. -[2025-11-09 22:33:42,148][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:34:06,388][__main__][INFO] - Number of regex retries in iteration 308: 0 -[2025-11-09 22:34:06,390][__main__][INFO] - agents played in iteration 308 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:34:07,278][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:34:07,319][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:34:07,349][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:34:07,375][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:34:07,376][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:34:07,377][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:34:08,155][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:34:08,616][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:34:09,121][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:34:09,622][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:34:10,137][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:34:10,642][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:34:11,144][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:34:11,645][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:34:12,145][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:34:12,645][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:34:13,147][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:34:13,653][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:34:14,156][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:34:14,664][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:34:15,164][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:34:15,664][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:34:16,163][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:34:16,664][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:34:17,169][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:34:17,672][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:34:18,175][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:34:18,677][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:34:19,183][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:34:19,687][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:34:20,191][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:34:20,692][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:34:21,194][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:34:21,696][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:34:22,198][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:34:22,701][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:34:23,205][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:34:23,704][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:34:24,204][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:34:24,707][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:34:25,212][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:34:25,716][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:34:26,219][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:34:26,721][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:34:27,224][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:34:27,725][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:34:28,231][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:34:28,735][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:34:29,234][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:34:29,732][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:34:30,231][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:34:30,731][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:34:31,230][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:34:31,728][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:34:32,226][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:34:32,723][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:34:33,222][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:34:33,721][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:34:34,219][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:34:34,717][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:34:35,214][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:34:35,716][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:34:36,218][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:34:36,715][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:34:37,215][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:34:37,715][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:34:38,214][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:34:38,718][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:34:39,218][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:34:39,716][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:34:40,214][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10425 tokens. -[2025-11-09 22:34:40,955][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.04%, Current % of VRAM taken: 55.54%, Block Peak % of device VRAM: 62.21%, ΔTime: 00:00:32 -[2025-11-09 22:34:41,579][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:34:41,580][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:34:41,582][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:34:42,469][__main__][INFO] - Iteration 309 took 1m 0s (40.19% Gen, 58.34% Train). Generation: 24s, Training: 35s. Estimated remaining time: 45h 40m 18s. Estimated total time: 50h 16m 4s. Time estimates for 10 more iterations: 10m 3s, 100 more iterations: 1h 40m 32s, 500 more iterations: 8h 22m 40s. -[2025-11-09 22:34:42,471][__main__][INFO] - Starting iteration 309. -[2025-11-09 22:34:42,943][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 30 and human policies 1. -[2025-11-09 22:34:42,944][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:35:14,589][__main__][INFO] - Number of regex retries in iteration 309: 0 -[2025-11-09 22:35:14,590][__main__][INFO] - agents played in iteration 309 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:35:15,412][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:35:15,440][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:35:15,468][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:35:15,492][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:35:15,492][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:35:15,493][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:35:16,243][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:35:16,701][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:35:17,207][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:35:17,708][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:35:18,207][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:35:18,708][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:35:19,212][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:35:19,715][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:35:20,215][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:35:20,721][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:35:21,249][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:35:21,757][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:35:22,257][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:35:22,775][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:35:23,280][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:35:23,782][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:35:24,284][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:35:24,784][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:35:25,282][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:35:25,782][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:35:26,284][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:35:26,785][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:35:27,284][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:35:27,795][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:35:28,295][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:35:28,797][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:35:29,297][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:35:29,794][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:35:30,292][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:35:30,790][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:35:31,287][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:35:31,805][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:35:32,302][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:35:32,800][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:35:33,298][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:35:33,795][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:35:34,301][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:35:34,799][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:35:35,295][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:35:35,795][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:35:36,293][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:35:36,791][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:35:37,287][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:35:37,782][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:35:38,291][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:35:38,788][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:35:39,286][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:35:39,784][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:35:40,281][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:35:40,781][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:35:41,281][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:35:41,781][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:35:42,294][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:35:42,795][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:35:43,296][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:35:43,795][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:35:44,298][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:35:44,798][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:35:45,297][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:35:45,793][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:35:46,304][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:35:46,799][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:35:47,298][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:35:47,794][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:35:48,293][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10270 tokens. -[2025-11-09 22:35:48,992][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 55.72%, Block Peak % of device VRAM: 62.28%, ΔTime: 00:00:32 -[2025-11-09 22:35:49,711][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:35:49,713][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:35:49,715][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:35:50,725][__main__][INFO] - Iteration 310 took 1m 7s (46.69% Gen, 51.82% Train). Generation: 31s, Training: 35s. Estimated remaining time: 51h 52m 12s. Estimated total time: 56h 29m 5s. Time estimates for 10 more iterations: 11m 17s, 100 more iterations: 1h 52m 58s, 500 more iterations: 9h 24m 50s. -[2025-11-09 22:35:50,727][__main__][INFO] - Starting iteration 310. -[2025-11-09 22:35:51,190][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 30 and human policies 1. -[2025-11-09 22:35:51,190][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:36:07,021][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 22:36:18,426][__main__][INFO] - Number of regex retries in iteration 310: 1 -[2025-11-09 22:36:18,427][__main__][INFO] - agents played in iteration 310 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:36:19,253][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:36:19,290][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:36:19,319][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:36:19,345][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:36:19,345][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:36:19,346][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:36:20,067][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:36:20,535][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:36:21,041][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:36:21,543][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:36:22,045][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:36:22,552][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:36:23,056][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:36:23,559][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:36:24,062][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:36:24,568][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:36:25,071][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:36:25,575][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:36:26,078][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:36:26,582][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:36:27,105][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:36:27,608][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:36:28,115][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:36:28,615][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:36:29,116][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:36:29,618][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:36:30,119][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:36:30,617][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:36:31,120][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:36:31,620][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:36:32,123][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:36:32,622][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:36:33,123][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:36:33,631][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:36:34,136][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:36:34,635][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:36:35,133][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:36:35,631][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:36:36,130][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:36:36,634][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:36:37,135][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:36:37,637][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:36:38,137][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:36:38,636][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:36:39,134][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:36:39,630][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:36:40,129][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:36:40,625][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:36:41,118][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:36:41,621][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:36:42,119][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:36:42,617][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:36:43,115][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:36:43,611][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:36:44,122][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:36:44,618][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:36:45,114][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:36:45,624][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:36:46,120][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:36:46,622][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:36:47,119][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:36:47,615][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:36:48,119][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:36:48,620][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:36:49,120][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:36:49,618][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:36:50,119][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:36:50,621][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:36:51,120][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:36:51,617][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:36:52,125][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10408 tokens. -[2025-11-09 22:36:52,910][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.25%, Current % of VRAM taken: 55.74%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:32 -[2025-11-09 22:36:53,545][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:36:53,547][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:36:53,549][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:36:55,401][__main__][INFO] - Iteration 311 took 1m 4s (42.41% Gen, 54.70% Train). Generation: 27s, Training: 35s. Estimated remaining time: 48h 52m 37s. Estimated total time: 53h 30m 35s. Time estimates for 10 more iterations: 10m 42s, 100 more iterations: 1h 47m 1s, 500 more iterations: 8h 55m 5s. -[2025-11-09 22:36:55,403][__main__][INFO] - Starting iteration 311. -[2025-11-09 22:36:55,875][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 31 and human policies 1. -[2025-11-09 22:36:55,875][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:37:21,792][__main__][INFO] - Number of regex retries in iteration 311: 0 -[2025-11-09 22:37:21,793][__main__][INFO] - agents played in iteration 311 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:37:22,691][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:37:22,720][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:37:22,747][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:37:22,771][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:37:22,771][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:37:22,772][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:37:23,525][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:37:23,985][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:37:24,490][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:37:25,016][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:37:25,518][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:37:26,018][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:37:26,524][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:37:27,023][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:37:27,523][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:37:28,021][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:37:28,519][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:37:29,030][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:37:29,529][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:37:30,030][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:37:30,530][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:37:31,027][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:37:31,528][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:37:32,028][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:37:32,526][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:37:33,031][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:37:33,533][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:37:34,709][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:37:35,297][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:37:35,812][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:37:36,312][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:37:36,820][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:37:37,319][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:37:37,819][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:37:38,328][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:37:38,831][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:37:39,331][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:37:39,828][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:37:40,328][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:37:40,827][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:37:41,326][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:37:41,826][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:37:42,326][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:37:42,827][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:37:43,327][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:37:43,824][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:37:44,325][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:37:44,826][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:37:45,321][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:37:45,820][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:37:46,319][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:37:46,820][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:37:47,318][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:37:47,822][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:37:48,325][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:37:48,832][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:37:49,353][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:37:49,857][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:37:50,359][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:37:50,859][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:37:51,362][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:37:51,862][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:37:52,366][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:37:52,863][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:37:53,360][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:37:53,858][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:37:54,359][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:37:54,861][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:37:55,365][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:37:55,864][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:37:56,363][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10389 tokens. -[2025-11-09 22:37:57,111][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 55.65%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:33 -[2025-11-09 22:37:57,785][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:37:57,787][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:37:57,789][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:37:58,755][__main__][INFO] - Iteration 312 took 1m 2s (41.22% Gen, 57.24% Train). Generation: 25s, Training: 35s. Estimated remaining time: 47h 44m 59s. Estimated total time: 52h 24m 1s. Time estimates for 10 more iterations: 10m 28s, 100 more iterations: 1h 44m 48s, 500 more iterations: 8h 44m 0s. -[2025-11-09 22:37:58,757][__main__][INFO] - Starting iteration 312. -[2025-11-09 22:37:59,239][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 31 and human policies 1. -[2025-11-09 22:37:59,240][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:38:31,702][__main__][INFO] - Number of regex retries in iteration 312: 0 -[2025-11-09 22:38:31,703][__main__][INFO] - agents played in iteration 312 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:38:32,538][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:38:32,561][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:38:32,585][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:38:32,607][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:38:32,608][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:38:32,610][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:38:33,358][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:38:33,821][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:38:34,329][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:38:34,833][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:38:35,338][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:38:35,840][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:38:36,340][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:38:36,843][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:38:37,350][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:38:37,852][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:38:38,359][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:38:38,861][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:38:39,368][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:38:39,873][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:38:40,373][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:38:40,879][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:38:41,382][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:38:41,884][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:38:42,382][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:38:42,883][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:38:43,387][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:38:43,886][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:38:44,386][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:38:44,883][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:38:45,380][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:38:45,876][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:38:46,373][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:38:46,867][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:38:47,363][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:38:47,858][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:38:48,354][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:38:48,852][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:38:49,346][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:38:49,843][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:38:50,343][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:38:50,840][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:38:51,343][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:38:51,852][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:38:52,354][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:38:52,853][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:38:53,354][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:38:53,855][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:38:54,360][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:38:54,860][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:38:55,358][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:38:55,862][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:38:56,363][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:38:56,865][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:38:57,365][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:38:57,863][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:38:58,363][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:38:58,868][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:38:59,372][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:38:59,875][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:39:00,378][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:39:00,880][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:39:01,378][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:39:01,880][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:39:02,386][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:39:02,889][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:39:03,391][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:39:03,895][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:39:04,397][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:39:04,899][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:39:05,401][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10350 tokens. -[2025-11-09 22:39:06,221][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 55.70%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:32 -[2025-11-09 22:39:06,974][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:39:06,976][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:39:06,978][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:39:07,913][__main__][INFO] - Iteration 313 took 1m 8s (47.27% Gen, 51.37% Train). Generation: 32s, Training: 35s. Estimated remaining time: 52h 33m 33s. Estimated total time: 57h 13m 43s. Time estimates for 10 more iterations: 11m 26s, 100 more iterations: 1h 54m 27s, 500 more iterations: 9h 32m 17s. -[2025-11-09 22:39:07,915][__main__][INFO] - Starting iteration 313. -[2025-11-09 22:39:08,383][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 31 and human policies 1. -[2025-11-09 22:39:08,384][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:39:28,442][__main__][INFO] - Number of regex retries in iteration 313: 0 -[2025-11-09 22:39:28,443][__main__][INFO] - agents played in iteration 313 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:39:29,268][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:39:29,297][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:39:29,324][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:39:29,348][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:39:29,348][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:39:29,349][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:39:30,015][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:39:30,476][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:39:30,978][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:39:31,475][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:39:31,977][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:39:32,479][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:39:32,986][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:39:33,487][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:39:33,988][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:39:34,488][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:39:34,991][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:39:35,500][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:39:36,015][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:39:36,517][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:39:37,025][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:39:37,524][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:39:38,025][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:39:38,545][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:39:39,049][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:39:39,557][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:39:40,058][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:39:40,558][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:39:41,478][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:39:42,382][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:39:42,882][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:39:43,382][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:39:43,887][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:39:44,388][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:39:44,891][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:39:45,391][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:39:45,894][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:39:46,397][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:39:46,898][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:39:47,401][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:39:47,926][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:39:48,424][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:39:48,929][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:39:49,431][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:39:49,938][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:39:50,455][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:39:50,956][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:39:51,458][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:39:51,958][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:39:52,456][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:39:52,962][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:39:53,458][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:39:53,955][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:39:54,458][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:39:54,960][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:39:55,461][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:39:55,959][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:39:56,462][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:39:56,964][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:39:57,468][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:39:57,974][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:39:58,476][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:39:58,980][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:39:59,485][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:39:59,988][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:40:00,490][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:40:00,996][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:40:01,502][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:40:02,006][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:40:02,521][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:40:03,029][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10290 tokens. -[2025-11-09 22:40:03,833][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 55.74%, Block Peak % of device VRAM: 62.22%, ΔTime: 00:00:33 -[2025-11-09 22:40:04,487][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:40:04,489][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:40:04,491][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:40:05,482][__main__][INFO] - Iteration 314 took 57s (35.13% Gen, 63.13% Train). Generation: 20s, Training: 36s. Estimated remaining time: 42h 53m 50s. Estimated total time: 47h 34m 59s. Time estimates for 10 more iterations: 9m 30s, 100 more iterations: 1h 35m 9s, 500 more iterations: 7h 55m 49s. -[2025-11-09 22:40:05,485][__main__][INFO] - Starting iteration 314. -[2025-11-09 22:40:05,952][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 31 and human policies 1. -[2025-11-09 22:40:05,954][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:40:35,072][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 22:40:41,728][__main__][INFO] - Number of regex retries in iteration 314: 1 -[2025-11-09 22:40:41,728][__main__][INFO] - agents played in iteration 314 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:40:42,541][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:40:42,567][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:40:42,593][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:40:42,616][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:40:42,616][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:40:42,617][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:40:43,372][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:40:43,830][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:40:44,338][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:40:44,844][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:40:45,348][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:40:45,846][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:40:46,351][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:40:46,862][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:40:47,362][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:40:47,866][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:40:48,365][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:40:48,868][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:40:49,373][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:40:49,874][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:40:50,372][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:40:50,874][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:40:51,373][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:40:51,884][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:40:52,405][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:40:52,907][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:40:53,414][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:40:53,920][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:40:54,420][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:40:54,924][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:40:55,426][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:40:55,922][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:40:56,418][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:40:56,916][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:40:57,419][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:40:57,916][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:40:58,414][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:40:58,913][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:40:59,414][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:40:59,912][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:41:00,411][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:41:00,912][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:41:01,414][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:41:01,915][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:41:02,416][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:41:02,928][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:41:03,427][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:41:03,956][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:41:04,453][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:41:04,952][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:41:05,457][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:41:05,956][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:41:06,461][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:41:06,961][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:41:07,465][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:41:07,980][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:41:08,481][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:41:08,982][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:41:09,485][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:41:09,988][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:41:10,489][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:41:10,989][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:41:11,491][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:41:11,997][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:41:12,498][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:41:12,999][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:41:13,498][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:41:14,003][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:41:14,504][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:41:15,001][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:41:15,499][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10219 tokens. -[2025-11-09 22:41:16,230][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 55.63%, Block Peak % of device VRAM: 62.42%, ΔTime: 00:00:32 -[2025-11-09 22:41:16,971][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:41:16,975][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:41:16,977][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:41:17,974][__main__][INFO] - Iteration 315 took 1m 12s (49.67% Gen, 48.94% Train). Generation: 35s, Training: 35s. Estimated remaining time: 55h 18m 48s. Estimated total time: 60h 1m 9s. Time estimates for 10 more iterations: 12m 0s, 100 more iterations: 2h 0m 2s, 500 more iterations: 10h 0m 11s. -[2025-11-09 22:41:17,976][__main__][INFO] - Starting iteration 315. -[2025-11-09 22:41:18,440][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 31 and human policies 1. -[2025-11-09 22:41:18,441][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:41:43,668][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 22:41:47,242][__main__][INFO] - Number of regex retries in iteration 315: 1 -[2025-11-09 22:41:47,243][__main__][INFO] - agents played in iteration 315 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:41:48,085][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:41:48,113][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:41:48,137][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:41:48,160][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:41:48,160][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:41:48,161][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:41:48,904][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:41:49,365][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:41:49,876][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:41:50,383][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:41:50,883][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:41:51,394][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:41:51,895][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:41:52,397][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:41:52,898][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:41:53,401][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:41:53,904][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:41:54,408][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:41:54,909][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:41:55,414][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:41:55,917][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:41:56,419][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:41:56,923][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:41:57,425][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:41:57,926][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:41:58,428][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:41:58,930][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:41:59,435][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:41:59,939][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:42:00,447][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:42:00,948][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:42:01,452][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:42:01,973][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:42:02,473][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:42:02,976][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:42:03,478][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:42:03,982][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:42:04,486][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:42:04,991][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:42:05,495][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:42:06,009][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:42:06,514][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:42:07,015][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:42:07,518][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:42:08,021][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:42:08,528][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:42:09,032][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:42:09,534][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:42:10,034][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:42:10,536][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:42:11,042][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:42:11,542][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:42:12,045][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:42:12,564][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:42:13,064][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:42:13,576][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:42:14,079][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:42:14,581][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:42:15,097][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:42:15,604][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:42:16,109][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:42:16,615][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:42:17,120][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:42:17,622][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:42:18,122][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:42:18,625][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:42:19,129][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:42:19,628][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:42:20,128][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:42:20,628][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:42:21,131][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10284 tokens. -[2025-11-09 22:42:21,943][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 55.57%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:33 -[2025-11-09 22:42:22,585][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:42:22,587][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:42:22,589][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:42:23,536][__main__][INFO] - Iteration 316 took 1m 5s (44.25% Gen, 54.30% Train). Generation: 28s, Training: 35s. Estimated remaining time: 49h 31m 21s. Estimated total time: 54h 14m 48s. Time estimates for 10 more iterations: 10m 50s, 100 more iterations: 1h 48m 29s, 500 more iterations: 9h 2m 28s. -[2025-11-09 22:42:23,538][__main__][INFO] - Starting iteration 316. -[2025-11-09 22:42:24,039][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 31 and human policies 1. -[2025-11-09 22:42:24,040][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:42:52,656][__main__][INFO] - Number of regex retries in iteration 316: 0 -[2025-11-09 22:42:52,657][__main__][INFO] - agents played in iteration 316 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:42:53,467][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:42:53,493][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:42:53,517][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:42:53,541][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:42:53,542][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:42:53,543][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:42:54,289][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:42:54,748][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:42:55,254][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:42:55,759][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:42:56,261][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:42:56,762][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:42:57,261][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:42:57,761][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:42:58,278][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:42:58,784][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:42:59,289][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:42:59,794][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:43:00,300][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:43:00,810][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:43:01,309][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:43:01,807][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:43:02,310][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:43:02,811][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:43:03,308][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:43:03,819][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:43:04,321][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:43:04,821][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:43:05,326][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:43:05,828][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:43:06,332][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:43:06,840][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:43:07,345][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:43:07,848][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:43:08,352][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:43:08,854][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:43:09,366][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:43:09,874][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:43:10,402][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:43:10,902][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:43:11,410][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:43:11,912][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:43:12,418][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:43:12,921][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:43:13,424][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:43:13,928][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:43:14,435][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:43:14,945][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:43:17,091][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:43:17,682][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:43:18,187][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:43:18,691][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:43:19,195][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:43:19,717][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:43:20,222][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:43:20,728][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:43:21,227][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:43:21,730][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:43:22,235][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:43:22,741][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:43:23,242][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:43:23,747][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:43:24,252][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:43:24,754][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:43:25,279][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:43:25,784][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:43:26,290][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:43:26,794][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:43:27,299][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:43:27,803][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:43:28,312][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10432 tokens. -[2025-11-09 22:43:29,118][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.25%, Current % of VRAM taken: 55.74%, Block Peak % of device VRAM: 62.24%, ΔTime: 00:00:34 -[2025-11-09 22:43:29,777][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:43:29,779][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:43:29,782][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:43:30,744][__main__][INFO] - Iteration 317 took 1m 6s (42.90% Gen, 55.66% Train). Generation: 28s, Training: 37s. Estimated remaining time: 50h 50m 43s. Estimated total time: 55h 35m 16s. Time estimates for 10 more iterations: 11m 7s, 100 more iterations: 1h 51m 10s, 500 more iterations: 9h 15m 52s. -[2025-11-09 22:43:30,748][__main__][INFO] - Starting iteration 317. -[2025-11-09 22:43:31,222][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 31 and human policies 1. -[2025-11-09 22:43:31,223][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:43:47,624][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 22:44:00,709][__main__][INFO] - Number of regex retries in iteration 317: 1 -[2025-11-09 22:44:00,710][__main__][INFO] - agents played in iteration 317 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:44:01,554][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:44:01,581][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:44:01,607][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:44:01,629][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:44:01,631][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:44:01,631][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:44:02,374][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:44:02,960][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:44:03,463][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:44:03,968][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:44:04,484][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:44:04,989][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:44:05,502][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:44:06,001][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:44:06,500][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:44:07,018][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:44:07,520][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:44:08,021][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:44:08,523][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:44:09,026][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:44:09,526][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:44:10,025][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:44:10,527][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:44:11,036][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:44:11,538][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:44:12,040][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:44:12,546][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:44:13,050][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:44:13,556][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:44:14,063][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:44:14,581][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:44:15,092][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:44:15,598][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:44:16,102][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:44:16,609][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:44:17,110][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:44:17,630][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:44:18,135][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:44:18,639][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:44:19,141][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:44:19,645][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:44:20,147][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:44:20,647][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:44:21,152][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:44:21,657][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:44:22,160][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:44:22,662][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:44:23,163][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:44:23,666][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:44:24,170][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:44:24,675][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:44:25,177][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:44:25,682][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:44:26,184][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:44:26,690][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:44:27,194][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:44:27,699][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:44:28,219][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:44:28,721][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:44:29,225][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:44:29,728][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:44:30,228][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:44:30,740][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:44:31,247][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:44:31,752][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:44:32,256][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:44:32,758][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:44:33,259][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:44:33,761][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:44:34,262][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:44:34,764][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10309 tokens. -[2025-11-09 22:44:35,501][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 55.63%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:33 -[2025-11-09 22:44:36,245][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:44:36,246][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:44:36,248][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:44:37,280][__main__][INFO] - Iteration 318 took 1m 6s (44.63% Gen, 53.80% Train). Generation: 29s, Training: 35s. Estimated remaining time: 50h 17m 16s. Estimated total time: 55h 2m 56s. Time estimates for 10 more iterations: 11m 0s, 100 more iterations: 1h 50m 5s, 500 more iterations: 9h 10m 29s. -[2025-11-09 22:44:37,282][__main__][INFO] - Starting iteration 318. -[2025-11-09 22:44:37,748][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 31 and human policies 1. -[2025-11-09 22:44:37,749][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:45:02,059][__main__][INFO] - Number of regex retries in iteration 318: 0 -[2025-11-09 22:45:02,060][__main__][INFO] - agents played in iteration 318 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:45:03,001][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:45:03,025][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:45:03,048][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:45:03,070][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:45:03,071][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:45:03,071][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:45:03,792][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:45:04,251][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:45:04,760][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:45:05,259][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:45:05,761][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:45:06,263][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:45:06,764][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:45:07,265][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:45:07,776][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:45:08,280][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:45:08,783][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:45:09,288][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:45:09,792][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:45:10,293][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:45:10,800][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:45:11,306][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:45:11,806][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:45:12,306][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:45:12,823][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:45:13,323][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:45:13,829][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:45:14,330][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:45:14,832][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:45:15,333][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:45:15,832][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:45:17,206][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:45:18,157][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:45:18,662][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:45:19,162][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:45:19,663][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:45:20,170][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:45:20,674][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:45:21,174][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:45:21,680][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:45:22,181][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:45:22,685][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:45:23,211][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:45:23,711][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:45:24,215][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:45:24,724][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:45:25,231][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:45:25,731][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:45:26,232][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:45:26,733][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:45:27,244][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:45:27,747][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:45:28,252][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:45:28,756][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:45:29,258][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:45:29,759][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:45:30,257][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:45:30,756][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:45:31,259][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:45:31,759][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:45:32,258][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:45:32,759][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:45:33,259][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:45:33,764][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:45:34,272][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:45:34,773][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:45:35,274][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:45:35,774][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:45:36,278][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:45:36,781][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:45:37,283][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10195 tokens. -[2025-11-09 22:45:38,064][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 55.62%, Block Peak % of device VRAM: 62.20%, ΔTime: 00:00:34 -[2025-11-09 22:45:38,714][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:45:38,716][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:45:38,718][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:45:39,673][__main__][INFO] - Iteration 319 took 1m 1s (39.26% Gen, 59.20% Train). Generation: 24s, Training: 36s. Estimated remaining time: 46h 49m 32s. Estimated total time: 51h 36m 15s. Time estimates for 10 more iterations: 10m 19s, 100 more iterations: 1h 43m 12s, 500 more iterations: 8h 36m 2s. -[2025-11-09 22:45:39,675][__main__][INFO] - Starting iteration 319. -[2025-11-09 22:45:40,136][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 31 and human policies 1. -[2025-11-09 22:45:40,136][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:46:01,980][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 22:46:14,870][__main__][INFO] - Number of regex retries in iteration 319: 1 -[2025-11-09 22:46:14,870][__main__][INFO] - agents played in iteration 319 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:46:15,700][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:46:15,725][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:46:15,750][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:46:15,774][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:46:15,774][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:46:15,775][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:46:16,534][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:46:16,995][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:46:17,500][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:46:18,002][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:46:18,515][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:46:19,018][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:46:19,517][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:46:20,021][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:46:20,525][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:46:21,043][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:46:21,549][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:46:22,054][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:46:22,559][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:46:23,097][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:46:23,602][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:46:24,110][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:46:24,617][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:46:25,136][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:46:25,633][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:46:26,132][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:46:26,632][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:46:27,132][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:46:27,639][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:46:28,138][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:46:28,637][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:46:29,140][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:46:29,639][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:46:30,140][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:46:30,643][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:46:31,143][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:46:31,654][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:46:32,154][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:46:32,658][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:46:33,159][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:46:33,663][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:46:34,177][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:46:34,680][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:46:35,182][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:46:35,684][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:46:36,187][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:46:36,693][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:46:37,194][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:46:37,696][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:46:38,207][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:46:38,708][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:46:39,213][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:46:39,712][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:46:40,214][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:46:40,716][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:46:41,216][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:46:41,715][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:46:42,215][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:46:42,717][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:46:43,215][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:46:43,711][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:46:44,207][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:46:44,725][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:46:45,221][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:46:45,717][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:46:46,232][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:46:46,729][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:46:47,226][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:46:47,727][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:46:48,228][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:46:48,739][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10174 tokens. -[2025-11-09 22:46:49,452][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 55.60%, Block Peak % of device VRAM: 62.38%, ΔTime: 00:00:32 -[2025-11-09 22:46:50,183][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:46:50,184][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:46:50,186][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:46:51,194][__main__][INFO] - Iteration 320 took 1m 11s (48.88% Gen, 49.70% Train). Generation: 34s, Training: 35s. Estimated remaining time: 54h 25m 2s. Estimated total time: 59h 12m 56s. Time estimates for 10 more iterations: 11m 50s, 100 more iterations: 1h 58m 25s, 500 more iterations: 9h 52m 9s. -[2025-11-09 22:46:51,196][__main__][INFO] - Starting iteration 320. -[2025-11-09 22:46:51,662][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 31 and human policies 1. -[2025-11-09 22:46:51,663][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:47:00,070][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 22:47:00,167][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 22:47:00,859][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 22:47:13,522][__main__][INFO] - Number of regex retries in iteration 320: 3 -[2025-11-09 22:47:13,522][__main__][INFO] - agents played in iteration 320 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:47:14,340][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:47:14,364][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:47:14,388][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:47:14,412][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:47:14,412][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:47:14,413][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:47:15,222][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:47:15,758][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:47:17,536][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:47:18,114][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:47:18,620][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:47:19,136][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:47:19,639][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:47:20,144][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:47:20,656][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:47:21,164][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:47:21,667][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:47:22,168][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:47:22,670][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:47:23,175][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:47:23,685][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:47:24,189][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:47:24,694][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:47:25,198][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:47:25,700][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:47:26,205][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:47:26,710][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:47:27,216][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:47:27,720][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:47:28,224][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:47:28,730][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:47:29,234][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:47:29,754][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:47:30,259][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:47:30,759][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:47:31,259][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:47:31,762][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:47:32,263][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:47:32,765][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:47:33,265][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:47:33,768][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:47:34,267][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:47:34,764][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:47:35,261][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:47:35,761][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:47:36,260][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:47:36,759][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:47:37,258][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:47:37,759][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:47:38,260][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:47:38,766][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:47:39,275][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:47:39,775][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:47:40,281][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:47:40,785][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:47:41,290][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:47:41,791][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:47:42,293][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:47:42,799][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:47:43,298][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:47:43,800][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:47:44,301][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:47:44,807][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:47:45,307][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:47:45,809][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:47:46,311][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:47:46,812][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:47:47,311][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:47:47,811][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:47:48,317][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:47:48,822][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10219 tokens. -[2025-11-09 22:47:49,620][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 55.59%, Block Peak % of device VRAM: 62.21%, ΔTime: 00:00:34 -[2025-11-09 22:47:50,267][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:47:50,269][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:47:50,270][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:47:52,214][__main__][INFO] - Iteration 321 took 1m 0s (36.10% Gen, 60.69% Train). Generation: 21s, Training: 36s. Estimated remaining time: 45h 38m 41s. Estimated total time: 50h 27m 36s. Time estimates for 10 more iterations: 10m 5s, 100 more iterations: 1h 40m 55s, 500 more iterations: 8h 24m 36s. -[2025-11-09 22:47:52,216][__main__][INFO] - Starting iteration 321. -[2025-11-09 22:47:52,700][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 32 and human policies 1. -[2025-11-09 22:47:52,701][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:48:11,963][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 22:48:12,890][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 22:48:19,419][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 22:48:25,897][__main__][INFO] - Number of regex retries in iteration 321: 3 -[2025-11-09 22:48:25,898][__main__][INFO] - agents played in iteration 321 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:48:26,706][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:48:26,733][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:48:26,760][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:48:26,783][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:48:26,784][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:48:26,785][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:48:27,537][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:48:27,994][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:48:28,507][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:48:29,006][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:48:29,508][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:48:30,012][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:48:30,514][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:48:31,019][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:48:31,519][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:48:32,018][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:48:32,527][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:48:33,026][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:48:33,527][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:48:34,029][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:48:34,553][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:48:35,062][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:48:35,559][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:48:36,055][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:48:36,568][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:48:37,066][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:48:37,567][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:48:38,067][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:48:38,567][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:48:39,075][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:48:39,570][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:48:40,072][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:48:40,569][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:48:41,071][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:48:41,576][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:48:42,076][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:48:42,577][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:48:43,077][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:48:43,579][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:48:44,080][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:48:44,584][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:48:45,083][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:48:45,584][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:48:46,086][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:48:46,587][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:48:47,089][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:48:47,597][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:48:48,100][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:48:48,604][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:48:49,108][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:48:49,615][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:48:50,120][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:48:50,620][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:48:51,126][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:48:51,630][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:48:52,128][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:48:52,628][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:48:53,131][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:48:53,642][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:48:54,141][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:48:54,657][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:48:55,159][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:48:55,663][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:48:56,164][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:48:56,660][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:48:57,164][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:48:57,661][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:48:58,162][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:48:58,670][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:48:59,171][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:48:59,668][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10233 tokens. -[2025-11-09 22:49:00,400][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 55.66%, Block Peak % of device VRAM: 62.20%, ΔTime: 00:00:32 -[2025-11-09 22:49:01,133][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:49:01,134][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:49:01,136][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:49:02,155][__main__][INFO] - Iteration 322 took 1m 9s (47.80% Gen, 50.74% Train). Generation: 33s, Training: 35s. Estimated remaining time: 53h 2m 40s. Estimated total time: 57h 52m 45s. Time estimates for 10 more iterations: 11m 34s, 100 more iterations: 1h 55m 45s, 500 more iterations: 9h 38m 47s. -[2025-11-09 22:49:02,157][__main__][INFO] - Starting iteration 322. -[2025-11-09 22:49:02,797][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 32 and human policies 1. -[2025-11-09 22:49:02,798][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:49:27,694][__main__][INFO] - Number of regex retries in iteration 322: 0 -[2025-11-09 22:49:27,696][__main__][INFO] - agents played in iteration 322 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:49:28,537][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:49:28,562][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:49:28,599][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:49:28,626][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:49:28,628][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:49:28,628][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:49:29,430][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:49:30,046][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:49:30,556][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:49:31,058][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:49:31,561][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:49:32,060][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:49:32,563][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:49:33,070][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:49:33,576][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:49:34,098][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:49:34,604][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:49:35,107][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:49:35,613][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:49:36,116][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:49:36,622][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:49:37,127][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:49:37,633][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:49:38,138][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:49:38,644][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:49:39,147][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:49:39,665][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:49:40,168][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:49:40,691][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:49:41,197][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:49:41,703][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:49:42,204][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:49:42,705][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:49:43,212][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:49:43,714][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:49:44,215][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:49:44,717][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:49:45,218][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:49:45,721][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:49:46,227][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:49:46,732][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:49:47,237][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:49:47,741][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:49:48,245][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:49:48,752][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:49:49,257][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:49:49,760][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:49:50,262][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:49:50,761][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:49:51,264][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:49:51,763][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:49:52,270][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:49:52,773][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:49:53,274][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:49:53,787][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:49:54,287][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:49:54,786][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:49:55,293][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:49:55,796][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:49:56,299][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:49:56,801][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:49:57,303][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:49:57,818][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:49:58,325][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:49:58,824][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:49:59,327][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:49:59,835][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:50:00,337][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:50:00,841][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:50:01,339][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:50:01,840][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10344 tokens. -[2025-11-09 22:50:02,611][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.05%, Current % of VRAM taken: 55.54%, Block Peak % of device VRAM: 62.22%, ΔTime: 00:00:33 -[2025-11-09 22:50:03,265][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:50:03,267][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:50:03,269][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:50:04,217][__main__][INFO] - Iteration 323 took 1m 1s (40.54% Gen, 57.92% Train). Generation: 24s, Training: 35s. Estimated remaining time: 46h 19m 54s. Estimated total time: 51h 11m 1s. Time estimates for 10 more iterations: 10m 14s, 100 more iterations: 1h 42m 22s, 500 more iterations: 8h 31m 50s. -[2025-11-09 22:50:04,219][__main__][INFO] - Starting iteration 323. -[2025-11-09 22:50:04,704][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 32 and human policies 1. -[2025-11-09 22:50:04,704][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:50:21,272][mllm.models.large_language_model_local][WARNING] - Response Proposals: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 22:50:34,491][__main__][INFO] - Number of regex retries in iteration 323: 1 -[2025-11-09 22:50:34,492][__main__][INFO] - agents played in iteration 323 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:50:35,302][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:50:35,328][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:50:35,354][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:50:35,378][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:50:35,379][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:50:35,380][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:50:36,145][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:50:36,607][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:50:37,120][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:50:37,621][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:50:38,125][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:50:38,633][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:50:39,133][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:50:39,641][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:50:40,144][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:50:40,651][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:50:41,167][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:50:41,668][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:50:42,168][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:50:42,670][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:50:43,171][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:50:43,673][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:50:44,173][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:50:44,675][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:50:45,175][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:50:45,674][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:50:46,199][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:50:46,705][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:50:47,206][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:50:47,706][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:50:48,202][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:50:48,701][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:50:49,199][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:50:49,697][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:50:50,206][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:50:50,703][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:50:51,201][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:50:51,698][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:50:52,200][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:50:52,705][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:50:53,204][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:50:53,703][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:50:54,204][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:50:54,702][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:50:55,206][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:50:55,707][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:50:56,210][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:50:56,715][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:50:57,218][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:50:57,720][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:50:58,222][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:50:58,723][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:50:59,228][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:50:59,728][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:51:00,227][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:51:00,731][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:51:01,233][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:51:01,732][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:51:02,232][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:51:02,734][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:51:03,236][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:51:03,735][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:51:04,237][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:51:04,739][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:51:05,237][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:51:05,740][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:51:06,240][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:51:06,739][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:51:07,236][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:51:07,734][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:51:08,236][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10119 tokens. -[2025-11-09 22:51:08,938][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.27%, Current % of VRAM taken: 55.76%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 -[2025-11-09 22:51:09,672][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:51:09,673][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:51:09,675][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:51:10,680][__main__][INFO] - Iteration 324 took 1m 5s (45.15% Gen, 53.33% Train). Generation: 29s, Training: 35s. Estimated remaining time: 50h 6m 34s. Estimated total time: 54h 58m 48s. Time estimates for 10 more iterations: 10m 59s, 100 more iterations: 1h 49m 57s, 500 more iterations: 9h 9m 48s. -[2025-11-09 22:51:10,682][__main__][INFO] - Starting iteration 324. -[2025-11-09 22:51:11,168][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 32 and human policies 1. -[2025-11-09 22:51:11,168][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:51:27,724][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 22:51:27,778][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 22:51:40,764][__main__][INFO] - Number of regex retries in iteration 324: 2 -[2025-11-09 22:51:40,764][__main__][INFO] - agents played in iteration 324 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:51:41,584][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:51:41,607][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:51:41,631][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:51:41,656][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:51:41,657][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:51:41,658][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:51:42,460][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:51:42,924][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:51:43,428][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:51:43,929][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:51:44,431][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:51:44,933][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:51:45,433][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:51:45,937][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:51:46,442][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:51:46,942][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:51:47,443][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:51:47,952][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:51:48,455][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:51:48,951][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:51:49,452][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:51:49,949][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:51:50,448][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:51:50,947][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:51:51,446][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:51:51,950][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:51:52,449][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:51:52,948][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:51:53,453][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:51:53,953][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:51:54,454][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:51:54,954][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:51:55,456][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:51:55,956][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:51:56,463][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:51:56,965][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:51:57,468][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:51:57,972][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:51:58,476][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:51:58,976][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:51:59,491][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:51:59,997][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:52:00,499][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:52:01,002][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:52:01,503][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:52:02,007][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:52:02,509][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:52:03,012][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:52:03,520][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:52:04,025][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:52:04,528][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:52:05,030][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:52:05,530][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:52:06,030][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:52:06,532][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:52:07,036][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:52:07,539][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:52:08,040][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:52:08,540][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:52:09,041][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:52:09,541][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:52:10,044][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:52:10,545][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:52:11,068][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:52:11,580][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:52:12,084][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:52:12,587][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:52:13,098][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:52:13,600][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:52:14,111][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:52:14,611][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10224 tokens. -[2025-11-09 22:52:15,346][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 55.66%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:32 -[2025-11-09 22:52:15,971][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:52:15,972][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:52:15,974][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:52:16,896][__main__][INFO] - Iteration 325 took 1m 5s (45.03% Gen, 53.57% Train). Generation: 29s, Training: 35s. Estimated remaining time: 49h 53m 6s. Estimated total time: 54h 46m 26s. Time estimates for 10 more iterations: 10m 57s, 100 more iterations: 1h 49m 32s, 500 more iterations: 9h 7m 44s. -[2025-11-09 22:52:16,898][__main__][INFO] - Starting iteration 325. -[2025-11-09 22:52:17,376][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 32 and human policies 1. -[2025-11-09 22:52:17,378][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:52:32,032][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 22:52:40,058][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 22:52:40,957][__main__][INFO] - Number of regex retries in iteration 325: 2 -[2025-11-09 22:52:40,958][__main__][INFO] - agents played in iteration 325 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:52:41,784][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:52:41,807][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:52:41,831][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:52:41,854][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:52:41,855][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:52:41,856][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:52:42,648][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:52:43,112][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:52:43,623][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:52:44,127][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:52:44,630][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:52:45,134][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:52:45,637][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:52:46,140][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:52:46,640][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:52:47,145][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:52:47,646][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:52:48,146][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:52:48,645][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:52:49,144][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:52:49,650][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:52:50,148][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:52:50,646][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:52:51,169][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:52:51,674][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:52:52,176][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:52:52,681][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:52:53,178][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:52:53,694][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:52:54,199][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:52:54,700][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:52:55,206][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:52:55,711][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:52:56,213][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:52:56,715][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:52:57,214][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:52:57,717][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:52:58,215][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:52:58,718][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:52:59,221][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:52:59,720][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:53:00,231][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:53:00,738][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:53:01,242][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:53:01,747][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:53:02,250][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:53:02,757][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:53:03,255][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:53:03,757][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:53:04,264][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:53:04,765][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:53:05,270][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:53:05,772][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:53:06,275][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:53:06,782][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:53:07,285][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:53:07,789][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:53:08,299][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:53:08,800][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:53:09,308][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:53:09,808][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:53:10,314][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:53:10,817][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:53:11,319][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:53:11,822][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:53:12,327][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:53:13,292][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:53:14,643][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:53:15,145][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:53:15,647][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:53:16,154][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10411 tokens. -[2025-11-09 22:53:16,963][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 55.72%, Block Peak % of device VRAM: 62.27%, ΔTime: 00:00:34 -[2025-11-09 22:53:17,642][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:53:17,643][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:53:17,646][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:53:18,646][__main__][INFO] - Iteration 326 took 1m 1s (38.48% Gen, 59.88% Train). Generation: 23s, Training: 36s. Estimated remaining time: 46h 9m 9s. Estimated total time: 51h 3m 30s. Time estimates for 10 more iterations: 10m 12s, 100 more iterations: 1h 42m 7s, 500 more iterations: 8h 30m 35s. -[2025-11-09 22:53:18,650][__main__][INFO] - Starting iteration 326. -[2025-11-09 22:53:19,117][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 32 and human policies 1. -[2025-11-09 22:53:19,117][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:53:50,466][__main__][INFO] - Number of regex retries in iteration 326: 0 -[2025-11-09 22:53:50,466][__main__][INFO] - agents played in iteration 326 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:53:51,287][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:53:51,311][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:53:51,337][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:53:51,360][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:53:51,361][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:53:51,362][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:53:52,130][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:53:52,591][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:53:53,106][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:53:53,612][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:53:54,129][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:53:54,633][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:53:55,138][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:53:55,644][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:53:56,146][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:53:56,650][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:53:57,150][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:53:57,654][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:53:58,158][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:53:58,659][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:53:59,161][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:53:59,660][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:54:00,159][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:54:00,659][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:54:01,160][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:54:01,660][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:54:02,166][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:54:02,664][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:54:03,165][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:54:03,666][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:54:04,169][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:54:04,719][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:54:05,220][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:54:05,723][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:54:06,225][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:54:06,728][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:54:07,234][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:54:07,736][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:54:08,243][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:54:08,749][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:54:09,249][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:54:09,750][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:54:10,255][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:54:10,757][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:54:11,262][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:54:11,764][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:54:12,268][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:54:12,774][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:54:13,273][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:54:13,776][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:54:14,280][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:54:14,782][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:54:15,288][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:54:15,791][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:54:16,297][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:54:16,802][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:54:17,304][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:54:17,806][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:54:18,309][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:54:18,812][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:54:19,320][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:54:19,824][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:54:20,337][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:54:20,836][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:54:21,335][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:54:21,835][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:54:22,336][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:54:22,836][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:54:23,338][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:54:23,839][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:54:24,352][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10377 tokens. -[2025-11-09 22:54:25,153][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.38%, ΔTime: 00:00:33 -[2025-11-09 22:54:25,923][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:54:25,925][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:54:25,927][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:54:27,044][__main__][INFO] - Iteration 327 took 1m 7s (46.15% Gen, 52.20% Train). Generation: 31s, Training: 35s. Estimated remaining time: 51h 40m 52s. Estimated total time: 56h 36m 22s. Time estimates for 10 more iterations: 11m 19s, 100 more iterations: 1h 53m 12s, 500 more iterations: 9h 26m 3s. -[2025-11-09 22:54:27,046][__main__][INFO] - Starting iteration 327. -[2025-11-09 22:54:27,508][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 32 and human policies 1. -[2025-11-09 22:54:27,509][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:54:52,983][__main__][INFO] - Number of regex retries in iteration 327: 0 -[2025-11-09 22:54:52,985][__main__][INFO] - agents played in iteration 327 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:54:53,831][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:54:53,862][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:54:53,889][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:54:53,913][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:54:53,913][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:54:53,914][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:54:54,719][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:54:55,185][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:54:55,692][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:54:56,200][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:54:56,702][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:54:57,210][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:54:57,715][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:54:58,217][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:54:58,722][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:54:59,224][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:54:59,747][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:55:00,247][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:55:00,747][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:55:01,247][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:55:01,752][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:55:02,263][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:55:02,762][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:55:03,268][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:55:03,777][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:55:04,283][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:55:04,792][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:55:05,295][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:55:05,796][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:55:06,299][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:55:06,799][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:55:07,298][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:55:07,800][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:55:08,299][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:55:08,801][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:55:09,302][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:55:09,804][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:55:10,308][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:55:10,808][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:55:11,310][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:55:11,811][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:55:12,315][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:55:12,817][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:55:13,317][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:55:13,816][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:55:14,318][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:55:14,821][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:55:15,325][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:55:15,822][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:55:16,324][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:55:16,838][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:55:17,339][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:55:17,847][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:55:18,347][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:55:18,851][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:55:19,356][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:55:19,859][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:55:20,361][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:55:20,861][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:55:21,360][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:55:21,874][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:55:22,374][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:55:22,879][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:55:23,381][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:55:23,886][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:55:24,390][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:55:24,895][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:55:25,400][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:55:25,906][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:55:26,411][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:55:26,921][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10315 tokens. -[2025-11-09 22:55:27,684][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 -[2025-11-09 22:55:28,348][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:55:28,350][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:55:28,352][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:55:29,325][__main__][INFO] - Iteration 328 took 1m 1s (41.21% Gen, 57.21% Train). Generation: 25s, Training: 35s. Estimated remaining time: 46h 34m 20s. Estimated total time: 51h 30m 53s. Time estimates for 10 more iterations: 10m 18s, 100 more iterations: 1h 43m 1s, 500 more iterations: 8h 35m 8s. -[2025-11-09 22:55:29,328][__main__][INFO] - Starting iteration 328. -[2025-11-09 22:55:30,060][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 32 and human policies 1. -[2025-11-09 22:55:30,061][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:56:03,789][__main__][INFO] - Number of regex retries in iteration 328: 0 -[2025-11-09 22:56:03,789][__main__][INFO] - agents played in iteration 328 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:56:04,578][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:56:04,615][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:56:04,637][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:56:04,659][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:56:04,660][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:56:04,661][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:56:05,365][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:56:05,821][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:56:06,330][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:56:06,830][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:56:07,335][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:56:07,834][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:56:08,336][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:56:08,849][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:56:09,348][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:56:09,847][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:56:10,346][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:56:10,845][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:56:11,352][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:56:11,852][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:56:12,350][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:56:12,870][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:56:13,367][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:56:13,865][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:56:14,364][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:56:14,862][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:56:15,377][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:56:15,875][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:56:16,374][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:56:16,873][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:56:17,373][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:56:17,882][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:56:18,387][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:56:18,890][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:56:19,397][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:56:19,896][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:56:20,395][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:56:20,895][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:56:21,398][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:56:21,902][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:56:22,399][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:56:22,899][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:56:23,403][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:56:23,902][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:56:24,404][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:56:24,910][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:56:25,410][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:56:25,945][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:56:26,447][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:56:26,946][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:56:27,448][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:56:27,955][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:56:28,460][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:56:28,966][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:56:29,474][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:56:29,991][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:56:30,496][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:56:30,997][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:56:31,503][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:56:32,003][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:56:32,503][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:56:33,005][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:56:33,506][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:56:34,010][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:56:34,511][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:56:35,014][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:56:35,520][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:56:36,022][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:56:36,525][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:56:37,028][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:56:37,529][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10217 tokens. -[2025-11-09 22:56:38,241][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 -[2025-11-09 22:56:38,966][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:56:38,968][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:56:38,970][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:56:40,059][__main__][INFO] - Iteration 329 took 1m 9s (48.18% Gen, 50.26% Train). Generation: 33s, Training: 35s. Estimated remaining time: 53h 22m 13s. Estimated total time: 58h 19m 57s. Time estimates for 10 more iterations: 11m 39s, 100 more iterations: 1h 56m 39s, 500 more iterations: 9h 43m 19s. -[2025-11-09 22:56:40,061][__main__][INFO] - Starting iteration 329. -[2025-11-09 22:56:40,531][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 32 and human policies 1. -[2025-11-09 22:56:40,531][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:56:55,682][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 22:57:05,300][__main__][INFO] - Number of regex retries in iteration 329: 1 -[2025-11-09 22:57:05,300][__main__][INFO] - agents played in iteration 329 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:57:06,107][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:57:06,134][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:57:06,161][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:57:06,184][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:57:06,184][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:57:06,185][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:57:06,940][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:57:07,400][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:57:07,904][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:57:08,406][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:57:08,907][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:57:09,407][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:57:09,910][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:57:10,409][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:57:10,917][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:57:11,422][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:57:11,922][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:57:12,422][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:57:12,922][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:57:13,421][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:57:13,925][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:57:14,423][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:57:14,928][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:57:15,425][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:57:15,926][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:57:16,425][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:57:16,923][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:57:17,432][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:57:17,933][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:57:18,435][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:57:18,936][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:57:19,433][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:57:19,930][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:57:20,428][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:57:20,925][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:57:21,421][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:57:21,917][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:57:22,415][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:57:22,916][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:57:23,414][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:57:23,943][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:57:24,443][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:57:24,946][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:57:25,444][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:57:25,947][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:57:26,450][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:57:26,963][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:57:27,469][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:57:27,981][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:57:28,481][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:57:28,980][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:57:29,485][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:57:29,979][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:57:30,480][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:57:30,983][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:57:31,493][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:57:32,002][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:57:32,501][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:57:33,007][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:57:33,509][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:57:34,014][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:57:34,518][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:57:35,020][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:57:35,520][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:57:36,033][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:57:36,536][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:57:37,038][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:57:37,541][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:57:38,044][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:57:38,546][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:57:39,049][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10292 tokens. -[2025-11-09 22:57:39,847][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 55.66%, Block Peak % of device VRAM: 62.21%, ΔTime: 00:00:32 -[2025-11-09 22:57:40,487][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:57:40,489][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:57:40,491][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:57:41,433][__main__][INFO] - Iteration 330 took 1m 0s (40.67% Gen, 57.78% Train). Generation: 24s, Training: 35s. Estimated remaining time: 45h 46m 24s. Estimated total time: 50h 45m 8s. Time estimates for 10 more iterations: 10m 9s, 100 more iterations: 1h 41m 30s, 500 more iterations: 8h 27m 31s. -[2025-11-09 22:57:41,435][__main__][INFO] - Starting iteration 330. -[2025-11-09 22:57:41,902][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 32 and human policies 1. -[2025-11-09 22:57:41,903][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:58:16,366][__main__][INFO] - Number of regex retries in iteration 330: 0 -[2025-11-09 22:58:16,367][__main__][INFO] - agents played in iteration 330 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:58:17,214][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:58:17,243][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:58:17,270][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:58:17,293][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:58:17,294][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:58:17,295][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:58:18,036][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:58:18,496][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:58:18,999][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:58:19,498][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:58:19,995][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:58:20,507][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:58:21,014][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:58:21,514][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:58:22,014][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:58:22,519][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:58:23,045][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:58:23,555][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:58:24,055][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:58:24,559][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:58:25,061][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:58:25,566][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:58:26,082][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:58:26,583][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:58:27,088][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:58:27,587][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:58:28,087][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:58:28,588][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:58:29,088][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:58:29,587][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:58:30,091][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:58:30,592][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:58:31,089][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:58:31,586][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:58:32,084][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:58:32,593][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:58:33,091][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:58:33,590][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:58:34,091][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:58:34,590][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:58:35,091][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:58:35,591][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:58:36,091][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:58:36,593][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:58:37,096][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:58:37,605][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:58:38,103][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:58:38,608][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:58:39,112][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:58:39,615][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:58:40,113][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:58:40,617][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:58:41,119][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:58:41,623][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:58:42,129][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:58:42,631][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:58:43,132][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:58:43,633][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:58:44,136][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:58:44,659][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:58:46,724][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:58:47,227][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:58:47,728][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:58:48,231][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:58:48,732][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:58:49,235][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:58:49,737][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:58:50,238][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 22:58:50,739][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 22:58:51,241][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 22:58:51,742][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10271 tokens. -[2025-11-09 22:58:52,578][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 55.67%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:34 -[2025-11-09 22:58:53,249][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 22:58:53,250][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 22:58:53,252][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 22:58:55,147][__main__][INFO] - Iteration 331 took 1m 13s (47.05% Gen, 50.36% Train). Generation: 34s, Training: 36s. Estimated remaining time: 56h 2m 19s. Estimated total time: 61h 2m 17s. Time estimates for 10 more iterations: 12m 12s, 100 more iterations: 2h 2m 4s, 500 more iterations: 10h 10m 22s. -[2025-11-09 22:58:55,150][__main__][INFO] - Starting iteration 331. -[2025-11-09 22:58:55,611][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 33 and human policies 1. -[2025-11-09 22:58:55,612][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 22:59:27,376][__main__][INFO] - Number of regex retries in iteration 331: 0 -[2025-11-09 22:59:27,376][__main__][INFO] - agents played in iteration 331 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 22:59:28,163][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:59:28,190][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:59:28,215][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:59:28,238][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 22:59:28,239][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 22:59:28,239][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 22:59:28,929][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 22:59:29,516][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 22:59:30,139][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 22:59:30,644][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 22:59:31,153][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 22:59:31,653][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 22:59:32,156][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 22:59:32,652][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 22:59:33,150][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 22:59:33,650][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 22:59:34,154][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 22:59:34,652][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 22:59:35,155][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 22:59:35,654][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 22:59:36,153][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 22:59:36,653][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 22:59:37,153][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 22:59:37,652][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 22:59:38,150][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 22:59:38,648][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 22:59:39,143][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 22:59:39,638][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 22:59:40,139][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 22:59:40,641][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 22:59:41,138][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 22:59:41,637][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 22:59:42,132][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 22:59:42,633][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 22:59:43,133][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 22:59:43,631][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 22:59:44,132][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 22:59:44,632][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 22:59:45,135][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 22:59:45,636][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 22:59:46,139][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 22:59:46,643][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 22:59:47,143][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 22:59:47,646][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 22:59:48,145][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 22:59:48,651][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 22:59:49,157][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 22:59:49,656][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 22:59:50,161][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 22:59:50,662][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 22:59:51,165][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 22:59:51,664][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 22:59:52,167][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 22:59:52,676][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 22:59:53,180][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 22:59:53,681][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 22:59:54,182][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 22:59:54,687][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 22:59:55,189][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 22:59:55,691][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 22:59:56,207][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 22:59:56,712][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 22:59:57,228][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 22:59:57,728][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 22:59:58,230][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 22:59:58,736][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 22:59:59,237][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 22:59:59,740][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:00:00,240][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:00:00,739][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:00:01,246][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10287 tokens. -[2025-11-09 23:00:02,058][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 55.57%, Block Peak % of device VRAM: 62.24%, ΔTime: 00:00:33 -[2025-11-09 23:00:02,826][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:00:02,828][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:00:02,830][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:00:03,827][__main__][INFO] - Iteration 332 took 1m 8s (46.56% Gen, 51.97% Train). Generation: 31s, Training: 35s. Estimated remaining time: 51h 49m 42s. Estimated total time: 56h 50m 49s. Time estimates for 10 more iterations: 11m 22s, 100 more iterations: 1h 53m 41s, 500 more iterations: 9h 28m 28s. -[2025-11-09 23:00:03,829][__main__][INFO] - Starting iteration 332. -[2025-11-09 23:00:04,292][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 33 and human policies 1. -[2025-11-09 23:00:04,293][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:00:13,750][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 23:00:25,746][__main__][INFO] - Number of regex retries in iteration 332: 1 -[2025-11-09 23:00:25,747][__main__][INFO] - agents played in iteration 332 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:00:26,530][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:00:26,556][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:00:26,581][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:00:26,604][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:00:26,605][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:00:26,606][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:00:27,300][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:00:27,752][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:00:28,255][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:00:28,759][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:00:29,258][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:00:29,755][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:00:30,251][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:00:30,752][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:00:31,254][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:00:31,754][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:00:32,258][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:00:32,761][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:00:33,257][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:00:33,756][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:00:34,253][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:00:34,748][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:00:35,253][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:00:35,748][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:00:36,244][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:00:36,755][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:00:37,251][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:00:37,745][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:00:38,240][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:00:38,736][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:00:39,243][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:00:39,741][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:00:40,236][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:00:40,735][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:00:41,232][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:00:41,729][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:00:42,230][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:00:42,726][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:00:43,247][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:00:43,743][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:00:44,247][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:00:44,750][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:00:45,251][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:00:45,753][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:00:46,250][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:00:46,751][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:00:47,258][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:00:47,782][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:00:48,282][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:00:48,782][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:00:49,283][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:00:49,788][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:00:50,295][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:00:50,799][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:00:51,302][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:00:51,802][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:00:52,304][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:00:52,807][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:00:53,310][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:00:53,823][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:00:54,326][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:00:54,831][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:00:55,331][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:00:55,836][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:00:56,340][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:00:56,844][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:00:57,345][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:00:57,847][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:00:58,351][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:00:58,853][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:00:59,356][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10249 tokens. -[2025-11-09 23:01:00,176][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 55.68%, Block Peak % of device VRAM: 62.28%, ΔTime: 00:00:32 -[2025-11-09 23:01:00,817][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:01:00,820][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:01:00,822][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:01:01,769][__main__][INFO] - Iteration 333 took 57s (37.33% Gen, 61.02% Train). Generation: 21s, Training: 35s. Estimated remaining time: 42h 51m 47s. Estimated total time: 47h 53m 52s. Time estimates for 10 more iterations: 9m 34s, 100 more iterations: 1h 35m 47s, 500 more iterations: 7h 58m 58s. -[2025-11-09 23:01:01,772][__main__][INFO] - Starting iteration 333. -[2025-11-09 23:01:02,234][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 33 and human policies 1. -[2025-11-09 23:01:02,234][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:01:15,462][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 23:01:29,543][__main__][INFO] - Number of regex retries in iteration 333: 1 -[2025-11-09 23:01:29,544][__main__][INFO] - agents played in iteration 333 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:01:30,456][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:01:30,484][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:01:30,511][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:01:30,535][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:01:30,535][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:01:30,537][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:01:31,220][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:01:31,680][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:01:32,189][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:01:32,691][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:01:33,192][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:01:33,692][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:01:34,203][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:01:34,702][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:01:35,201][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:01:35,702][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:01:36,200][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:01:36,704][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:01:37,204][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:01:37,705][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:01:38,207][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:01:38,708][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:01:39,205][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:01:39,702][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:01:40,202][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:01:40,697][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:01:41,195][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:01:41,691][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:01:42,194][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:01:42,695][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:01:43,191][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:01:43,692][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:01:44,194][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:01:44,693][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:01:45,189][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:01:45,687][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:01:46,188][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:01:46,687][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:01:47,189][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:01:47,687][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:01:48,184][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:01:48,687][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:01:49,189][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:01:49,685][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:01:50,183][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:01:50,680][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:01:51,177][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:01:51,676][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:01:52,174][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:01:52,668][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:01:53,167][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:01:53,665][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:01:54,163][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:01:54,660][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:01:55,178][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:01:55,673][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:01:56,171][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:01:56,675][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:01:57,178][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:01:57,680][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:01:58,182][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:01:58,682][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:01:59,191][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:01:59,691][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:02:00,196][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:02:00,698][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:02:01,199][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:02:01,704][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:02:02,208][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:02:02,710][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:02:03,216][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10321 tokens. -[2025-11-09 23:02:04,030][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 55.61%, Block Peak % of device VRAM: 62.33%, ΔTime: 00:00:32 -[2025-11-09 23:02:04,768][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:02:04,769][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:02:04,771][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:02:05,780][__main__][INFO] - Iteration 334 took 1m 3s (42.97% Gen, 55.43% Train). Generation: 27s, Training: 35s. Estimated remaining time: 47h 54m 13s. Estimated total time: 52h 57m 22s. Time estimates for 10 more iterations: 10m 35s, 100 more iterations: 1h 45m 54s, 500 more iterations: 8h 49m 33s. -[2025-11-09 23:02:05,783][__main__][INFO] - Starting iteration 334. -[2025-11-09 23:02:06,267][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 33 and human policies 1. -[2025-11-09 23:02:06,268][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:02:15,811][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 23:02:29,386][__main__][INFO] - Number of regex retries in iteration 334: 1 -[2025-11-09 23:02:29,387][__main__][INFO] - agents played in iteration 334 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:02:30,211][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:02:30,235][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:02:30,258][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:02:30,282][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:02:30,283][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:02:30,284][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:02:30,993][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:02:31,448][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:02:31,948][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:02:32,451][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:02:32,951][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:02:33,452][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:02:33,949][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:02:34,448][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:02:34,944][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:02:35,442][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:02:35,948][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:02:36,448][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:02:36,961][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:02:37,463][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:02:37,962][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:02:38,468][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:02:38,963][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:02:39,472][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:02:39,974][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:02:40,470][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:02:40,975][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:02:41,470][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:02:41,968][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:02:42,472][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:02:42,972][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:02:43,488][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:02:43,988][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:02:44,488][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:02:45,007][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:02:45,512][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:02:46,019][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:02:46,519][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:02:47,019][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:02:47,517][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:02:48,016][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:02:48,516][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:02:49,013][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:02:49,513][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:02:50,009][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:02:50,507][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:02:51,004][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:02:51,504][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:02:52,003][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:02:52,503][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:02:52,998][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:02:53,498][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:02:53,999][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:02:54,497][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:02:54,996][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:02:55,492][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:02:55,988][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:02:56,494][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:02:56,992][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:02:57,493][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:02:57,991][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:02:58,492][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:02:58,999][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:02:59,507][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:03:00,010][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:03:00,514][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:03:01,019][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:03:01,520][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:03:02,023][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:03:02,527][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:03:03,031][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10275 tokens. -[2025-11-09 23:03:03,833][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 55.68%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:32 -[2025-11-09 23:03:04,503][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:03:04,505][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:03:04,507][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:03:05,471][__main__][INFO] - Iteration 335 took 59s (39.05% Gen, 59.32% Train). Generation: 23s, Training: 35s. Estimated remaining time: 44h 16m 6s. Estimated total time: 49h 20m 14s. Time estimates for 10 more iterations: 9m 52s, 100 more iterations: 1h 38m 40s, 500 more iterations: 8h 13m 22s. -[2025-11-09 23:03:05,473][__main__][INFO] - Starting iteration 335. -[2025-11-09 23:03:05,939][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 33 and human policies 1. -[2025-11-09 23:03:05,940][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:03:35,374][__main__][INFO] - Number of regex retries in iteration 335: 0 -[2025-11-09 23:03:35,375][__main__][INFO] - agents played in iteration 335 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:03:36,226][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:03:36,252][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:03:36,278][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:03:36,300][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:03:36,301][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:03:36,301][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:03:36,995][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:03:37,449][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:03:37,949][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:03:38,446][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:03:38,948][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:03:39,448][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:03:39,945][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:03:40,441][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:03:40,942][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:03:41,440][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:03:41,937][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:03:42,440][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:03:42,943][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:03:43,444][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:03:43,944][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:03:44,442][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:03:44,941][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:03:45,437][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:03:45,936][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:03:46,433][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:03:46,928][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:03:47,426][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:03:47,925][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:03:48,419][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:03:48,920][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:03:49,417][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:03:49,916][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:03:50,411][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:03:50,908][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:03:51,407][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:03:51,902][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:03:52,400][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:03:52,905][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:03:53,402][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:03:53,899][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:03:54,396][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:03:54,893][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:03:55,402][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:03:55,899][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:03:56,397][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:03:56,923][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:03:57,420][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:03:57,923][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:03:58,420][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:03:58,922][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:03:59,421][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:03:59,918][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:04:00,418][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:04:00,921][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:04:01,417][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:04:01,918][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:04:02,417][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:04:02,916][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:04:03,414][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:04:03,910][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:04:04,413][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:04:04,913][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:04:05,413][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:04:05,918][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:04:06,423][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:04:06,924][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:04:07,425][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:04:07,929][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:04:08,431][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:04:08,932][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10219 tokens. -[2025-11-09 23:04:09,755][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 55.63%, Block Peak % of device VRAM: 62.21%, ΔTime: 00:00:32 -[2025-11-09 23:04:10,467][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:04:10,469][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:04:10,471][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:04:11,491][__main__][INFO] - Iteration 336 took 1m 5s (44.90% Gen, 53.54% Train). Generation: 29s, Training: 35s. Estimated remaining time: 49h 32m 23s. Estimated total time: 54h 37m 37s. Time estimates for 10 more iterations: 10m 55s, 100 more iterations: 1h 49m 15s, 500 more iterations: 9h 6m 16s. -[2025-11-09 23:04:11,494][__main__][INFO] - Starting iteration 336. -[2025-11-09 23:04:11,963][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 33 and human policies 1. -[2025-11-09 23:04:11,964][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:04:31,806][__main__][INFO] - Number of regex retries in iteration 336: 0 -[2025-11-09 23:04:31,808][__main__][INFO] - agents played in iteration 336 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:04:32,624][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:04:32,648][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:04:32,671][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:04:32,694][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:04:32,695][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:04:32,695][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:04:33,469][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:04:33,929][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:04:34,436][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:04:34,943][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:04:35,447][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:04:35,950][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:04:36,458][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:04:36,959][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:04:37,460][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:04:37,961][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:04:38,469][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:04:38,981][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:04:39,479][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:04:39,982][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:04:40,481][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:04:40,981][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:04:41,481][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:04:41,980][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:04:42,482][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:04:42,991][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:04:43,492][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:04:43,995][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:04:44,495][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:04:44,996][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:04:45,511][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:04:46,013][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:04:46,515][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:04:47,014][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:04:47,514][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:04:48,022][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:04:48,522][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:04:49,022][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:04:49,519][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:04:50,018][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:04:50,519][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:04:51,022][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:04:51,521][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:04:52,027][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:04:52,550][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:04:53,055][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:04:53,561][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:04:54,066][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:04:54,568][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:04:55,066][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:04:55,563][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:04:56,062][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:04:56,560][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:04:57,059][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:04:57,561][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:04:58,060][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:04:58,562][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:04:59,063][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:04:59,566][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:05:00,068][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:05:00,571][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:05:01,081][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:05:01,580][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:05:02,086][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:05:02,605][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:05:03,110][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:05:03,617][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:05:04,120][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:05:04,621][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:05:05,130][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:05:05,632][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10307 tokens. -[2025-11-09 23:05:06,485][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 55.62%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:33 -[2025-11-09 23:05:07,137][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:05:07,138][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:05:07,140][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:05:08,127][__main__][INFO] - Iteration 337 took 56s (35.33% Gen, 62.91% Train). Generation: 19s, Training: 35s. Estimated remaining time: 41h 42m 3s. Estimated total time: 46h 48m 14s. Time estimates for 10 more iterations: 9m 21s, 100 more iterations: 1h 33m 36s, 500 more iterations: 7h 48m 2s. -[2025-11-09 23:05:08,129][__main__][INFO] - Starting iteration 337. -[2025-11-09 23:05:08,602][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 33 and human policies 1. -[2025-11-09 23:05:08,603][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:05:36,577][__main__][INFO] - Number of regex retries in iteration 337: 0 -[2025-11-09 23:05:36,578][__main__][INFO] - agents played in iteration 337 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:05:37,367][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:05:37,392][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:05:37,416][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:05:37,438][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:05:37,438][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:05:37,440][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:05:38,206][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:05:38,665][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:05:39,172][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:05:39,676][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:05:40,177][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:05:40,682][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:05:41,186][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:05:41,685][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:05:42,186][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:05:42,687][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:05:43,184][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:05:43,685][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:05:44,183][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:05:44,682][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:05:45,188][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:05:45,691][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:05:46,195][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:05:46,693][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:05:47,192][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:05:47,697][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:05:48,219][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:05:48,736][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:05:49,240][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:05:49,739][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:05:50,238][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:05:50,738][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:05:51,242][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:05:51,741][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:05:52,241][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:05:52,747][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:05:53,244][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:05:53,739][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:05:54,237][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:05:54,734][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:05:55,241][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:05:55,738][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:05:56,235][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:05:56,731][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:05:57,228][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:05:57,724][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:05:58,223][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:05:58,723][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:05:59,221][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:05:59,721][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:06:00,226][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:06:00,743][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:06:01,246][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:06:01,750][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:06:02,249][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:06:02,747][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:06:03,246][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:06:03,747][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:06:04,246][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:06:04,761][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:06:05,260][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:06:05,760][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:06:06,262][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:06:06,762][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:06:07,267][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:06:07,771][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:06:08,271][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:06:08,774][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:06:09,278][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:06:09,779][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:06:10,282][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10208 tokens. -[2025-11-09 23:06:11,051][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 55.67%, Block Peak % of device VRAM: 62.14%, ΔTime: 00:00:32 -[2025-11-09 23:06:11,789][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:06:11,790][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:06:11,792][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:06:12,869][__main__][INFO] - Iteration 338 took 1m 4s (43.53% Gen, 54.79% Train). Generation: 27s, Training: 35s. Estimated remaining time: 48h 26m 7s. Estimated total time: 53h 33m 23s. Time estimates for 10 more iterations: 10m 42s, 100 more iterations: 1h 47m 6s, 500 more iterations: 8h 55m 33s. -[2025-11-09 23:06:12,872][__main__][INFO] - Starting iteration 338. -[2025-11-09 23:06:13,341][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 33 and human policies 1. -[2025-11-09 23:06:13,342][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:06:41,993][__main__][INFO] - Number of regex retries in iteration 338: 0 -[2025-11-09 23:06:41,995][__main__][INFO] - agents played in iteration 338 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:06:42,858][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:06:42,881][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:06:42,905][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:06:42,929][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:06:42,929][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:06:42,930][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:06:43,714][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:06:44,177][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:06:44,686][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:06:45,186][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:06:45,689][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:06:46,187][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:06:46,686][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:06:47,187][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:06:47,691][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:06:48,193][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:06:48,697][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:06:49,195][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:06:49,693][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:06:50,188][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:06:50,684][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:06:51,181][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:06:51,681][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:06:52,184][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:06:52,685][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:06:53,183][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:06:53,685][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:06:54,180][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:06:54,679][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:06:55,179][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:06:55,676][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:06:56,184][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:06:56,682][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:06:57,183][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:06:57,712][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:06:58,217][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:06:58,728][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:06:59,232][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:06:59,735][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:07:00,233][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:07:00,737][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:07:01,257][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:07:01,760][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:07:02,261][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:07:02,764][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:07:03,269][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:07:03,777][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:07:04,280][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:07:04,782][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:07:05,289][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:07:05,791][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:07:06,294][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:07:06,798][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:07:07,298][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:07:07,800][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:07:08,302][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:07:08,809][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:07:09,314][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:07:09,818][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:07:10,319][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:07:10,822][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:07:11,328][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:07:11,840][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:07:12,343][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:07:12,846][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:07:13,351][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:07:13,852][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:07:14,365][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:07:14,871][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:07:15,375][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:07:15,887][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10257 tokens. -[2025-11-09 23:07:16,683][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.29%, Current % of VRAM taken: 55.78%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:32 -[2025-11-09 23:07:17,341][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:07:17,342][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:07:17,344][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:07:18,282][__main__][INFO] - Iteration 339 took 1m 4s (44.12% Gen, 54.43% Train). Generation: 28s, Training: 35s. Estimated remaining time: 48h 58m 42s. Estimated total time: 54h 7m 3s. Time estimates for 10 more iterations: 10m 49s, 100 more iterations: 1h 48m 14s, 500 more iterations: 9h 1m 10s. -[2025-11-09 23:07:18,284][__main__][INFO] - Starting iteration 339. -[2025-11-09 23:07:18,771][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 33 and human policies 1. -[2025-11-09 23:07:18,771][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:07:47,950][__main__][INFO] - Number of regex retries in iteration 339: 0 -[2025-11-09 23:07:47,950][__main__][INFO] - agents played in iteration 339 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:07:48,744][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:07:48,772][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:07:48,797][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:07:48,821][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:07:48,822][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:07:48,823][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:07:49,571][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:07:50,028][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:07:50,535][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:07:51,037][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:07:51,537][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:07:52,038][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:07:52,551][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:07:53,051][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:07:53,560][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:07:54,058][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:07:54,556][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:07:55,055][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:07:55,562][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:07:56,063][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:07:56,578][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:07:57,081][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:07:57,594][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:07:58,098][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:07:58,601][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:07:59,104][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:07:59,606][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:08:00,107][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:08:00,605][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:08:01,105][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:08:01,606][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:08:02,105][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:08:02,607][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:08:03,108][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:08:03,609][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:08:04,109][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:08:04,607][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:08:05,104][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:08:05,605][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:08:06,104][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:08:06,606][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:08:07,106][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:08:07,605][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:08:08,107][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:08:08,607][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:08:09,104][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:08:09,605][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:08:10,109][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:08:10,609][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:08:11,111][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:08:11,612][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:08:12,118][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:08:12,621][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:08:13,123][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:08:13,629][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:08:14,159][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:08:14,662][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:08:15,166][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:08:16,491][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:08:17,362][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:08:17,866][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:08:18,381][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:08:18,885][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:08:19,386][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:08:19,900][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:08:20,401][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:08:20,905][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:08:21,406][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:08:21,907][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:08:22,411][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:08:22,915][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10332 tokens. -[2025-11-09 23:08:23,732][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 55.69%, Block Peak % of device VRAM: 62.24%, ΔTime: 00:00:34 -[2025-11-09 23:08:24,384][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:08:24,385][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:08:24,387][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:08:25,360][__main__][INFO] - Iteration 340 took 1m 6s (43.82% Gen, 54.72% Train). Generation: 29s, Training: 36s. Estimated remaining time: 50h 20m 1s. Estimated total time: 55h 29m 29s. Time estimates for 10 more iterations: 11m 5s, 100 more iterations: 1h 50m 58s, 500 more iterations: 9h 14m 54s. -[2025-11-09 23:08:25,363][__main__][INFO] - Starting iteration 340. -[2025-11-09 23:08:25,830][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 33 and human policies 1. -[2025-11-09 23:08:25,831][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:08:48,087][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 23:09:00,644][__main__][INFO] - Number of regex retries in iteration 340: 1 -[2025-11-09 23:09:00,645][__main__][INFO] - agents played in iteration 340 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:09:01,459][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:09:01,487][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:09:01,513][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:09:01,537][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:09:01,537][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:09:01,538][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:09:02,226][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:09:02,783][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:09:03,301][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:09:03,811][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:09:04,319][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:09:04,823][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:09:05,322][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:09:05,823][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:09:06,322][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:09:06,820][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:09:07,320][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:09:07,823][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:09:08,322][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:09:08,821][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:09:09,319][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:09:09,820][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:09:10,319][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:09:10,820][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:09:11,321][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:09:11,822][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:09:12,319][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:09:12,820][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:09:13,321][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:09:13,823][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:09:14,324][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:09:14,823][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:09:15,324][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:09:15,824][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:09:16,333][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:09:16,835][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:09:17,337][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:09:17,839][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:09:18,338][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:09:18,840][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:09:19,339][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:09:19,842][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:09:20,345][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:09:20,844][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:09:21,345][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:09:21,849][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:09:22,350][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:09:22,851][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:09:23,354][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:09:23,857][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:09:24,359][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:09:24,858][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:09:25,361][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:09:25,861][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:09:26,364][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:09:26,880][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:09:27,381][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:09:27,883][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:09:28,385][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:09:28,888][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:09:29,386][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:09:29,882][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:09:30,385][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:09:30,887][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:09:31,380][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:09:31,875][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:09:32,370][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:09:32,866][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:09:33,361][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:09:33,860][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:09:34,356][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10325 tokens. -[2025-11-09 23:09:35,082][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 55.62%, Block Peak % of device VRAM: 62.34%, ΔTime: 00:00:32 -[2025-11-09 23:09:35,822][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:09:35,824][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:09:35,825][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:09:37,762][__main__][INFO] - Iteration 341 took 1m 11s (48.40% Gen, 48.91% Train). Generation: 34s, Training: 35s. Estimated remaining time: 54h 45m 57s. Estimated total time: 59h 56m 38s. Time estimates for 10 more iterations: 11m 59s, 100 more iterations: 1h 59m 53s, 500 more iterations: 9h 59m 26s. -[2025-11-09 23:09:37,765][__main__][INFO] - Starting iteration 341. -[2025-11-09 23:09:38,235][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 34 and human policies 1. -[2025-11-09 23:09:38,236][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:09:58,825][__main__][INFO] - Number of regex retries in iteration 341: 0 -[2025-11-09 23:09:58,827][__main__][INFO] - agents played in iteration 341 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:09:59,688][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:09:59,726][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:09:59,755][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:09:59,780][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:09:59,781][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:09:59,782][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:10:00,535][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:10:00,994][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:10:01,500][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:10:02,004][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:10:02,506][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:10:03,012][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:10:03,517][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:10:04,012][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:10:04,513][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:10:05,009][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:10:05,514][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:10:06,011][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:10:06,512][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:10:07,009][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:10:07,528][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:10:08,025][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:10:08,523][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:10:09,027][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:10:09,527][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:10:10,029][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:10:10,523][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:10:11,019][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:10:11,525][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:10:12,023][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:10:12,521][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:10:13,028][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:10:13,528][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:10:14,023][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:10:14,528][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:10:15,025][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:10:15,530][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:10:16,031][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:10:16,536][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:10:17,033][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:10:17,531][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:10:18,036][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:10:18,533][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:10:19,031][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:10:19,537][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:10:20,042][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:10:20,542][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:10:21,042][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:10:21,547][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:10:22,049][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:10:22,549][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:10:23,050][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:10:23,549][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:10:24,049][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:10:24,553][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:10:25,055][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:10:25,559][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:10:26,060][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:10:26,564][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:10:27,064][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:10:27,565][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:10:28,065][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:10:28,570][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:10:29,068][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:10:29,576][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:10:30,083][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:10:30,587][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:10:31,092][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:10:31,591][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:10:32,094][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:10:32,609][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10278 tokens. -[2025-11-09 23:10:33,396][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.05%, Current % of VRAM taken: 55.54%, Block Peak % of device VRAM: 62.14%, ΔTime: 00:00:32 -[2025-11-09 23:10:34,068][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:10:34,070][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:10:34,072][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:10:35,000][__main__][INFO] - Iteration 342 took 56s (36.27% Gen, 62.09% Train). Generation: 20s, Training: 35s. Estimated remaining time: 42h 6m 37s. Estimated total time: 47h 18m 15s. Time estimates for 10 more iterations: 9m 27s, 100 more iterations: 1h 34m 36s, 500 more iterations: 7h 53m 2s. -[2025-11-09 23:10:35,002][__main__][INFO] - Starting iteration 342. -[2025-11-09 23:10:35,476][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 34 and human policies 1. -[2025-11-09 23:10:35,477][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:11:05,596][__main__][INFO] - Number of regex retries in iteration 342: 0 -[2025-11-09 23:11:05,596][__main__][INFO] - agents played in iteration 342 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:11:06,392][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:11:06,418][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:11:06,442][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:11:06,465][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:11:06,466][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:11:06,467][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:11:07,190][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:11:07,645][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:11:08,148][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:11:08,650][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:11:09,153][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:11:09,654][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:11:10,154][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:11:10,657][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:11:11,156][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:11:11,653][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:11:12,155][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:11:12,653][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:11:13,154][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:11:13,651][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:11:14,147][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:11:14,647][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:11:15,144][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:11:15,641][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:11:16,138][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:11:16,636][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:11:17,137][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:11:17,634][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:11:18,132][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:11:18,629][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:11:19,126][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:11:19,624][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:11:20,122][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:11:20,620][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:11:21,138][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:11:21,634][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:11:22,135][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:11:22,632][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:11:23,129][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:11:23,636][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:11:24,135][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:11:24,635][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:11:25,137][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:11:25,635][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:11:26,135][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:11:26,638][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:11:27,141][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:11:27,647][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:11:28,149][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:11:28,651][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:11:29,158][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:11:29,658][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:11:30,158][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:11:30,660][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:11:31,159][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:11:31,660][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:11:32,157][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:11:32,660][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:11:33,159][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:11:33,659][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:11:34,163][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:11:34,659][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:11:35,154][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:11:35,660][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:11:36,177][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:11:36,678][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:11:37,179][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:11:37,681][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:11:38,181][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:11:38,678][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:11:39,179][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10226 tokens. -[2025-11-09 23:11:39,889][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 55.72%, Block Peak % of device VRAM: 62.39%, ΔTime: 00:00:32 -[2025-11-09 23:11:40,616][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:11:40,618][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:11:40,620][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:11:41,610][__main__][INFO] - Iteration 343 took 1m 6s (45.54% Gen, 52.96% Train). Generation: 30s, Training: 35s. Estimated remaining time: 49h 53m 58s. Estimated total time: 55h 6m 43s. Time estimates for 10 more iterations: 11m 1s, 100 more iterations: 1h 50m 13s, 500 more iterations: 9h 11m 7s. -[2025-11-09 23:11:41,612][__main__][INFO] - Starting iteration 343. -[2025-11-09 23:11:42,095][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 34 and human policies 1. -[2025-11-09 23:11:42,096][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:12:12,384][__main__][INFO] - Number of regex retries in iteration 343: 0 -[2025-11-09 23:12:12,386][__main__][INFO] - agents played in iteration 343 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:12:13,190][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:12:13,216][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:12:13,242][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:12:13,265][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:12:13,265][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:12:13,266][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:12:14,034][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:12:14,494][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:12:15,000][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:12:15,509][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:12:16,013][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:12:16,517][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:12:17,014][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:12:17,525][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:12:18,022][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:12:18,521][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:12:19,038][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:12:19,537][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:12:20,040][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:12:20,542][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:12:21,044][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:12:21,550][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:12:22,050][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:12:22,552][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:12:23,062][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:12:23,564][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:12:24,067][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:12:24,569][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:12:25,070][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:12:25,573][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:12:26,074][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:12:26,577][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:12:27,078][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:12:27,580][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:12:28,087][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:12:28,597][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:12:29,099][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:12:29,599][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:12:30,102][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:12:30,607][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:12:31,111][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:12:31,612][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:12:32,130][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:12:32,634][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:12:33,134][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:12:33,642][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:12:34,145][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:12:34,654][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:12:35,158][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:12:35,659][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:12:36,161][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:12:36,661][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:12:37,162][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:12:37,667][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:12:38,168][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:12:38,677][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:12:39,178][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:12:39,680][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:12:40,181][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:12:40,679][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:12:41,180][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:12:41,678][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:12:42,179][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:12:42,678][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:12:43,176][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:12:43,676][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:12:44,177][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:12:44,675][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:12:45,175][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:12:45,675][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:12:46,177][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10412 tokens. -[2025-11-09 23:12:46,943][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 55.70%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:32 -[2025-11-09 23:12:47,583][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:12:47,584][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:12:47,586][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:12:48,471][__main__][INFO] - Iteration 344 took 1m 6s (45.63% Gen, 53.03% Train). Generation: 30s, Training: 35s. Estimated remaining time: 50h 5m 0s. Estimated total time: 55h 18m 51s. Time estimates for 10 more iterations: 11m 3s, 100 more iterations: 1h 50m 37s, 500 more iterations: 9h 13m 8s. -[2025-11-09 23:12:48,473][__main__][INFO] - Starting iteration 344. -[2025-11-09 23:12:48,939][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 34 and human policies 1. -[2025-11-09 23:12:48,940][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:13:09,890][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 23:13:14,579][__main__][INFO] - Number of regex retries in iteration 344: 1 -[2025-11-09 23:13:14,580][__main__][INFO] - agents played in iteration 344 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:13:15,380][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:13:15,410][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:13:15,437][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:13:15,460][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:13:15,461][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:13:15,462][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:13:16,241][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:13:16,704][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:13:17,211][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:13:17,707][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:13:18,204][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:13:18,703][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:13:19,201][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:13:19,702][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:13:20,198][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:13:20,700][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:13:21,198][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:13:21,696][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:13:22,206][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:13:22,707][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:13:23,204][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:13:23,701][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:13:24,202][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:13:24,706][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:13:25,208][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:13:25,714][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:13:26,220][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:13:26,721][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:13:27,220][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:13:27,720][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:13:28,220][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:13:28,726][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:13:29,264][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:13:29,767][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:13:30,272][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:13:30,801][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:13:31,306][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:13:31,806][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:13:32,308][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:13:32,814][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:13:33,317][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:13:33,820][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:13:34,325][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:13:34,827][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:13:35,328][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:13:35,832][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:13:36,336][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:13:36,837][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:13:37,333][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:13:37,839][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:13:38,339][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:13:38,839][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:13:39,339][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:13:39,839][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:13:40,346][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:13:40,846][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:13:41,345][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:13:41,848][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:13:42,347][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:13:42,847][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:13:43,343][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:13:43,841][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:13:44,344][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:13:44,841][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:13:45,340][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:13:45,844][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:13:46,346][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:13:46,844][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:13:47,344][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:13:47,846][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:13:48,359][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10282 tokens. -[2025-11-09 23:13:49,071][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 55.58%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:32 -[2025-11-09 23:13:49,810][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:13:49,812][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:13:49,814][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:13:50,870][__main__][INFO] - Iteration 345 took 1m 1s (41.40% Gen, 56.89% Train). Generation: 25s, Training: 35s. Estimated remaining time: 46h 21m 39s. Estimated total time: 51h 36m 33s. Time estimates for 10 more iterations: 10m 19s, 100 more iterations: 1h 43m 13s, 500 more iterations: 8h 36m 5s. -[2025-11-09 23:13:50,872][__main__][INFO] - Starting iteration 345. -[2025-11-09 23:13:51,337][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 34 and human policies 1. -[2025-11-09 23:13:51,338][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:14:22,931][__main__][INFO] - Number of regex retries in iteration 345: 0 -[2025-11-09 23:14:22,932][__main__][INFO] - agents played in iteration 345 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:14:23,776][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:14:23,810][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:14:23,838][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:14:23,862][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:14:23,863][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:14:23,864][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:14:24,606][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:14:25,082][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:14:25,588][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:14:26,090][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:14:26,588][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:14:27,089][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:14:27,599][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:14:28,102][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:14:28,605][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:14:29,118][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:14:29,621][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:14:30,123][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:14:30,624][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:14:31,125][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:14:31,631][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:14:32,133][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:14:32,638][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:14:33,145][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:14:33,647][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:14:34,154][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:14:34,659][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:14:35,163][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:14:35,663][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:14:36,166][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:14:36,668][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:14:37,170][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:14:37,677][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:14:38,197][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:14:38,703][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:14:39,205][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:14:39,713][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:14:40,218][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:14:40,726][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:14:41,228][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:14:41,725][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:14:42,225][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:14:42,725][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:14:43,222][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:14:43,719][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:14:44,216][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:14:44,716][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:14:45,212][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:14:45,709][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:14:46,207][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:14:46,704][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:14:47,201][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:14:47,698][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:14:48,197][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:14:48,702][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:14:49,202][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:14:49,710][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:14:50,217][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:14:50,719][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:14:51,222][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:14:51,726][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:14:52,226][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:14:52,724][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:14:53,223][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:14:53,722][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:14:54,224][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:14:54,725][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:14:55,227][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:14:55,725][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:14:56,223][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:14:56,726][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10292 tokens. -[2025-11-09 23:14:57,528][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 55.65%, Block Peak % of device VRAM: 62.52%, ΔTime: 00:00:32 -[2025-11-09 23:14:58,182][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:14:58,185][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:14:58,186][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:14:59,137][__main__][INFO] - Iteration 346 took 1m 7s (46.60% Gen, 52.00% Train). Generation: 31s, Training: 35s. Estimated remaining time: 51h 13m 57s. Estimated total time: 56h 29m 59s. Time estimates for 10 more iterations: 11m 17s, 100 more iterations: 1h 52m 59s, 500 more iterations: 9h 24m 59s. -[2025-11-09 23:14:59,139][__main__][INFO] - Starting iteration 346. -[2025-11-09 23:14:59,607][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 34 and human policies 1. -[2025-11-09 23:14:59,608][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:15:23,617][__main__][INFO] - Number of regex retries in iteration 346: 0 -[2025-11-09 23:15:23,618][__main__][INFO] - agents played in iteration 346 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:15:24,449][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:15:24,476][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:15:24,501][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:15:24,524][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:15:24,524][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:15:24,525][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:15:25,346][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:15:25,809][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:15:26,320][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:15:26,827][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:15:27,331][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:15:27,831][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:15:28,335][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:15:28,835][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:15:29,344][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:15:29,846][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:15:30,350][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:15:30,855][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:15:31,363][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:15:31,868][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:15:32,374][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:15:32,876][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:15:33,375][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:15:33,875][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:15:34,377][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:15:34,889][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:15:35,389][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:15:35,892][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:15:36,401][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:15:36,903][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:15:37,980][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:15:38,956][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:15:39,461][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:15:39,964][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:15:40,468][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:15:40,968][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:15:41,471][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:15:41,980][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:15:42,482][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:15:42,999][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:15:43,503][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:15:44,019][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:15:44,521][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:15:45,021][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:15:45,532][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:15:46,032][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:15:46,535][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:15:47,033][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:15:47,537][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:15:48,039][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:15:48,540][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:15:49,040][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:15:49,537][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:15:50,034][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:15:50,537][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:15:51,041][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:15:51,542][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:15:52,039][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:15:52,539][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:15:53,044][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:15:53,543][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:15:54,044][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:15:54,544][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:15:55,040][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:15:55,538][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:15:56,038][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:15:56,534][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:15:57,034][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:15:57,531][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:15:58,028][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:15:58,540][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10354 tokens. -[2025-11-09 23:15:59,282][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 55.60%, Block Peak % of device VRAM: 62.20%, ΔTime: 00:00:33 -[2025-11-09 23:15:59,925][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:15:59,927][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:15:59,929][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:16:00,919][__main__][INFO] - Iteration 347 took 1m 1s (39.16% Gen, 59.22% Train). Generation: 24s, Training: 36s. Estimated remaining time: 45h 48m 34s. Estimated total time: 51h 5m 38s. Time estimates for 10 more iterations: 10m 13s, 100 more iterations: 1h 42m 11s, 500 more iterations: 8h 30m 56s. -[2025-11-09 23:16:00,922][__main__][INFO] - Starting iteration 347. -[2025-11-09 23:16:01,409][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 34 and human policies 1. -[2025-11-09 23:16:01,410][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:16:33,091][__main__][INFO] - Number of regex retries in iteration 347: 0 -[2025-11-09 23:16:33,092][__main__][INFO] - agents played in iteration 347 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:16:33,945][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:16:33,972][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:16:33,996][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:16:34,020][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:16:34,020][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:16:34,021][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:16:34,801][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:16:35,269][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:16:35,773][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:16:36,275][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:16:36,781][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:16:37,284][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:16:37,784][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:16:38,285][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:16:38,785][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:16:39,288][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:16:39,790][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:16:40,295][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:16:40,798][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:16:41,302][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:16:41,815][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:16:42,319][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:16:42,824][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:16:43,327][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:16:43,827][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:16:44,343][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:16:44,842][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:16:45,344][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:16:45,855][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:16:46,368][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:16:46,874][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:16:47,379][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:16:47,883][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:16:48,390][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:16:48,894][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:16:49,401][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:16:49,906][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:16:50,409][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:16:50,913][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:16:51,415][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:16:51,921][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:16:52,423][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:16:52,924][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:16:53,442][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:16:53,944][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:16:54,444][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:16:54,947][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:16:55,445][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:16:55,948][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:16:56,451][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:16:56,954][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:16:57,457][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:16:57,955][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:16:58,457][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:16:58,956][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:16:59,460][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:16:59,964][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:17:00,462][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:17:00,963][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:17:01,461][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:17:01,960][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:17:02,458][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:17:02,959][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:17:03,454][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:17:03,951][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:17:04,445][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:17:04,940][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:17:05,434][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:17:05,929][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:17:06,423][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:17:06,921][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10340 tokens. -[2025-11-09 23:17:07,627][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 55.71%, Block Peak % of device VRAM: 62.28%, ΔTime: 00:00:32 -[2025-11-09 23:17:08,378][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:17:08,379][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:17:08,381][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:17:09,391][__main__][INFO] - Iteration 348 took 1m 7s (46.60% Gen, 51.91% Train). Generation: 31s, Training: 35s. Estimated remaining time: 51h 20m 54s. Estimated total time: 56h 39m 6s. Time estimates for 10 more iterations: 11m 19s, 100 more iterations: 1h 53m 18s, 500 more iterations: 9h 26m 31s. -[2025-11-09 23:17:09,393][__main__][INFO] - Starting iteration 348. -[2025-11-09 23:17:09,860][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 34 and human policies 1. -[2025-11-09 23:17:09,861][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:17:32,048][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 23:17:37,284][__main__][INFO] - Number of regex retries in iteration 348: 1 -[2025-11-09 23:17:37,285][__main__][INFO] - agents played in iteration 348 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:17:38,153][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:17:38,189][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:17:38,216][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:17:38,242][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:17:38,243][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:17:38,243][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:17:39,043][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:17:39,508][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:17:40,014][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:17:40,520][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:17:41,040][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:17:41,545][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:17:42,053][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:17:42,556][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:17:43,062][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:17:43,569][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:17:44,078][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:17:44,584][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:17:45,092][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:17:45,606][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:17:46,125][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:17:46,630][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:17:47,135][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:17:47,641][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:17:48,147][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:17:48,652][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:17:49,155][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:17:49,659][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:17:50,162][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:17:50,663][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:17:51,165][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:17:51,670][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:17:52,174][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:17:52,678][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:17:53,188][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:17:53,709][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:17:54,219][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:17:54,721][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:17:55,225][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:17:55,725][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:17:56,227][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:17:56,726][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:17:57,227][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:17:57,730][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:17:58,234][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:17:58,736][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:17:59,238][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:17:59,736][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:18:00,237][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:18:00,736][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:18:01,235][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:18:01,737][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:18:02,238][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:18:02,738][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:18:03,242][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:18:03,744][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:18:04,255][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:18:04,760][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:18:05,259][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:18:05,766][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:18:06,268][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:18:06,770][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:18:07,266][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:18:07,764][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:18:08,264][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:18:08,761][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:18:09,260][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:18:09,761][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:18:10,258][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:18:10,759][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:18:11,259][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10325 tokens. -[2025-11-09 23:18:11,974][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 55.67%, Block Peak % of device VRAM: 62.33%, ΔTime: 00:00:32 -[2025-11-09 23:18:12,622][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:18:12,623][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:18:12,625][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:18:13,518][__main__][INFO] - Iteration 349 took 1m 3s (43.08% Gen, 55.51% Train). Generation: 27s, Training: 35s. Estimated remaining time: 47h 43m 41s. Estimated total time: 53h 2m 58s. Time estimates for 10 more iterations: 10m 36s, 100 more iterations: 1h 46m 5s, 500 more iterations: 8h 50m 29s. -[2025-11-09 23:18:13,521][__main__][INFO] - Starting iteration 349. -[2025-11-09 23:18:13,997][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 34 and human policies 1. -[2025-11-09 23:18:13,997][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:18:38,372][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 23:18:44,826][__main__][INFO] - Number of regex retries in iteration 349: 1 -[2025-11-09 23:18:44,827][__main__][INFO] - agents played in iteration 349 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:18:45,643][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:18:45,666][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:18:45,690][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:18:45,713][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:18:45,714][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:18:45,715][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:18:46,497][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:18:46,955][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:18:47,471][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:18:47,976][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:18:48,479][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:18:48,988][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:18:49,494][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:18:49,997][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:18:50,501][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:18:51,003][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:18:51,507][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:18:52,015][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:18:52,515][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:18:53,021][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:18:53,523][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:18:54,025][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:18:54,526][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:18:55,028][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:18:55,546][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:18:56,046][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:18:56,546][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:18:57,047][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:18:57,545][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:18:58,054][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:18:58,558][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:18:59,061][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:18:59,563][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:19:00,065][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:19:00,569][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:19:01,070][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:19:01,572][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:19:02,080][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:19:02,581][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:19:03,081][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:19:03,580][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:19:04,084][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:19:04,587][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:19:05,087][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:19:05,588][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:19:06,085][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:19:06,582][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:19:07,102][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:19:07,603][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:19:08,102][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:19:08,602][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:19:09,102][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:19:09,606][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:19:10,106][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:19:10,606][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:19:11,106][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:19:11,608][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:19:12,106][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:19:12,611][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:19:13,112][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:19:13,611][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:19:14,112][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:19:14,614][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:19:15,115][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:19:15,615][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:19:16,112][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:19:16,612][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:19:17,113][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:19:17,610][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:19:18,112][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:19:18,611][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10449 tokens. -[2025-11-09 23:19:19,320][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.47%, ΔTime: 00:00:32 -[2025-11-09 23:19:20,053][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:19:20,054][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:19:20,056][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:19:21,104][__main__][INFO] - Iteration 350 took 1m 7s (45.94% Gen, 52.50% Train). Generation: 30s, Training: 35s. Estimated remaining time: 50h 35m 0s. Estimated total time: 55h 55m 24s. Time estimates for 10 more iterations: 11m 11s, 100 more iterations: 1h 51m 50s, 500 more iterations: 9h 19m 14s. -[2025-11-09 23:19:21,106][__main__][INFO] - Starting iteration 350. -[2025-11-09 23:19:21,584][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 34 and human policies 1. -[2025-11-09 23:19:21,585][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:19:51,827][__main__][INFO] - Number of regex retries in iteration 350: 0 -[2025-11-09 23:19:51,829][__main__][INFO] - agents played in iteration 350 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:19:52,655][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:19:52,686][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:19:52,710][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:19:52,734][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:19:52,734][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:19:52,736][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:19:53,528][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:19:53,991][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:19:54,498][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:19:55,004][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:19:55,504][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:19:56,010][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:19:56,520][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:19:57,028][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:19:57,534][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:19:58,037][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:19:58,542][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:19:59,052][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:19:59,557][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:20:00,059][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:20:00,560][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:20:01,062][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:20:01,576][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:20:02,075][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:20:02,573][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:20:03,077][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:20:03,575][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:20:04,075][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:20:04,578][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:20:05,075][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:20:05,583][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:20:06,084][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:20:06,583][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:20:07,082][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:20:07,583][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:20:08,086][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:20:08,589][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:20:09,090][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:20:09,594][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:20:10,097][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:20:10,605][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:20:11,111][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:20:11,613][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:20:12,115][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:20:12,618][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:20:13,120][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:20:13,632][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:20:14,134][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:20:14,637][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:20:15,132][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:20:15,628][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:20:16,130][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:20:16,626][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:20:17,126][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:20:17,632][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:20:18,131][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:20:18,628][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:20:19,125][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:20:19,622][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:20:20,126][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:20:20,626][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:20:21,126][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:20:21,624][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:20:22,124][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:20:22,622][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:20:23,121][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:20:23,621][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:20:24,123][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:20:24,621][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:20:25,119][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:20:25,620][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10300 tokens. -[2025-11-09 23:20:26,353][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.25%, Current % of VRAM taken: 55.74%, Block Peak % of device VRAM: 62.48%, ΔTime: 00:00:32 -[2025-11-09 23:20:26,997][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:20:26,999][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:20:27,000][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:20:28,866][__main__][INFO] - Iteration 351 took 1m 7s (44.95% Gen, 52.27% Train). Generation: 30s, Training: 35s. Estimated remaining time: 50h 42m 35s. Estimated total time: 56h 4m 7s. Time estimates for 10 more iterations: 11m 12s, 100 more iterations: 1h 52m 8s, 500 more iterations: 9h 20m 41s. -[2025-11-09 23:20:28,868][__main__][INFO] - Starting iteration 351. -[2025-11-09 23:20:29,334][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 35 and human policies 1. -[2025-11-09 23:20:29,335][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:20:53,682][__main__][INFO] - Number of regex retries in iteration 351: 0 -[2025-11-09 23:20:53,683][__main__][INFO] - agents played in iteration 351 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:20:54,481][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:20:54,505][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:20:54,528][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:20:54,551][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:20:54,551][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:20:54,552][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:20:55,362][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:20:55,829][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:20:56,338][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:20:56,840][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:20:57,343][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:20:58,326][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:20:59,318][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:20:59,822][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:21:00,324][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:21:00,828][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:21:01,331][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:21:01,834][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:21:02,356][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:21:02,861][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:21:03,381][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:21:03,888][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:21:04,394][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:21:04,899][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:21:05,401][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:21:05,905][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:21:06,406][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:21:06,907][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:21:07,412][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:21:07,914][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:21:08,421][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:21:08,923][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:21:09,422][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:21:09,922][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:21:10,420][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:21:10,919][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:21:11,417][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:21:11,916][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:21:12,414][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:21:12,915][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:21:13,414][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:21:13,918][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:21:14,418][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:21:14,916][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:21:15,415][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:21:15,915][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:21:16,414][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:21:16,915][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:21:17,417][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:21:17,916][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:21:18,415][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:21:18,917][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:21:19,414][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:21:19,915][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:21:20,416][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:21:20,913][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:21:21,409][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:21:21,907][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:21:22,403][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:21:22,899][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:21:23,395][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:21:23,891][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:21:24,391][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:21:24,887][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:21:25,383][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:21:25,882][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:21:26,384][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:21:26,885][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:21:27,380][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:21:27,877][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:21:28,383][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10266 tokens. -[2025-11-09 23:21:29,130][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 55.71%, Block Peak % of device VRAM: 62.28%, ΔTime: 00:00:33 -[2025-11-09 23:21:29,791][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:21:29,793][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:21:29,794][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:21:30,817][__main__][INFO] - Iteration 352 took 1m 1s (39.60% Gen, 58.73% Train). Generation: 24s, Training: 36s. Estimated remaining time: 45h 51m 38s. Estimated total time: 51h 14m 12s. Time estimates for 10 more iterations: 10m 14s, 100 more iterations: 1h 42m 28s, 500 more iterations: 8h 32m 22s. -[2025-11-09 23:21:30,820][__main__][INFO] - Starting iteration 352. -[2025-11-09 23:21:31,304][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 35 and human policies 1. -[2025-11-09 23:21:31,305][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:21:49,330][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 23:22:01,712][__main__][INFO] - Number of regex retries in iteration 352: 1 -[2025-11-09 23:22:01,712][__main__][INFO] - agents played in iteration 352 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:22:02,638][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:22:02,661][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:22:02,685][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:22:02,708][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:22:02,709][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:22:02,710][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:22:03,469][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:22:03,931][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:22:04,441][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:22:04,945][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:22:05,448][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:22:05,955][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:22:06,458][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:22:06,959][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:22:07,459][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:22:07,972][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:22:08,474][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:22:08,975][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:22:09,479][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:22:09,979][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:22:10,491][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:22:10,997][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:22:11,499][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:22:12,999][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:22:12,499][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:22:13,000][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:22:13,499][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:22:13,999][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:22:14,503][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:22:15,005][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:22:15,506][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:22:16,005][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:22:16,504][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:22:17,006][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:22:17,506][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:22:18,007][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:22:18,506][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:22:19,005][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:22:19,508][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:22:20,011][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:22:20,513][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:22:21,011][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:22:21,509][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:22:22,010][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:22:22,509][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:22:23,008][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:22:23,504][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:22:24,001][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:22:24,498][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:22:24,996][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:22:25,491][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:22:25,986][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:22:26,486][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:22:26,984][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:22:27,482][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:22:27,979][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:22:28,479][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:22:28,975][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:22:29,472][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:22:29,969][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:22:30,469][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:22:30,965][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:22:31,465][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:22:31,963][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:22:32,461][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:22:32,966][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:22:33,461][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:22:33,958][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:22:34,454][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:22:34,949][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:22:35,448][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10299 tokens. -[2025-11-09 23:22:36,158][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 55.70%, Block Peak % of device VRAM: 62.38%, ΔTime: 00:00:32 -[2025-11-09 23:22:36,910][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:22:36,912][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:22:36,913][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:22:37,924][__main__][INFO] - Iteration 353 took 1m 6s (45.64% Gen, 52.84% Train). Generation: 30s, Training: 35s. Estimated remaining time: 50h 7m 19s. Estimated total time: 55h 31m 0s. Time estimates for 10 more iterations: 11m 6s, 100 more iterations: 1h 51m 2s, 500 more iterations: 9h 15m 10s. -[2025-11-09 23:22:37,926][__main__][INFO] - Starting iteration 353. -[2025-11-09 23:22:38,398][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 35 and human policies 1. -[2025-11-09 23:22:38,398][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:22:50,346][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 23:23:04,265][__main__][INFO] - Number of regex retries in iteration 353: 1 -[2025-11-09 23:23:04,267][__main__][INFO] - agents played in iteration 353 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:23:05,114][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:23:05,138][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:23:05,162][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:23:05,187][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:23:05,188][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:23:05,188][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:23:05,995][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:23:06,462][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:23:06,967][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:23:07,469][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:23:07,980][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:23:08,496][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:23:09,000][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:23:09,503][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:23:10,015][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:23:10,515][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:23:11,016][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:23:11,518][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:23:12,026][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:23:12,527][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:23:13,029][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:23:13,534][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:23:14,038][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:23:14,542][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:23:15,052][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:23:15,556][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:23:16,061][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:23:16,568][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:23:17,069][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:23:17,571][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:23:18,070][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:23:18,570][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:23:19,070][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:23:19,570][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:23:20,073][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:23:20,581][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:23:21,079][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:23:21,592][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:23:22,093][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:23:22,592][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:23:23,095][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:23:23,596][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:23:24,094][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:23:24,594][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:23:25,090][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:23:25,603][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:23:26,107][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:23:26,606][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:23:27,106][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:23:27,604][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:23:28,103][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:23:28,602][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:23:29,102][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:23:29,611][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:23:30,108][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:23:30,608][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:23:31,106][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:23:31,614][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:23:32,121][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:23:32,622][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:23:33,123][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:23:33,628][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:23:34,129][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:23:34,628][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:23:35,126][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:23:35,627][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:23:36,126][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:23:36,623][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:23:37,123][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:23:37,627][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:23:38,128][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10221 tokens. -[2025-11-09 23:23:38,941][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 55.64%, Block Peak % of device VRAM: 62.47%, ΔTime: 00:00:32 -[2025-11-09 23:23:39,583][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:23:39,584][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:23:39,586][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:23:40,575][__main__][INFO] - Iteration 354 took 1m 2s (41.60% Gen, 56.80% Train). Generation: 25s, Training: 35s. Estimated remaining time: 46h 24m 10s. Estimated total time: 51h 48m 54s. Time estimates for 10 more iterations: 10m 21s, 100 more iterations: 1h 43m 37s, 500 more iterations: 8h 38m 9s. -[2025-11-09 23:23:40,578][__main__][INFO] - Starting iteration 354. -[2025-11-09 23:23:41,050][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 35 and human policies 1. -[2025-11-09 23:23:41,051][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:24:10,965][__main__][INFO] - Number of regex retries in iteration 354: 0 -[2025-11-09 23:24:10,966][__main__][INFO] - agents played in iteration 354 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:24:11,790][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:24:11,814][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:24:11,838][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:24:11,861][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:24:11,862][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:24:11,862][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:24:12,652][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:24:13,110][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:24:13,616][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:24:14,125][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:24:14,627][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:24:15,127][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:24:15,628][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:24:16,128][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:24:16,627][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:24:17,129][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:24:17,631][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:24:18,135][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:24:18,641][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:24:19,142][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:24:19,644][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:24:20,145][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:24:20,647][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:24:21,146][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:24:21,683][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:24:22,197][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:24:22,703][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:24:23,221][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:24:23,725][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:24:24,231][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:24:24,737][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:24:25,240][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:24:25,744][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:24:26,243][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:24:26,745][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:24:27,248][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:24:27,749][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:24:28,246][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:24:28,745][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:24:29,245][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:24:29,751][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:24:30,251][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:24:30,749][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:24:31,248][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:24:31,747][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:24:32,251][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:24:32,752][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:24:33,248][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:24:33,751][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:24:34,248][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:24:34,748][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:24:35,243][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:24:35,740][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:24:36,239][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:24:36,734][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:24:37,228][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:24:37,729][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:24:38,223][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:24:38,719][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:24:39,215][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:24:39,712][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:24:40,224][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:24:40,719][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:24:41,216][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:24:41,714][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:24:42,210][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:24:42,716][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:24:43,217][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:24:43,710][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:24:44,212][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:24:44,712][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10396 tokens. -[2025-11-09 23:24:45,467][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.29%, Current % of VRAM taken: 55.79%, Block Peak % of device VRAM: 62.51%, ΔTime: 00:00:32 -[2025-11-09 23:24:46,188][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:24:46,190][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:24:46,191][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:24:47,240][__main__][INFO] - Iteration 355 took 1m 6s (45.19% Gen, 53.22% Train). Generation: 29s, Training: 35s. Estimated remaining time: 49h 43m 40s. Estimated total time: 55h 9m 30s. Time estimates for 10 more iterations: 11m 1s, 100 more iterations: 1h 50m 19s, 500 more iterations: 9h 11m 35s. -[2025-11-09 23:24:47,242][__main__][INFO] - Starting iteration 355. -[2025-11-09 23:24:47,705][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 35 and human policies 1. -[2025-11-09 23:24:47,705][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:25:05,385][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 23:25:18,006][__main__][INFO] - Number of regex retries in iteration 355: 1 -[2025-11-09 23:25:18,007][__main__][INFO] - agents played in iteration 355 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:25:18,836][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.67%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:25:18,861][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.67%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:25:18,885][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.67%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:25:18,909][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.67%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:25:18,910][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:25:18,911][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:25:19,723][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:25:20,188][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:25:20,700][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:25:21,213][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:25:21,712][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:25:22,228][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:25:22,731][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:25:23,239][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:25:23,750][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:25:24,259][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:25:24,762][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:25:25,264][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:25:25,770][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:25:26,271][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:25:26,774][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:25:27,276][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:25:27,781][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:25:28,277][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:25:28,779][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:25:29,278][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:25:29,782][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:25:30,282][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:25:30,781][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:25:31,280][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:25:31,781][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:25:32,279][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:25:32,777][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:25:33,274][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:25:33,779][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:25:34,285][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:25:34,786][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:25:35,283][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:25:35,785][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:25:36,285][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:25:36,790][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:25:37,296][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:25:37,797][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:25:38,298][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:25:38,798][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:25:39,299][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:25:39,801][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:25:40,300][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:25:40,822][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:25:41,320][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:25:41,822][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:25:42,324][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:25:42,824][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:25:43,334][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:25:43,832][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:25:44,331][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:25:44,844][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:25:45,342][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:25:45,839][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:25:46,335][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:25:46,833][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:25:47,348][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:25:47,843][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:25:48,342][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:25:48,839][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:25:49,336][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:25:49,840][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:25:50,337][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:25:50,833][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:25:51,338][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:25:51,838][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10328 tokens. -[2025-11-09 23:25:52,626][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 55.57%, Block Peak % of device VRAM: 62.50%, ΔTime: 00:00:32 -[2025-11-09 23:25:53,272][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:25:53,274][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:25:53,275][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:25:54,247][__main__][INFO] - Iteration 356 took 1m 6s (45.54% Gen, 53.00% Train). Generation: 30s, Training: 35s. Estimated remaining time: 50h 0m 11s. Estimated total time: 55h 27m 8s. Time estimates for 10 more iterations: 11m 5s, 100 more iterations: 1h 50m 54s, 500 more iterations: 9h 14m 31s. -[2025-11-09 23:25:54,250][__main__][INFO] - Starting iteration 356. -[2025-11-09 23:25:54,714][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 35 and human policies 1. -[2025-11-09 23:25:54,715][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:26:16,129][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 23:26:17,173][__main__][INFO] - Number of regex retries in iteration 356: 1 -[2025-11-09 23:26:17,175][__main__][INFO] - agents played in iteration 356 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:26:18,051][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:26:18,075][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:26:18,099][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:26:18,121][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:26:18,123][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:26:18,123][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:26:18,957][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:26:19,492][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:26:20,009][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:26:20,514][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:26:21,023][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:26:21,528][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:26:22,030][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:26:22,542][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:26:23,043][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:26:23,546][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:26:24,058][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:26:24,560][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:26:25,065][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:26:25,565][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:26:26,065][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:26:26,571][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:26:27,077][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:26:27,575][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:26:28,075][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:26:28,580][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:26:29,081][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:26:29,580][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:26:30,081][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:26:30,583][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:26:31,081][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:26:31,584][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:26:32,082][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:26:32,581][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:26:33,086][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:26:33,588][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:26:34,086][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:26:34,585][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:26:35,089][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:26:35,590][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:26:36,093][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:26:36,593][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:26:37,096][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:26:37,611][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:26:38,109][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:26:38,613][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:26:39,118][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:26:39,629][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:26:40,125][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:26:40,623][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:26:41,128][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:26:41,627][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:26:42,131][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:26:42,629][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:26:43,129][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:26:43,632][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:26:44,135][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:26:44,632][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:26:45,129][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:26:45,628][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:26:46,142][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:26:46,642][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:26:47,141][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:26:47,642][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:26:48,139][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:26:48,636][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:26:49,133][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:26:49,633][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:26:50,147][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:26:50,641][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:26:51,138][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10351 tokens. -[2025-11-09 23:26:51,874][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 55.67%, Block Peak % of device VRAM: 62.23%, ΔTime: 00:00:32 -[2025-11-09 23:26:52,520][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:26:52,521][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:26:52,523][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:26:53,436][__main__][INFO] - Iteration 357 took 58s (38.25% Gen, 60.19% Train). Generation: 22s, Training: 35s. Estimated remaining time: 43h 28m 10s. Estimated total time: 48h 56m 6s. Time estimates for 10 more iterations: 9m 47s, 100 more iterations: 1h 37m 52s, 500 more iterations: 8h 9m 21s. -[2025-11-09 23:26:53,440][__main__][INFO] - Starting iteration 357. -[2025-11-09 23:26:53,910][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 35 and human policies 1. -[2025-11-09 23:26:53,910][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:27:14,611][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 23:27:14,777][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 23:27:27,557][__main__][INFO] - Number of regex retries in iteration 357: 2 -[2025-11-09 23:27:27,558][__main__][INFO] - agents played in iteration 357 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:27:28,369][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:27:28,398][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:27:28,425][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:27:28,449][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:27:28,449][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:27:28,450][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:27:29,195][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:27:29,660][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:27:30,167][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:27:30,811][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:27:31,312][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:27:31,816][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:27:32,321][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:27:32,822][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:27:33,323][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:27:33,836][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:27:34,333][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:27:34,836][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:27:35,339][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:27:35,840][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:27:36,355][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:27:36,856][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:27:37,356][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:27:37,858][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:27:38,358][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:27:38,866][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:27:39,364][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:27:39,864][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:27:40,378][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:27:40,878][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:27:41,381][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:27:41,885][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:27:42,387][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:27:42,889][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:27:43,389][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:27:43,887][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:27:44,384][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:27:44,882][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:27:45,385][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:27:45,886][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:27:46,388][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:27:46,888][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:27:47,387][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:27:47,886][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:27:48,384][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:27:48,884][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:27:49,384][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:27:49,883][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:27:50,381][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:27:50,886][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:27:51,387][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:27:51,884][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:27:52,381][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:27:52,881][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:27:53,382][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:27:53,881][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:27:54,381][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:27:54,881][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:27:55,379][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:27:55,878][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:27:56,376][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:27:56,878][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:27:57,384][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:27:57,883][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:27:58,382][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:27:58,882][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:27:59,383][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:27:59,882][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:28:00,379][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:28:00,880][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:28:01,383][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10348 tokens. -[2025-11-09 23:28:02,180][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 55.69%, Block Peak % of device VRAM: 62.50%, ΔTime: 00:00:32 -[2025-11-09 23:28:02,921][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:28:02,923][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:28:02,924][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:28:03,961][__main__][INFO] - Iteration 358 took 1m 10s (48.03% Gen, 50.49% Train). Generation: 33s, Training: 35s. Estimated remaining time: 52h 53m 30s. Estimated total time: 58h 22m 37s. Time estimates for 10 more iterations: 11m 40s, 100 more iterations: 1h 56m 45s, 500 more iterations: 9h 43m 46s. -[2025-11-09 23:28:03,964][__main__][INFO] - Starting iteration 358. -[2025-11-09 23:28:04,452][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 35 and human policies 1. -[2025-11-09 23:28:04,453][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:28:28,450][__main__][INFO] - Number of regex retries in iteration 358: 0 -[2025-11-09 23:28:28,451][__main__][INFO] - agents played in iteration 358 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:28:29,285][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:28:29,324][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:28:29,357][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:28:29,387][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:28:29,387][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:28:29,389][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:28:30,199][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:28:30,658][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:28:31,166][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:28:31,686][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:28:32,186][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:28:32,696][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:28:33,197][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:28:33,701][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:28:34,215][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:28:34,719][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:28:35,244][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:28:35,750][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:28:36,251][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:28:36,767][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:28:37,274][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:28:37,776][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:28:38,279][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:28:38,781][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:28:39,286][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:28:39,787][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:28:40,292][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:28:40,800][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:28:41,298][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:28:41,799][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:28:42,302][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:28:42,801][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:28:43,302][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:28:43,803][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:28:44,302][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:28:44,803][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:28:45,300][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:28:45,802][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:28:46,303][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:28:46,801][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:28:47,299][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:28:47,797][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:28:48,301][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:28:48,806][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:28:49,304][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:28:49,804][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:28:50,304][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:28:50,802][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:28:51,301][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:28:51,798][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:28:52,297][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:28:52,800][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:28:53,297][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:28:53,796][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:28:54,296][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:28:54,794][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:28:55,298][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:28:55,798][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:28:56,296][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:28:56,801][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:28:57,297][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:28:57,794][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:28:58,293][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:28:58,788][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:28:59,302][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:28:59,800][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:29:00,297][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:29:00,801][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:29:01,301][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:29:01,804][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:29:02,300][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10242 tokens. -[2025-11-09 23:29:03,016][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 55.59%, Block Peak % of device VRAM: 62.49%, ΔTime: 00:00:32 -[2025-11-09 23:29:03,653][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:29:03,655][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:29:03,656][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:29:04,597][__main__][INFO] - Iteration 359 took 1m 0s (39.90% Gen, 58.53% Train). Generation: 23s, Training: 35s. Estimated remaining time: 44h 37m 8s. Estimated total time: 50h 7m 15s. Time estimates for 10 more iterations: 10m 1s, 100 more iterations: 1h 40m 14s, 500 more iterations: 8h 21m 12s. -[2025-11-09 23:29:04,599][__main__][INFO] - Starting iteration 359. -[2025-11-09 23:29:05,061][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 35 and human policies 1. -[2025-11-09 23:29:05,062][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:29:23,121][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 23:29:27,538][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 20 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 23:29:36,729][__main__][INFO] - Number of regex retries in iteration 359: 2 -[2025-11-09 23:29:36,729][__main__][INFO] - agents played in iteration 359 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:29:37,516][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:29:37,539][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:29:37,562][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:29:37,584][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:29:37,585][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:29:37,586][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:29:38,409][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:29:38,874][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:29:39,380][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:29:39,906][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:29:40,405][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:29:40,903][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:29:41,418][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:29:41,915][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:29:42,412][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:29:42,910][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:29:43,409][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:29:43,922][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:29:44,422][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:29:44,923][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:29:45,425][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:29:45,925][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:29:46,425][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:29:46,923][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:29:47,422][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:29:47,927][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:29:48,434][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:29:48,934][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:29:49,429][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:29:49,925][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:29:50,425][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:29:50,922][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:29:51,421][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:29:51,917][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:29:52,414][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:29:52,912][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:29:53,409][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:29:53,905][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:29:54,403][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:29:54,900][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:29:55,399][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:29:55,895][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:29:56,398][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:29:56,898][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:29:57,399][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:29:57,900][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:29:58,399][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:29:58,896][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:29:59,406][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:29:59,901][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:30:00,398][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:30:00,897][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:30:01,393][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:30:01,889][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:30:02,394][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:30:02,891][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:30:03,387][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:30:03,884][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:30:04,381][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:30:04,888][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:30:05,384][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:30:05,886][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:30:06,384][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:30:06,879][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:30:07,376][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:30:07,871][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:30:08,369][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:30:08,868][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:30:09,367][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:30:09,870][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:30:10,365][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10180 tokens. -[2025-11-09 23:30:11,106][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 55.57%, Block Peak % of device VRAM: 62.22%, ΔTime: 00:00:32 -[2025-11-09 23:30:11,839][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:30:11,840][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:30:11,842][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:30:12,888][__main__][INFO] - Iteration 360 took 1m 7s (46.69% Gen, 51.77% Train). Generation: 31s, Training: 35s. Estimated remaining time: 51h 0m 3s. Estimated total time: 56h 31m 19s. Time estimates for 10 more iterations: 11m 18s, 100 more iterations: 1h 53m 2s, 500 more iterations: 9h 25m 13s. -[2025-11-09 23:30:12,890][__main__][INFO] - Starting iteration 360. -[2025-11-09 23:30:13,354][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 35 and human policies 1. -[2025-11-09 23:30:13,354][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:30:39,157][__main__][INFO] - Number of regex retries in iteration 360: 0 -[2025-11-09 23:30:39,159][__main__][INFO] - agents played in iteration 360 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:30:39,965][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:30:39,998][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:30:40,026][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:30:40,051][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:30:40,052][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:30:40,054][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:30:40,797][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:30:41,259][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:30:41,765][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:30:42,268][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:30:42,772][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:30:43,279][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:30:43,784][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:30:44,305][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:30:44,820][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:30:45,328][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:30:45,829][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:30:46,334][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:30:46,840][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:30:47,339][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:30:47,841][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:30:48,339][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:30:48,838][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:30:49,340][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:30:49,839][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:30:50,337][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:30:50,846][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:30:51,348][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:30:51,850][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:30:52,350][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:30:52,851][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:30:53,353][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:30:53,852][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:30:54,353][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:30:54,856][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:30:55,355][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:30:55,856][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:30:56,355][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:30:56,852][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:30:57,355][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:30:57,854][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:30:58,350][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:30:58,848][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:30:59,344][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:30:59,841][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:31:00,338][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:31:00,834][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:31:01,334][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:31:01,830][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:31:02,327][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:31:02,825][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:31:03,320][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:31:03,824][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:31:04,320][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:31:04,817][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:31:05,317][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:31:05,817][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:31:06,318][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:31:06,818][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:31:07,314][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:31:07,826][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:31:08,324][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:31:08,824][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:31:09,324][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:31:09,822][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:31:10,328][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:31:10,826][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:31:11,321][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:31:11,835][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:31:12,332][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:31:12,830][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10210 tokens. -[2025-11-09 23:31:13,537][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 55.64%, Block Peak % of device VRAM: 62.46%, ΔTime: 00:00:32 -[2025-11-09 23:31:14,180][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:31:14,181][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:31:14,183][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:31:16,141][__main__][INFO] - Iteration 361 took 1m 2s (41.10% Gen, 55.78% Train). Generation: 25s, Training: 35s. Estimated remaining time: 46h 47m 4s. Estimated total time: 52h 19m 23s. Time estimates for 10 more iterations: 10m 27s, 100 more iterations: 1h 44m 38s, 500 more iterations: 8h 43m 13s. -[2025-11-09 23:31:16,143][__main__][INFO] - Starting iteration 361. -[2025-11-09 23:31:16,604][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 36 and human policies 1. -[2025-11-09 23:31:16,604][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:31:40,697][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 23:31:41,615][__main__][INFO] - Number of regex retries in iteration 361: 1 -[2025-11-09 23:31:41,616][__main__][INFO] - agents played in iteration 361 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:31:42,474][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:31:42,497][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:31:42,524][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:31:42,547][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:31:42,548][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:31:42,548][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:31:43,366][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:31:43,929][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:31:44,457][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:31:44,961][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:31:45,465][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:31:45,981][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:31:46,493][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:31:46,995][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:31:47,508][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:31:48,011][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:31:48,510][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:31:49,011][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:31:49,509][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:31:50,013][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:31:50,508][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:31:51,005][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:31:51,502][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:31:52,000][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:31:52,498][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:31:52,995][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:31:53,493][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:31:53,989][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:31:54,486][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:31:54,982][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:31:55,479][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:31:55,977][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:31:56,481][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:31:56,981][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:31:57,479][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:31:57,981][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:31:58,480][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:31:58,980][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:31:59,476][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:31:59,971][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:32:00,476][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:32:00,974][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:32:01,468][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:32:01,964][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:32:02,459][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:32:02,955][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:32:03,453][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:32:03,955][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:32:04,459][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:32:04,956][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:32:05,455][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:32:05,953][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:32:06,447][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:32:06,945][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:32:07,438][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:32:07,933][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:32:08,439][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:32:08,935][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:32:09,431][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:32:09,927][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:32:10,422][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:32:10,928][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:32:11,422][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:32:11,920][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:32:12,455][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:32:12,954][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:32:13,451][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:32:13,949][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:32:14,446][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:32:14,945][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:32:15,442][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10185 tokens. -[2025-11-09 23:32:16,211][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 55.68%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 -[2025-11-09 23:32:16,835][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:32:16,836][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:32:16,838][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:32:17,766][__main__][INFO] - Iteration 362 took 1m 1s (40.89% Gen, 57.59% Train). Generation: 25s, Training: 35s. Estimated remaining time: 45h 24m 48s. Estimated total time: 50h 58m 9s. Time estimates for 10 more iterations: 10m 11s, 100 more iterations: 1h 41m 56s, 500 more iterations: 8h 29m 41s. -[2025-11-09 23:32:17,768][__main__][INFO] - Starting iteration 362. -[2025-11-09 23:32:18,245][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 36 and human policies 1. -[2025-11-09 23:32:18,245][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:32:44,266][__main__][INFO] - Number of regex retries in iteration 362: 0 -[2025-11-09 23:32:44,266][__main__][INFO] - agents played in iteration 362 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:32:45,086][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:32:45,110][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:32:45,133][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:32:45,166][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:32:45,167][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:32:45,168][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:32:45,954][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:32:46,415][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:32:46,933][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:32:47,430][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:32:47,943][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:32:48,450][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:32:48,952][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:32:49,454][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:32:49,959][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:32:50,472][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:32:50,976][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:32:51,477][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:32:51,979][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:32:52,481][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:32:52,981][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:32:53,482][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:32:53,982][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:32:54,480][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:32:54,980][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:32:55,476][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:32:55,976][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:32:56,473][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:32:56,972][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:32:57,469][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:32:57,965][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:32:58,463][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:32:58,960][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:32:59,457][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:32:59,956][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:33:00,451][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:33:00,951][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:33:01,447][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:33:01,942][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:33:02,439][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:33:02,948][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:33:03,449][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:33:03,946][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:33:04,443][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:33:04,942][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:33:05,440][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:33:05,939][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:33:06,439][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:33:06,936][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:33:07,431][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:33:07,925][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:33:08,422][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:33:08,940][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:33:09,440][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:33:09,935][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:33:10,433][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:33:10,934][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:33:11,432][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:33:11,932][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:33:12,428][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:33:12,937][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:33:13,436][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:33:13,932][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:33:14,430][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:33:14,929][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:33:15,435][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:33:15,934][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:33:16,430][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:33:16,930][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:33:17,426][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:33:17,925][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10224 tokens. -[2025-11-09 23:33:18,633][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 55.64%, Block Peak % of device VRAM: 62.17%, ΔTime: 00:00:32 -[2025-11-09 23:33:19,364][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:33:19,366][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:33:19,367][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:33:20,441][__main__][INFO] - Iteration 363 took 1m 2s (41.84% Gen, 56.44% Train). Generation: 26s, Training: 35s. Estimated remaining time: 46h 15m 26s. Estimated total time: 51h 49m 49s. Time estimates for 10 more iterations: 10m 21s, 100 more iterations: 1h 43m 39s, 500 more iterations: 8h 38m 18s. -[2025-11-09 23:33:20,443][__main__][INFO] - Starting iteration 363. -[2025-11-09 23:33:20,906][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 36 and human policies 1. -[2025-11-09 23:33:20,906][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:33:49,877][__main__][INFO] - Number of regex retries in iteration 363: 0 -[2025-11-09 23:33:49,879][__main__][INFO] - agents played in iteration 363 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:33:50,715][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:33:50,739][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:33:50,766][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:33:50,789][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:33:50,790][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:33:50,791][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:33:51,585][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:33:52,049][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:33:52,567][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:33:53,068][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:33:53,572][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:33:54,077][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:33:54,579][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:33:55,087][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:33:55,587][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:33:56,095][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:33:56,601][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:33:57,103][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:33:57,605][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:33:58,106][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:33:58,608][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:33:59,109][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:33:59,608][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:34:00,106][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:34:00,609][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:34:01,109][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:34:01,610][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:34:02,110][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:34:02,606][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:34:03,108][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:34:03,607][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:34:04,106][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:34:04,604][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:34:05,106][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:34:05,606][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:34:06,110][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:34:06,607][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:34:07,106][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:34:07,609][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:34:08,110][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:34:08,607][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:34:09,107][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:34:09,606][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:34:10,106][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:34:10,605][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:34:11,110][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:34:11,608][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:34:12,109][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:34:12,608][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:34:13,109][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:34:13,609][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:34:14,116][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:34:14,619][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:34:15,122][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:34:15,622][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:34:16,119][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:34:16,617][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:34:17,114][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:34:17,613][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:34:18,111][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:34:18,607][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:34:19,106][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:34:19,602][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:34:20,102][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:34:20,601][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:34:21,099][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:34:21,599][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:34:22,100][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:34:22,596][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:34:23,095][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:34:23,591][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10185 tokens. -[2025-11-09 23:34:24,340][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 55.58%, Block Peak % of device VRAM: 62.24%, ΔTime: 00:00:32 -[2025-11-09 23:34:24,967][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:34:24,969][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:34:24,970][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:34:25,865][__main__][INFO] - Iteration 364 took 1m 4s (44.60% Gen, 54.02% Train). Generation: 28s, Training: 35s. Estimated remaining time: 48h 32m 29s. Estimated total time: 54h 7m 58s. Time estimates for 10 more iterations: 10m 49s, 100 more iterations: 1h 48m 15s, 500 more iterations: 9h 1m 19s. -[2025-11-09 23:34:25,867][__main__][INFO] - Starting iteration 364. -[2025-11-09 23:34:26,335][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 36 and human policies 1. -[2025-11-09 23:34:26,335][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:34:37,776][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 23:34:38,000][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 23:34:51,073][__main__][INFO] - Number of regex retries in iteration 364: 2 -[2025-11-09 23:34:51,075][__main__][INFO] - agents played in iteration 364 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:34:51,943][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:34:51,973][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:34:51,999][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:34:52,022][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:34:52,023][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:34:52,023][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:34:52,804][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:34:53,280][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:34:53,786][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:34:54,288][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:34:54,798][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:34:55,300][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:34:55,807][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:34:56,308][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:34:56,812][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:34:57,316][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:34:57,817][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:34:58,322][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:34:58,824][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:34:59,329][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:34:59,834][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:35:00,337][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:35:00,845][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:35:01,364][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:35:01,868][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:35:02,383][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:35:02,885][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:35:03,387][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:35:03,892][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:35:04,390][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:35:04,898][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:35:05,399][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:35:05,898][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:35:06,401][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:35:06,901][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:35:07,399][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:35:07,901][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:35:08,399][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:35:08,899][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:35:09,408][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:35:09,909][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:35:10,411][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:35:10,911][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:35:11,415][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:35:11,921][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:35:12,422][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:35:12,925][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:35:13,423][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:35:13,921][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:35:14,422][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:35:14,923][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:35:15,423][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:35:15,928][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:35:16,430][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:35:16,928][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:35:17,430][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:35:17,929][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:35:18,432][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:35:18,935][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:35:19,434][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:35:19,930][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:35:20,429][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:35:20,926][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:35:21,426][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:35:21,923][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:35:22,422][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:35:22,921][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:35:23,422][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:35:23,922][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:35:24,419][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:35:24,921][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10231 tokens. -[2025-11-09 23:35:25,658][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 55.74%, Block Peak % of device VRAM: 62.50%, ΔTime: 00:00:32 -[2025-11-09 23:35:26,288][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:35:26,290][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:35:26,292][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:35:27,204][__main__][INFO] - Iteration 365 took 1m 0s (40.64% Gen, 57.86% Train). Generation: 24s, Training: 35s. Estimated remaining time: 45h 7m 0s. Estimated total time: 50h 43m 30s. Time estimates for 10 more iterations: 10m 8s, 100 more iterations: 1h 41m 27s, 500 more iterations: 8h 27m 15s. -[2025-11-09 23:35:27,208][__main__][INFO] - Starting iteration 365. -[2025-11-09 23:35:27,700][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 36 and human policies 1. -[2025-11-09 23:35:27,700][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:35:52,825][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 23:35:57,813][__main__][INFO] - Number of regex retries in iteration 365: 1 -[2025-11-09 23:35:57,814][__main__][INFO] - agents played in iteration 365 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:35:58,644][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:35:58,667][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:35:58,691][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:35:58,713][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:35:58,714][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:35:58,716][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:35:59,481][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:35:59,941][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:36:00,452][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:36:00,951][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:36:01,459][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:36:01,962][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:36:02,466][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:36:02,985][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:36:03,488][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:36:04,009][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:36:04,511][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:36:05,016][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:36:05,517][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:36:06,022][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:36:06,525][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:36:07,025][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:36:07,528][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:36:08,029][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:36:08,532][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:36:09,034][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:36:09,533][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:36:10,035][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:36:10,532][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:36:11,029][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:36:11,527][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:36:12,025][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:36:12,526][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:36:13,024][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:36:13,523][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:36:14,020][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:36:14,524][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:36:15,020][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:36:15,521][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:36:16,018][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:36:16,516][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:36:17,013][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:36:17,511][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:36:18,008][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:36:18,512][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:36:19,010][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:36:19,509][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:36:20,008][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:36:20,512][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:36:21,017][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:36:21,516][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:36:22,017][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:36:22,520][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:36:23,023][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:36:23,529][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:36:24,028][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:36:24,525][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:36:25,026][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:36:25,524][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:36:26,020][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:36:26,517][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:36:27,017][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:36:27,522][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:36:28,019][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:36:28,517][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:36:29,013][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:36:29,513][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:36:30,018][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:36:30,517][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:36:31,018][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:36:31,516][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10226 tokens. -[2025-11-09 23:36:32,227][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.09%, Current % of VRAM taken: 55.58%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 -[2025-11-09 23:36:32,972][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:36:32,974][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:36:32,975][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:36:33,969][__main__][INFO] - Iteration 366 took 1m 6s (45.44% Gen, 53.06% Train). Generation: 30s, Training: 35s. Estimated remaining time: 49h 35m 52s. Estimated total time: 55h 13m 29s. Time estimates for 10 more iterations: 11m 2s, 100 more iterations: 1h 50m 26s, 500 more iterations: 9h 12m 14s. -[2025-11-09 23:36:33,971][__main__][INFO] - Starting iteration 366. -[2025-11-09 23:36:34,437][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 36 and human policies 1. -[2025-11-09 23:36:34,437][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:36:51,782][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 23:37:03,950][__main__][INFO] - Number of regex retries in iteration 366: 1 -[2025-11-09 23:37:03,950][__main__][INFO] - agents played in iteration 366 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:37:04,822][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:37:04,867][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:37:04,902][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:37:04,932][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:37:04,932][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:37:04,933][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:37:05,737][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:37:06,196][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:37:06,728][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:37:07,232][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:37:07,735][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:37:08,234][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:37:08,739][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:37:09,246][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:37:09,759][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:37:10,262][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:37:10,766][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:37:11,275][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:37:11,779][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:37:12,282][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:37:12,785][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:37:13,306][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:37:13,810][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:37:14,314][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:37:14,817][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:37:15,321][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:37:15,831][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:37:16,338][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:37:16,838][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:37:17,343][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:37:17,844][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:37:18,348][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:37:18,853][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:37:19,352][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:37:19,854][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:37:20,351][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:37:20,853][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:37:21,350][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:37:21,851][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:37:22,355][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:37:22,854][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:37:23,352][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:37:23,851][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:37:24,349][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:37:24,848][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:37:25,345][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:37:25,843][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:37:26,341][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:37:26,839][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:37:27,339][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:37:27,841][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:37:28,337][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:37:28,839][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:37:29,337][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:37:29,836][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:37:30,338][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:37:30,858][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:37:31,359][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:37:31,860][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:37:32,362][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:37:32,866][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:37:33,369][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:37:33,870][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:37:34,373][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:37:34,873][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:37:35,374][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:37:35,879][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:37:36,378][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:37:36,888][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:37:37,387][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:37:37,887][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10356 tokens. -[2025-11-09 23:37:38,603][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 55.74%, Block Peak % of device VRAM: 62.47%, ΔTime: 00:00:32 -[2025-11-09 23:37:39,258][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:37:39,260][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:37:39,261][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:37:40,164][__main__][INFO] - Iteration 367 took 1m 5s (44.90% Gen, 53.72% Train). Generation: 29s, Training: 35s. Estimated remaining time: 49h 7m 40s. Estimated total time: 54h 46m 23s. Time estimates for 10 more iterations: 10m 57s, 100 more iterations: 1h 49m 32s, 500 more iterations: 9h 7m 43s. -[2025-11-09 23:37:40,166][__main__][INFO] - Starting iteration 367. -[2025-11-09 23:37:40,641][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 36 and human policies 1. -[2025-11-09 23:37:40,642][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:38:04,426][__main__][INFO] - Number of regex retries in iteration 367: 0 -[2025-11-09 23:38:04,428][__main__][INFO] - agents played in iteration 367 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:38:05,272][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:38:05,311][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:38:05,341][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:38:05,368][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:38:05,369][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:38:05,370][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:38:06,178][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:38:06,638][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:38:07,150][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:38:07,658][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:38:08,161][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:38:08,658][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:38:09,160][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:38:09,667][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:38:10,168][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:38:10,674][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:38:11,179][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:38:11,682][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:38:12,187][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:38:12,689][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:38:13,192][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:38:13,705][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:38:14,210][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:38:14,709][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:38:15,243][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:38:15,748][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:38:16,256][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:38:16,762][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:38:17,265][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:38:17,772][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:38:18,279][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:38:18,782][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:38:19,282][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:38:19,783][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:38:20,291][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:38:20,792][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:38:21,292][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:38:21,794][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:38:22,295][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:38:22,799][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:38:23,301][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:38:23,804][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:38:24,316][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:38:24,818][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:38:25,320][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:38:25,820][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:38:26,320][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:38:26,841][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:38:27,341][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:38:27,836][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:38:28,337][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:38:28,836][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:38:29,340][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:38:29,837][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:38:30,333][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:38:30,833][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:38:31,330][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:38:31,827][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:38:32,330][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:38:32,830][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:38:33,327][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:38:33,826][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:38:34,325][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:38:34,838][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:38:35,338][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:38:35,836][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:38:36,335][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:38:36,834][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:38:37,335][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:38:37,838][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:38:38,338][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10360 tokens. -[2025-11-09 23:38:39,071][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 55.64%, Block Peak % of device VRAM: 62.38%, ΔTime: 00:00:32 -[2025-11-09 23:38:39,706][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:38:39,708][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:38:39,709][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:38:40,655][__main__][INFO] - Iteration 368 took 1m 0s (39.63% Gen, 58.79% Train). Generation: 23s, Training: 35s. Estimated remaining time: 44h 20m 58s. Estimated total time: 50h 0m 42s. Time estimates for 10 more iterations: 10m 0s, 100 more iterations: 1h 40m 1s, 500 more iterations: 8h 20m 7s. -[2025-11-09 23:38:40,657][__main__][INFO] - Starting iteration 368. -[2025-11-09 23:38:41,128][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 36 and human policies 1. -[2025-11-09 23:38:41,129][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:39:13,270][__main__][INFO] - Number of regex retries in iteration 368: 0 -[2025-11-09 23:39:13,270][__main__][INFO] - agents played in iteration 368 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:39:14,221][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:39:14,250][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:39:14,278][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:39:14,301][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:39:14,301][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:39:14,303][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:39:15,088][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:39:15,555][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:39:16,065][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:39:16,562][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:39:17,065][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:39:17,567][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:39:18,071][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:39:18,570][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:39:19,077][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:39:19,582][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:39:20,083][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:39:20,585][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:39:21,085][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:39:21,584][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:39:22,088][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:39:22,593][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:39:23,095][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:39:23,600][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:39:24,103][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:39:24,609][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:39:25,112][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:39:25,612][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:39:26,117][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:39:26,620][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:39:27,122][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:39:27,636][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:39:28,140][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:39:28,658][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:39:29,157][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:39:29,673][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:39:30,171][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:39:30,669][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:39:31,169][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:39:31,687][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:39:32,189][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:39:32,690][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:39:33,187][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:39:33,686][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:39:34,182][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:39:34,683][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:39:35,183][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:39:35,683][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:39:36,180][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:39:36,683][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:39:37,180][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:39:37,681][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:39:38,177][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:39:38,674][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:39:39,180][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:39:39,680][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:39:40,180][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:39:40,678][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:39:41,175][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:39:41,677][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:39:42,180][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:39:42,684][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:39:43,185][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:39:43,685][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:39:44,185][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:39:44,682][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:39:45,180][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:39:45,680][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:39:46,181][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:39:46,679][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:39:47,180][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10214 tokens. -[2025-11-09 23:39:47,985][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 55.70%, Block Peak % of device VRAM: 62.14%, ΔTime: 00:00:32 -[2025-11-09 23:39:48,736][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:39:48,738][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:39:48,740][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:39:49,770][__main__][INFO] - Iteration 369 took 1m 8s (46.82% Gen, 51.67% Train). Generation: 32s, Training: 35s. Estimated remaining time: 51h 31m 18s. Estimated total time: 57h 12m 10s. Time estimates for 10 more iterations: 11m 26s, 100 more iterations: 1h 54m 24s, 500 more iterations: 9h 32m 1s. -[2025-11-09 23:39:49,772][__main__][INFO] - Starting iteration 369. -[2025-11-09 23:39:50,309][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 36 and human policies 1. -[2025-11-09 23:39:50,309][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:40:21,542][__main__][INFO] - Number of regex retries in iteration 369: 0 -[2025-11-09 23:40:21,545][__main__][INFO] - agents played in iteration 369 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:40:22,365][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:40:22,389][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:40:22,412][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:40:22,435][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:40:22,435][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:40:22,436][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:40:23,216][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:40:23,677][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:40:24,186][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:40:24,696][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:40:25,199][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:40:25,706][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:40:26,209][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:40:26,710][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:40:27,212][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:40:27,716][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:40:28,217][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:40:28,723][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:40:29,225][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:40:29,729][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:40:30,238][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:40:30,744][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:40:31,252][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:40:31,754][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:40:32,255][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:40:32,759][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:40:33,263][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:40:33,764][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:40:34,268][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:40:34,767][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:40:35,286][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:40:35,789][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:40:36,292][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:40:36,799][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:40:37,303][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:40:37,813][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:40:38,317][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:40:38,818][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:40:39,320][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:40:39,825][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:40:40,324][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:40:40,827][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:40:41,327][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:40:41,830][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:40:42,329][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:40:42,829][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:40:43,330][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:40:43,830][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:40:44,331][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:40:44,829][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:40:45,331][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:40:45,832][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:40:46,334][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:40:46,831][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:40:47,333][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:40:47,834][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:40:48,335][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:40:48,838][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:40:49,338][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:40:49,840][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:40:50,341][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:40:50,841][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:40:51,353][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:40:51,852][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:40:52,351][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:40:52,850][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:40:53,349][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:40:53,868][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:40:54,372][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:40:54,880][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:40:55,380][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10337 tokens. -[2025-11-09 23:40:56,176][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.05%, Current % of VRAM taken: 55.55%, Block Peak % of device VRAM: 62.34%, ΔTime: 00:00:32 -[2025-11-09 23:40:56,797][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:40:56,798][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:40:56,800][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:40:57,722][__main__][INFO] - Iteration 370 took 1m 7s (46.33% Gen, 52.30% Train). Generation: 31s, Training: 35s. Estimated remaining time: 50h 28m 40s. Estimated total time: 56h 10m 41s. Time estimates for 10 more iterations: 11m 14s, 100 more iterations: 1h 52m 21s, 500 more iterations: 9h 21m 46s. -[2025-11-09 23:40:57,724][__main__][INFO] - Starting iteration 370. -[2025-11-09 23:40:58,190][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 36 and human policies 1. -[2025-11-09 23:40:58,191][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:41:11,576][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 23:41:23,946][__main__][INFO] - Number of regex retries in iteration 370: 1 -[2025-11-09 23:41:23,947][__main__][INFO] - agents played in iteration 370 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:41:24,762][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:41:24,789][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:41:24,815][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:41:24,838][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:41:24,838][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:41:24,839][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:41:25,531][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:41:25,987][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:41:26,505][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:41:27,005][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:41:27,509][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:41:28,010][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:41:28,512][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:41:29,020][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:41:29,523][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:41:30,025][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:41:30,528][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:41:31,031][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:41:31,530][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:41:32,032][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:41:32,531][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:41:34,225][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:41:34,850][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:41:35,351][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:41:35,853][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:41:36,353][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:41:36,851][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:41:37,364][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:41:37,867][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:41:38,383][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:41:38,886][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:41:39,399][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:41:39,908][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:41:40,413][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:41:40,920][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:41:41,431][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:41:41,933][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:41:42,434][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:41:42,942][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:41:43,441][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:41:43,943][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:41:44,447][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:41:44,948][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:41:45,447][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:41:45,953][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:41:46,453][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:41:46,956][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:41:47,455][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:41:47,954][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:41:48,454][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:41:48,950][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:41:49,451][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:41:49,950][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:41:50,454][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:41:50,953][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:41:51,463][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:41:51,969][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:41:52,471][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:41:52,975][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:41:53,480][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:41:54,002][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:41:54,506][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:41:55,011][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:41:55,513][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:41:56,016][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:41:56,524][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:41:57,028][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:41:57,532][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:41:58,061][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:41:58,568][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:41:59,073][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10393 tokens. -[2025-11-09 23:41:59,872][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.25%, Current % of VRAM taken: 55.74%, Block Peak % of device VRAM: 62.46%, ΔTime: 00:00:34 -[2025-11-09 23:42:00,523][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:42:00,525][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:42:00,528][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:42:02,436][__main__][INFO] - Iteration 371 took 1m 4s (40.09% Gen, 56.94% Train). Generation: 25s, Training: 36s. Estimated remaining time: 47h 49m 17s. Estimated total time: 53h 32m 22s. Time estimates for 10 more iterations: 10m 42s, 100 more iterations: 1h 47m 4s, 500 more iterations: 8h 55m 23s. -[2025-11-09 23:42:02,439][__main__][INFO] - Starting iteration 371. -[2025-11-09 23:42:03,035][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 37 and human policies 1. -[2025-11-09 23:42:03,035][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:42:36,316][__main__][INFO] - Number of regex retries in iteration 371: 0 -[2025-11-09 23:42:36,317][__main__][INFO] - agents played in iteration 371 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:42:37,150][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:42:37,172][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:42:37,197][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:42:37,219][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:42:37,220][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:42:37,220][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:42:38,007][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:42:38,468][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:42:38,979][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:42:39,483][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:42:39,986][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:42:40,507][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:42:41,011][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:42:41,515][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:42:42,020][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:42:42,524][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:42:43,027][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:42:43,528][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:42:44,031][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:42:44,537][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:42:45,042][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:42:45,546][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:42:46,058][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:42:46,563][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:42:47,082][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:42:47,587][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:42:48,088][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:42:48,587][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:42:49,091][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:42:49,607][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:42:50,108][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:42:50,612][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:42:51,115][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:42:51,618][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:42:52,127][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:42:52,628][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:42:53,136][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:42:53,643][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:42:54,143][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:42:54,641][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:42:55,141][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:42:55,637][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:42:56,140][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:42:56,636][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:42:57,132][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:42:57,636][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:42:58,139][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:42:58,639][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:42:59,142][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:42:59,644][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:43:00,147][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:43:00,648][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:43:01,153][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:43:01,664][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:43:02,167][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:43:02,677][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:43:03,188][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:43:03,691][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:43:04,211][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:43:04,715][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:43:05,221][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:43:05,727][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:43:06,233][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:43:06,738][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:43:07,237][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:43:07,739][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:43:08,242][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:43:08,740][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:43:09,242][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:43:09,750][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:43:10,256][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10400 tokens. -[2025-11-09 23:43:11,010][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 55.68%, Block Peak % of device VRAM: 62.47%, ΔTime: 00:00:33 -[2025-11-09 23:43:11,734][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:43:11,736][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:43:11,739][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:43:12,743][__main__][INFO] - Iteration 372 took 1m 9s (47.74% Gen, 50.81% Train). Generation: 33s, Training: 35s. Estimated remaining time: 52h 21m 11s. Estimated total time: 58h 5m 27s. Time estimates for 10 more iterations: 11m 37s, 100 more iterations: 1h 56m 10s, 500 more iterations: 9h 40m 54s. -[2025-11-09 23:43:12,745][__main__][INFO] - Starting iteration 372. -[2025-11-09 23:43:13,232][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 37 and human policies 1. -[2025-11-09 23:43:13,233][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:43:34,864][__main__][INFO] - Number of regex retries in iteration 372: 0 -[2025-11-09 23:43:34,866][__main__][INFO] - agents played in iteration 372 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:43:35,721][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:43:35,767][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:43:35,800][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:43:35,827][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:43:35,828][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:43:35,829][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:43:36,609][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:43:37,078][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:43:37,584][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:43:38,100][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:43:38,601][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:43:39,104][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:43:39,620][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:43:40,122][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:43:40,627][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:43:41,126][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:43:41,623][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:43:42,125][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:43:42,625][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:43:43,125][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:43:43,624][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:43:44,128][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:43:44,636][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:43:45,136][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:43:45,639][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:43:46,142][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:43:46,647][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:43:47,150][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:43:47,650][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:43:48,150][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:43:48,660][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:43:49,165][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:43:49,670][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:43:50,185][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:43:50,690][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:43:51,190][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:43:51,700][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:43:52,202][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:43:52,719][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:43:53,221][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:43:53,723][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:43:54,224][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:43:54,725][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:43:55,232][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:43:55,738][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:43:56,239][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:43:56,743][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:43:57,247][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:43:57,747][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:43:58,250][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:43:58,750][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:43:59,257][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:43:59,760][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:44:00,263][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:44:00,766][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:44:01,275][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:44:01,800][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:44:02,306][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:44:02,811][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:44:03,315][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:44:03,819][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:44:04,327][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:44:04,831][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:44:05,340][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:44:05,843][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:44:06,349][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:44:06,852][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:44:07,355][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:44:07,862][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:44:08,366][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:44:08,866][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10448 tokens. -[2025-11-09 23:44:09,664][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 55.66%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:33 -[2025-11-09 23:44:10,304][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:44:10,305][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:44:10,307][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:44:11,232][__main__][INFO] - Iteration 373 took 58s (37.30% Gen, 61.10% Train). Generation: 21s, Training: 35s. Estimated remaining time: 42h 34m 48s. Estimated total time: 48h 20m 2s. Time estimates for 10 more iterations: 9m 40s, 100 more iterations: 1h 36m 40s, 500 more iterations: 8h 3m 20s. -[2025-11-09 23:44:11,235][__main__][INFO] - Starting iteration 373. -[2025-11-09 23:44:11,699][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 37 and human policies 1. -[2025-11-09 23:44:11,700][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:44:30,664][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 23:44:42,733][__main__][INFO] - Number of regex retries in iteration 373: 1 -[2025-11-09 23:44:42,734][__main__][INFO] - agents played in iteration 373 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:44:43,525][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:44:43,551][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:44:43,574][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:44:43,597][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:44:43,597][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:44:43,598][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:44:44,306][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:44:44,764][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:44:45,263][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:44:45,758][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:44:46,253][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:44:46,749][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:44:47,250][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:44:47,747][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:44:48,244][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:44:48,741][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:44:49,240][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:44:49,738][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:44:50,236][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:44:50,736][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:44:51,231][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:44:51,733][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:44:52,233][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:44:52,730][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:44:53,227][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:44:53,725][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:44:54,233][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:44:54,729][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:44:55,230][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:44:55,736][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:44:56,239][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:44:56,739][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:44:57,239][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:44:57,743][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:44:58,250][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:44:58,751][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:44:59,251][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:44:59,755][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:45:00,254][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:45:00,765][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:45:01,267][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:45:01,771][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:45:02,275][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:45:02,775][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:45:03,278][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:45:05,488][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:45:05,997][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:45:06,499][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:45:07,000][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:45:07,502][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:45:08,006][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:45:08,506][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:45:09,009][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:45:09,515][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:45:10,019][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:45:10,523][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:45:11,025][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:45:11,524][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:45:12,028][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:45:12,530][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:45:13,036][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:45:13,536][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:45:14,041][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:45:14,542][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:45:15,043][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:45:15,544][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:45:16,048][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:45:16,553][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:45:17,055][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:45:17,555][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:45:18,061][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10193 tokens. -[2025-11-09 23:45:18,908][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 55.60%, Block Peak % of device VRAM: 62.24%, ΔTime: 00:00:34 -[2025-11-09 23:45:19,581][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:45:19,584][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:45:19,586][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:45:20,513][__main__][INFO] - Iteration 374 took 1m 8s (45.10% Gen, 53.55% Train). Generation: 31s, Training: 36s. Estimated remaining time: 51h 34m 19s. Estimated total time: 57h 20m 42s. Time estimates for 10 more iterations: 11m 28s, 100 more iterations: 1h 54m 41s, 500 more iterations: 9h 33m 27s. -[2025-11-09 23:45:20,519][__main__][INFO] - Starting iteration 374. -[2025-11-09 23:45:21,013][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 37 and human policies 1. -[2025-11-09 23:45:21,014][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:45:40,302][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 23:45:52,343][__main__][INFO] - Number of regex retries in iteration 374: 1 -[2025-11-09 23:45:52,343][__main__][INFO] - agents played in iteration 374 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:45:53,145][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:45:53,168][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:45:53,192][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:45:53,214][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:45:53,215][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:45:53,215][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:45:53,986][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:45:54,444][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:45:54,954][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:45:55,454][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:45:55,954][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:45:56,455][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:45:56,965][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:45:57,470][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:45:57,970][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:45:58,468][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:45:58,969][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:45:59,470][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:45:59,983][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:46:00,486][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:46:00,986][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:46:01,485][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:46:01,985][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:46:02,487][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:46:02,994][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:46:03,495][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:46:03,996][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:46:04,499][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:46:05,004][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:46:05,524][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:46:06,031][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:46:06,532][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:46:07,033][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:46:07,541][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:46:08,045][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:46:08,547][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:46:09,049][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:46:09,554][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:46:10,061][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:46:10,561][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:46:11,066][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:46:11,567][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:46:12,071][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:46:12,572][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:46:13,074][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:46:13,579][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:46:14,079][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:46:14,581][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:46:15,083][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:46:15,584][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:46:16,085][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:46:16,585][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:46:17,086][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:46:17,586][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:46:18,087][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:46:18,590][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:46:19,091][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:46:19,591][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:46:20,097][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:46:20,602][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:46:21,106][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:46:21,621][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:46:22,125][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:46:22,639][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:46:23,141][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:46:23,643][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:46:24,146][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:46:24,649][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:46:25,148][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:46:25,649][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:46:26,149][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10260 tokens. -[2025-11-09 23:46:26,914][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 55.72%, Block Peak % of device VRAM: 62.21%, ΔTime: 00:00:32 -[2025-11-09 23:46:27,646][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:46:27,648][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:46:27,649][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:46:28,676][__main__][INFO] - Iteration 375 took 1m 7s (46.30% Gen, 52.18% Train). Generation: 31s, Training: 35s. Estimated remaining time: 50h 35m 39s. Estimated total time: 56h 23m 11s. Time estimates for 10 more iterations: 11m 16s, 100 more iterations: 1h 52m 46s, 500 more iterations: 9h 23m 51s. -[2025-11-09 23:46:28,679][__main__][INFO] - Starting iteration 375. -[2025-11-09 23:46:29,149][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 37 and human policies 1. -[2025-11-09 23:46:29,149][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:46:58,633][__main__][INFO] - Number of regex retries in iteration 375: 0 -[2025-11-09 23:46:58,635][__main__][INFO] - agents played in iteration 375 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:46:59,445][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:46:59,482][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:46:59,513][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:46:59,539][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:46:59,539][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:46:59,540][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:47:00,311][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:47:00,772][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:47:01,276][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:47:01,778][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:47:02,287][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:47:02,789][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:47:03,293][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:47:03,794][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:47:04,297][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:47:04,800][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:47:05,304][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:47:05,806][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:47:06,325][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:47:06,826][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:47:07,326][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:47:07,834][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:47:08,333][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:47:08,841][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:47:09,345][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:47:09,848][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:47:10,353][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:47:10,858][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:47:11,364][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:47:11,868][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:47:12,389][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:47:12,893][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:47:13,401][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:47:13,904][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:47:14,409][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:47:14,912][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:47:15,413][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:47:15,915][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:47:16,422][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:47:16,922][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:47:17,427][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:47:17,945][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:47:18,450][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:47:18,952][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:47:19,456][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:47:19,961][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:47:20,462][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:47:20,968][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:47:21,473][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:47:21,977][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:47:22,481][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:47:22,986][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:47:23,489][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:47:23,993][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:47:24,500][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:47:25,004][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:47:25,506][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:47:26,012][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:47:26,515][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:47:27,022][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:47:27,523][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:47:28,027][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:47:28,526][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:47:29,030][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:47:29,534][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:47:30,035][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:47:30,534][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:47:31,035][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:47:31,534][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:47:32,035][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:47:32,551][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10338 tokens. -[2025-11-09 23:47:33,283][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.25%, Current % of VRAM taken: 55.75%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:32 -[2025-11-09 23:47:33,904][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:47:33,905][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:47:33,907][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:47:34,826][__main__][INFO] - Iteration 376 took 1m 5s (44.89% Gen, 53.71% Train). Generation: 29s, Training: 35s. Estimated remaining time: 48h 55m 15s. Estimated total time: 54h 43m 53s. Time estimates for 10 more iterations: 10m 56s, 100 more iterations: 1h 49m 27s, 500 more iterations: 9h 7m 18s. -[2025-11-09 23:47:34,828][__main__][INFO] - Starting iteration 376. -[2025-11-09 23:47:35,296][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 37 and human policies 1. -[2025-11-09 23:47:35,297][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:47:47,210][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 23:47:58,476][__main__][INFO] - Number of regex retries in iteration 376: 1 -[2025-11-09 23:47:58,477][__main__][INFO] - agents played in iteration 376 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:47:59,275][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:47:59,303][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:47:59,331][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:47:59,354][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:47:59,355][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:47:59,356][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:48:00,140][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:48:00,597][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:48:01,108][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:48:01,605][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:48:02,108][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:48:02,612][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:48:03,112][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:48:03,611][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:48:04,112][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:48:04,618][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:48:05,121][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:48:05,626][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:48:06,130][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:48:06,635][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:48:07,144][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:48:07,650][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:48:08,154][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:48:08,654][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:48:09,169][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:48:09,669][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:48:10,172][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:48:10,672][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:48:11,171][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:48:11,673][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:48:12,176][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:48:12,678][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:48:13,185][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:48:13,689][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:48:14,190][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:48:14,696][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:48:15,198][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:48:15,703][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:48:16,206][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:48:16,711][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:48:17,212][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:48:17,714][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:48:18,215][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:48:18,715][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:48:19,215][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:48:19,723][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:48:20,225][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:48:20,733][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:48:21,239][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:48:21,743][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:48:22,253][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:48:22,757][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:48:23,769][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:48:24,790][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:48:25,338][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:48:25,839][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:48:26,344][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:48:26,849][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:48:27,364][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:48:27,865][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:48:28,368][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:48:28,874][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:48:29,371][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:48:29,876][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:48:30,373][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:48:30,870][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:48:31,383][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:48:31,886][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:48:32,387][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:48:32,894][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:48:33,396][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10224 tokens. -[2025-11-09 23:48:34,176][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.01%, Current % of VRAM taken: 55.51%, Block Peak % of device VRAM: 62.39%, ΔTime: 00:00:34 -[2025-11-09 23:48:34,808][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:48:34,810][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:48:34,812][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:48:35,735][__main__][INFO] - Iteration 377 took 1m 0s (38.35% Gen, 60.12% Train). Generation: 23s, Training: 36s. Estimated remaining time: 44h 32m 19s. Estimated total time: 50h 21m 58s. Time estimates for 10 more iterations: 10m 4s, 100 more iterations: 1h 40m 43s, 500 more iterations: 8h 23m 39s. -[2025-11-09 23:48:35,737][__main__][INFO] - Starting iteration 377. -[2025-11-09 23:48:36,224][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 37 and human policies 1. -[2025-11-09 23:48:36,226][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:49:07,410][__main__][INFO] - Number of regex retries in iteration 377: 0 -[2025-11-09 23:49:07,411][__main__][INFO] - agents played in iteration 377 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:49:08,216][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:49:08,239][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:49:08,264][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:49:08,287][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:49:08,288][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:49:08,289][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:49:09,086][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:49:09,545][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:49:10,057][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:49:10,573][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:49:11,077][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:49:11,579][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:49:12,084][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:49:12,587][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:49:13,087][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:49:13,593][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:49:14,096][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:49:14,596][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:49:15,101][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:49:15,606][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:49:16,111][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:49:16,627][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:49:17,136][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:49:17,644][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:49:18,159][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:49:18,672][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:49:19,177][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:49:19,690][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:49:20,193][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:49:20,699][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:49:21,209][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:49:21,744][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:49:22,259][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:49:22,765][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:49:23,270][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:49:23,785][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:49:24,289][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:49:24,797][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:49:25,301][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:49:25,804][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:49:26,311][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:49:26,814][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:49:27,317][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:49:27,821][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:49:28,323][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:49:28,829][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:49:29,334][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:49:29,837][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:49:30,344][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:49:30,845][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:49:31,350][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:49:31,864][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:49:32,366][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:49:32,871][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:49:33,369][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:49:33,867][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:49:34,370][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:49:34,870][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:49:35,370][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:49:35,869][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:49:36,372][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:49:36,869][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:49:37,368][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:49:37,870][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:49:38,371][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:49:38,873][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:49:39,390][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:49:39,893][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:49:40,395][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:49:40,897][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:49:41,398][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10424 tokens. -[2025-11-09 23:49:42,184][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 55.68%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:33 -[2025-11-09 23:49:42,917][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:49:42,919][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:49:42,920][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:49:43,979][__main__][INFO] - Iteration 378 took 1m 7s (46.02% Gen, 52.41% Train). Generation: 31s, Training: 35s. Estimated remaining time: 50h 36m 59s. Estimated total time: 56h 27m 46s. Time estimates for 10 more iterations: 11m 17s, 100 more iterations: 1h 52m 55s, 500 more iterations: 9h 24m 37s. -[2025-11-09 23:49:43,981][__main__][INFO] - Starting iteration 378. -[2025-11-09 23:49:44,444][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 37 and human policies 1. -[2025-11-09 23:49:44,445][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:50:09,019][__main__][INFO] - Number of regex retries in iteration 378: 0 -[2025-11-09 23:50:09,020][__main__][INFO] - agents played in iteration 378 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:50:09,821][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:50:09,844][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:50:09,868][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:50:09,890][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:50:09,891][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:50:09,892][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:50:10,675][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:50:11,144][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:50:11,649][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:50:12,151][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:50:12,652][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:50:13,157][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:50:13,663][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:50:14,163][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:50:14,663][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:50:15,212][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:50:15,719][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:50:16,223][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:50:16,728][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:50:17,236][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:50:17,744][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:50:18,248][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:50:18,758][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:50:20,654][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:50:21,162][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:50:21,667][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:50:22,171][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:50:22,676][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:50:23,181][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:50:23,684][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:50:24,204][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:50:24,712][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:50:25,220][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:50:25,727][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:50:26,235][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:50:26,739][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:50:27,250][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:50:27,759][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:50:28,266][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:50:28,766][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:50:29,268][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:50:29,769][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:50:30,273][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:50:30,773][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:50:31,275][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:50:31,781][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:50:32,286][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:50:32,790][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:50:33,307][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:50:33,809][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:50:34,318][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:50:34,822][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:50:35,328][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:50:35,830][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:50:36,332][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:50:36,833][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:50:37,333][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:50:37,837][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:50:38,340][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:50:38,841][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:50:39,342][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:50:39,845][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:50:40,345][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:50:40,845][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:50:41,343][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:50:41,841][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:50:42,345][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:50:42,844][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:50:43,341][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:50:43,843][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:50:44,341][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10395 tokens. -[2025-11-09 23:50:45,107][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 55.67%, Block Peak % of device VRAM: 62.52%, ΔTime: 00:00:34 -[2025-11-09 23:50:45,727][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:50:45,729][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:50:45,731][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:50:46,683][__main__][INFO] - Iteration 379 took 1m 2s (39.48% Gen, 58.98% Train). Generation: 24s, Training: 36s. Estimated remaining time: 46h 0m 9s. Estimated total time: 51h 51m 59s. Time estimates for 10 more iterations: 10m 22s, 100 more iterations: 1h 43m 43s, 500 more iterations: 8h 38m 39s. -[2025-11-09 23:50:46,685][__main__][INFO] - Starting iteration 379. -[2025-11-09 23:50:47,151][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 37 and human policies 1. -[2025-11-09 23:50:47,152][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:51:16,981][__main__][INFO] - Number of regex retries in iteration 379: 0 -[2025-11-09 23:51:16,982][__main__][INFO] - agents played in iteration 379 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:51:17,790][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:51:17,817][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:51:17,843][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:51:17,867][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:51:17,867][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:51:17,868][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:51:18,662][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:51:19,134][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:51:19,646][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:51:20,151][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:51:20,657][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:51:21,160][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:51:21,665][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:51:22,169][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:51:22,667][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:51:23,170][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:51:23,672][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:51:24,174][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:51:24,678][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:51:25,180][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:51:25,684][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:51:26,185][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:51:26,693][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:51:27,207][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:51:27,713][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:51:28,219][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:51:28,728][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:51:29,233][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:51:29,750][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:51:30,255][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:51:30,762][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:51:31,271][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:51:31,773][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:51:32,276][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:51:32,778][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:51:33,281][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:51:33,784][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:51:34,288][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:51:34,791][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:51:35,291][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:51:35,791][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:51:36,293][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:51:36,792][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:51:37,291][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:51:37,791][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:51:38,292][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:51:38,790][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:51:40,028][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:51:40,590][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:51:41,089][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:51:41,589][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:51:42,095][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:51:42,595][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:51:43,097][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:51:43,603][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:51:44,105][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:51:44,614][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:51:45,122][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:51:45,636][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:51:46,142][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:51:46,643][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:51:47,147][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:51:47,651][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:51:48,156][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:51:48,658][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:51:49,160][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:51:49,666][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:51:50,166][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:51:50,671][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:51:51,182][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:51:51,686][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10206 tokens. -[2025-11-09 23:51:52,505][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 55.68%, Block Peak % of device VRAM: 62.44%, ΔTime: 00:00:33 -[2025-11-09 23:51:53,172][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:51:53,174][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:51:53,176][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:51:54,196][__main__][INFO] - Iteration 380 took 1m 7s (44.49% Gen, 53.98% Train). Generation: 29s, Training: 36s. Estimated remaining time: 49h 59m 20s. Estimated total time: 55h 52m 17s. Time estimates for 10 more iterations: 11m 10s, 100 more iterations: 1h 51m 44s, 500 more iterations: 9h 18m 42s. -[2025-11-09 23:51:54,200][__main__][INFO] - Starting iteration 380. -[2025-11-09 23:51:54,666][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 37 and human policies 1. -[2025-11-09 23:51:54,667][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:52:15,013][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 23:52:26,728][__main__][INFO] - Number of regex retries in iteration 380: 1 -[2025-11-09 23:52:26,729][__main__][INFO] - agents played in iteration 380 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:52:27,553][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:52:27,581][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:52:27,609][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:52:27,634][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:52:27,634][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:52:27,635][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:52:28,444][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:52:28,904][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:52:29,412][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:52:29,915][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:52:30,434][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:52:30,936][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:52:31,438][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:52:31,941][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:52:32,445][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:52:32,964][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:52:33,465][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:52:33,968][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:52:34,473][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:52:34,977][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:52:35,485][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:52:35,988][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:52:36,494][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:52:37,004][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:52:37,506][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:52:38,013][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:52:38,525][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:52:39,028][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:52:39,544][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:52:40,048][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:52:40,553][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:52:41,063][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:52:41,567][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:52:42,074][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:52:42,578][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:52:43,084][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:52:43,589][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:52:44,092][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:52:44,594][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:52:45,092][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:52:45,593][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:52:46,094][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:52:46,591][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:52:47,094][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:52:47,595][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:52:48,094][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:52:48,591][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:52:49,093][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:52:49,592][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:52:50,095][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:52:50,600][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:52:51,097][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:52:51,596][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:52:52,097][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:52:52,596][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:52:53,095][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:52:53,598][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:52:54,102][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:52:54,601][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:52:55,103][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:52:55,604][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:52:56,105][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:52:56,617][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:52:57,116][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:52:57,619][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:52:58,129][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:52:58,627][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:52:59,122][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:52:59,623][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:53:00,117][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:53:00,614][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10194 tokens. -[2025-11-09 23:53:01,351][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.00%, Current % of VRAM taken: 55.50%, Block Peak % of device VRAM: 62.33%, ΔTime: 00:00:32 -[2025-11-09 23:53:02,128][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:53:02,130][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:53:02,131][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:53:04,770][__main__][INFO] - Iteration 381 took 1m 10s (45.74% Gen, 50.50% Train). Generation: 32s, Training: 35s. Estimated remaining time: 52h 31m 3s. Estimated total time: 58h 25m 11s. Time estimates for 10 more iterations: 11m 41s, 100 more iterations: 1h 56m 50s, 500 more iterations: 9h 44m 11s. -[2025-11-09 23:53:04,772][__main__][INFO] - Starting iteration 381. -[2025-11-09 23:53:05,236][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 38 and human policies 1. -[2025-11-09 23:53:05,236][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:53:29,203][__main__][INFO] - Number of regex retries in iteration 381: 0 -[2025-11-09 23:53:29,204][__main__][INFO] - agents played in iteration 381 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:53:30,230][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:53:30,257][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:53:30,284][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:53:30,310][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:53:30,311][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:53:30,312][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:53:31,113][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:53:31,583][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:53:32,093][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:53:32,595][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:53:33,100][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:53:33,603][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:53:34,106][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:53:34,609][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:53:35,110][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:53:36,407][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:53:37,005][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:53:37,509][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:53:38,018][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:53:38,528][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:53:39,030][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:53:39,532][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:53:40,038][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:53:40,545][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:53:41,047][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:53:41,549][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:53:42,052][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:53:42,558][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:53:43,059][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:53:43,570][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:53:44,071][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:53:44,574][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:53:45,084][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:53:45,590][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:53:46,110][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:53:46,619][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:53:47,123][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:53:47,627][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:53:48,134][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:53:48,634][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:53:49,140][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:53:49,641][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:53:50,139][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:53:50,647][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:53:51,147][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:53:51,651][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:53:52,150][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:53:52,654][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:53:53,156][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:53:53,658][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:53:54,160][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:53:54,662][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:53:55,164][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:53:55,670][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:53:56,173][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:53:56,694][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:53:57,199][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:53:57,701][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:53:58,206][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:53:58,706][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:53:59,207][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:53:59,707][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:54:00,202][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:54:00,700][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:54:01,196][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:54:01,691][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:54:02,189][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:54:02,686][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:54:03,181][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:54:03,680][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:54:04,175][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10229 tokens. -[2025-11-09 23:54:04,940][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 55.56%, Block Peak % of device VRAM: 62.37%, ΔTime: 00:00:33 -[2025-11-09 23:54:05,577][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:54:05,579][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:54:05,581][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:54:06,506][__main__][INFO] - Iteration 382 took 1m 1s (39.12% Gen, 59.37% Train). Generation: 23s, Training: 36s. Estimated remaining time: 45h 8m 24s. Estimated total time: 51h 3m 33s. Time estimates for 10 more iterations: 10m 12s, 100 more iterations: 1h 42m 7s, 500 more iterations: 8h 30m 35s. -[2025-11-09 23:54:06,509][__main__][INFO] - Starting iteration 382. -[2025-11-09 23:54:06,972][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 38 and human policies 1. -[2025-11-09 23:54:06,972][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:54:34,090][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 23:54:38,669][__main__][INFO] - Number of regex retries in iteration 382: 1 -[2025-11-09 23:54:38,669][__main__][INFO] - agents played in iteration 382 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:54:39,488][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:54:39,511][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:54:39,534][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:54:39,559][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:54:39,560][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:54:39,561][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:54:40,381][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:54:40,846][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:54:41,358][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:54:41,872][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:54:42,378][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:54:42,885][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:54:43,391][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:54:43,895][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:54:44,410][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:54:44,917][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:54:45,424][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:54:45,935][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:54:46,442][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:54:46,946][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:54:47,448][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:54:47,951][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:54:48,457][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:54:48,956][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:54:49,455][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:54:49,963][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:54:50,464][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:54:50,967][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:54:51,464][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:54:51,968][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:54:52,487][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:54:52,987][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:54:53,489][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:54:53,991][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:54:54,495][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:54:55,006][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:54:55,504][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:54:56,009][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:54:56,510][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:54:57,012][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:54:57,515][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:54:58,015][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:54:58,518][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:54:59,026][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:54:59,528][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:55:00,028][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:55:00,526][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:55:01,028][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:55:01,532][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:55:02,036][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:55:02,536][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:55:03,039][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:55:03,539][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:55:04,038][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:55:04,538][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:55:05,037][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:55:05,533][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:55:06,029][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:55:06,529][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:55:07,027][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:55:07,526][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:55:08,023][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:55:08,520][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:55:09,018][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:55:09,521][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:55:10,021][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:55:10,518][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:55:11,023][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:55:11,519][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:55:12,018][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:55:12,513][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10253 tokens. -[2025-11-09 23:55:13,235][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 55.68%, Block Peak % of device VRAM: 62.42%, ΔTime: 00:00:32 -[2025-11-09 23:55:13,980][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:55:13,982][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:55:13,984][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:55:15,039][__main__][INFO] - Iteration 383 took 1m 8s (46.57% Gen, 51.88% Train). Generation: 31s, Training: 35s. Estimated remaining time: 50h 47m 7s. Estimated total time: 56h 43m 25s. Time estimates for 10 more iterations: 11m 20s, 100 more iterations: 1h 53m 26s, 500 more iterations: 9h 27m 14s. -[2025-11-09 23:55:15,041][__main__][INFO] - Starting iteration 383. -[2025-11-09 23:55:15,502][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 38 and human policies 1. -[2025-11-09 23:55:15,503][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:55:42,026][__main__][INFO] - Number of regex retries in iteration 383: 0 -[2025-11-09 23:55:42,028][__main__][INFO] - agents played in iteration 383 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:55:42,880][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:55:42,910][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:55:42,936][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:55:42,960][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:55:42,961][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:55:42,962][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:55:43,781][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:55:44,246][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:55:44,753][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:55:45,260][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:55:45,764][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:55:46,270][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:55:46,772][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:55:47,282][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:55:47,800][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:55:48,306][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:55:48,809][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:55:49,313][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:55:49,818][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:55:50,328][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:55:50,835][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:55:51,336][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:55:51,840][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:55:52,343][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:55:52,846][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:55:53,347][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:55:53,844][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:55:54,344][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:55:54,841][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:55:55,338][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:55:55,839][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:55:56,338][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:55:56,837][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:55:57,336][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:55:57,838][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:55:58,342][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:55:58,840][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:55:59,339][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:55:59,844][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:56:00,342][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:56:00,844][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:56:01,342][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:56:01,839][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:56:02,341][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:56:02,841][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:56:03,341][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:56:03,846][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:56:04,353][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:56:04,852][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:56:05,354][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:56:05,859][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:56:06,361][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:56:06,859][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:56:07,356][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:56:07,859][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:56:08,361][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:56:08,863][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:56:09,366][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:56:09,864][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:56:10,367][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:56:10,864][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:56:11,363][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:56:11,864][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:56:12,363][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:56:12,863][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:56:13,363][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:56:13,862][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:56:14,361][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:56:14,861][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:56:15,358][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:56:15,859][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10253 tokens. -[2025-11-09 23:56:16,610][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 55.74%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 -[2025-11-09 23:56:17,248][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:56:17,250][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:56:17,252][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:56:18,174][__main__][INFO] - Iteration 384 took 1m 2s (42.32% Gen, 56.20% Train). Generation: 26s, Training: 35s. Estimated remaining time: 46h 16m 17s. Estimated total time: 52h 13m 38s. Time estimates for 10 more iterations: 10m 26s, 100 more iterations: 1h 44m 27s, 500 more iterations: 8h 42m 16s. -[2025-11-09 23:56:18,177][__main__][INFO] - Starting iteration 384. -[2025-11-09 23:56:18,643][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 38 and human policies 1. -[2025-11-09 23:56:18,643][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:56:35,600][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 23:56:48,280][__main__][INFO] - Number of regex retries in iteration 384: 1 -[2025-11-09 23:56:48,284][__main__][INFO] - agents played in iteration 384 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:56:49,199][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:56:49,222][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:56:49,245][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:56:49,268][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:56:49,268][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:56:49,269][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:56:50,088][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:56:50,634][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:56:51,149][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:56:51,659][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:56:52,168][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:56:52,672][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:56:53,176][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:56:53,681][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:56:54,188][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:56:54,688][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:56:55,195][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:56:55,698][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:56:56,199][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:56:56,707][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:56:57,213][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:56:57,723][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:56:58,231][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:56:58,734][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:56:59,242][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:56:59,745][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:57:00,247][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:57:00,748][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:57:01,250][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:57:01,752][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:57:02,253][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:57:02,750][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:57:03,258][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:57:03,770][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:57:04,277][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:57:04,780][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:57:05,283][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:57:05,785][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:57:06,285][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:57:06,791][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:57:07,299][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:57:07,800][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:57:08,301][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:57:08,804][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:57:09,305][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:57:09,808][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:57:10,309][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:57:10,806][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:57:11,307][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:57:11,809][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:57:12,310][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:57:12,812][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:57:13,309][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:57:13,821][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:57:14,322][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:57:14,826][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:57:15,328][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:57:15,832][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:57:16,335][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:57:16,833][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:57:17,329][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:57:17,837][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:57:18,334][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:57:18,847][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:57:19,343][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:57:19,840][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:57:20,346][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:57:20,841][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:57:21,349][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:57:21,848][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:57:22,347][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10272 tokens. -[2025-11-09 23:57:23,094][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 55.64%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:33 -[2025-11-09 23:57:23,737][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:57:23,738][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:57:23,740][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:57:24,721][__main__][INFO] - Iteration 385 took 1m 6s (44.86% Gen, 53.66% Train). Generation: 29s, Training: 35s. Estimated remaining time: 49h 5m 28s. Estimated total time: 55h 3m 56s. Time estimates for 10 more iterations: 11m 0s, 100 more iterations: 1h 50m 7s, 500 more iterations: 9h 10m 39s. -[2025-11-09 23:57:24,724][__main__][INFO] - Starting iteration 385. -[2025-11-09 23:57:25,195][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 38 and human policies 1. -[2025-11-09 23:57:25,195][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:57:56,899][__main__][INFO] - Number of regex retries in iteration 385: 0 -[2025-11-09 23:57:56,901][__main__][INFO] - agents played in iteration 385 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:57:57,711][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:57:57,734][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:57:57,759][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:57:57,782][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:57:57,783][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:57:57,784][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:57:58,583][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:57:59,043][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:57:59,563][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:58:00,068][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:58:00,577][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:58:01,079][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:58:01,579][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:58:02,083][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:58:02,585][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:58:03,092][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:58:03,595][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:58:04,097][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:58:04,604][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:58:05,106][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:58:05,609][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:58:06,111][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:58:06,610][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:58:07,107][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:58:07,611][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:58:08,112][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:58:08,613][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:58:09,112][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:58:09,611][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:58:10,114][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:58:10,615][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:58:11,121][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:58:11,623][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:58:12,127][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:58:12,631][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:58:13,134][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:58:13,636][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:58:14,136][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:58:14,637][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:58:15,135][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:58:15,634][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:58:16,131][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:58:16,631][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:58:17,128][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:58:17,630][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:58:18,126][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:58:18,626][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:58:19,121][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:58:19,616][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:58:20,114][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:58:20,611][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:58:21,107][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:58:21,605][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:58:22,102][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:58:22,599][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:58:23,094][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:58:23,591][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:58:24,092][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:58:24,590][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:58:25,086][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:58:25,592][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:58:26,087][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:58:26,585][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:58:27,084][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:58:27,583][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:58:28,089][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:58:28,587][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:58:29,085][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:58:29,592][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:58:30,090][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:58:30,592][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10210 tokens. -[2025-11-09 23:58:31,328][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.26%, Current % of VRAM taken: 55.75%, Block Peak % of device VRAM: 62.22%, ΔTime: 00:00:32 -[2025-11-09 23:58:32,081][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:58:32,083][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:58:32,084][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:58:33,082][__main__][INFO] - Iteration 386 took 1m 7s (46.70% Gen, 51.83% Train). Generation: 31s, Training: 35s. Estimated remaining time: 50h 34m 45s. Estimated total time: 56h 34m 21s. Time estimates for 10 more iterations: 11m 18s, 100 more iterations: 1h 53m 8s, 500 more iterations: 9h 25m 43s. -[2025-11-09 23:58:33,084][__main__][INFO] - Starting iteration 386. -[2025-11-09 23:58:33,549][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 38 and human policies 1. -[2025-11-09 23:58:33,550][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-09 23:58:55,036][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-09 23:59:00,699][__main__][INFO] - Number of regex retries in iteration 386: 1 -[2025-11-09 23:59:00,699][__main__][INFO] - agents played in iteration 386 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-09 23:59:01,530][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:59:01,559][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:59:01,584][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:59:01,608][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-09 23:59:01,609][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-09 23:59:01,610][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-09 23:59:02,438][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-09 23:59:03,011][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-09 23:59:03,519][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-09 23:59:04,028][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-09 23:59:04,534][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-09 23:59:05,041][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-09 23:59:05,545][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-09 23:59:06,047][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-09 23:59:06,551][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-09 23:59:07,060][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-09 23:59:07,564][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-09 23:59:08,068][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-09 23:59:08,567][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-09 23:59:09,066][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-09 23:59:09,565][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-09 23:59:10,067][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-09 23:59:10,568][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-09 23:59:11,067][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-09 23:59:11,566][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-09 23:59:12,065][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-09 23:59:12,564][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-09 23:59:13,064][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-09 23:59:13,565][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-09 23:59:14,068][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-09 23:59:14,566][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-09 23:59:15,068][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-09 23:59:15,577][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-09 23:59:16,080][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-09 23:59:16,590][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-09 23:59:17,094][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-09 23:59:17,598][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-09 23:59:18,102][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-09 23:59:18,603][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-09 23:59:19,107][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-09 23:59:19,614][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-09 23:59:20,120][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-09 23:59:20,627][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-09 23:59:21,133][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-09 23:59:21,635][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-09 23:59:22,135][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-09 23:59:22,635][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-09 23:59:23,138][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-09 23:59:23,634][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-09 23:59:24,129][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-09 23:59:24,633][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-09 23:59:25,129][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-09 23:59:25,626][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-09 23:59:26,126][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-09 23:59:26,623][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-09 23:59:27,125][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-09 23:59:27,622][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-09 23:59:28,122][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-09 23:59:28,620][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-09 23:59:29,119][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-09 23:59:29,618][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-09 23:59:30,117][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-09 23:59:30,624][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-09 23:59:31,131][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-09 23:59:31,628][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-09 23:59:32,128][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-09 23:59:32,624][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-09 23:59:33,121][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-09 23:59:33,620][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-09 23:59:34,118][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-09 23:59:34,615][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10312 tokens. -[2025-11-09 23:59:35,335][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 55.56%, Block Peak % of device VRAM: 62.34%, ΔTime: 00:00:32 -[2025-11-09 23:59:35,961][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-09 23:59:35,963][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-09 23:59:35,965][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-09 23:59:36,970][__main__][INFO] - Iteration 387 took 1m 3s (42.81% Gen, 55.60% Train). Generation: 27s, Training: 35s. Estimated remaining time: 46h 50m 24s. Estimated total time: 52h 51m 4s. Time estimates for 10 more iterations: 10m 34s, 100 more iterations: 1h 45m 42s, 500 more iterations: 8h 48m 30s. -[2025-11-09 23:59:36,973][__main__][INFO] - Starting iteration 387. -[2025-11-09 23:59:37,434][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 38 and human policies 1. -[2025-11-09 23:59:37,435][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:00:08,768][__main__][INFO] - Number of regex retries in iteration 387: 0 -[2025-11-10 00:00:08,770][__main__][INFO] - agents played in iteration 387 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:00:09,654][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:00:09,689][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:00:09,720][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:00:09,747][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:00:09,749][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:00:09,750][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:00:10,522][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:00:11,066][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:00:11,588][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:00:12,101][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:00:12,600][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:00:13,113][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:00:13,614][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:00:14,115][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:00:14,615][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:00:15,114][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:00:15,620][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:00:16,122][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:00:16,624][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:00:17,129][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:00:17,629][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:00:18,128][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:00:18,631][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:00:19,134][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:00:19,641][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:00:20,144][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:00:20,646][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:00:21,150][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:00:21,654][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:00:22,156][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:00:22,662][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:00:23,165][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:00:23,667][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:00:24,164][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:00:24,662][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:00:25,171][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:00:25,670][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:00:26,166][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:00:26,667][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:00:27,167][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:00:27,673][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:00:28,173][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:00:28,670][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:00:29,167][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:00:29,664][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:00:30,161][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:00:30,660][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:00:31,161][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:00:31,664][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:00:32,163][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:00:32,662][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:00:33,162][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:00:33,662][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:00:34,163][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:00:34,664][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:00:35,161][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:00:35,662][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:00:36,158][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:00:36,656][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:00:37,157][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:00:37,653][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:00:38,150][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:00:38,650][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:00:39,145][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:00:39,642][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:00:40,144][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:00:40,640][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:00:41,141][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:00:41,638][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:00:42,138][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:00:42,635][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10231 tokens. -[2025-11-10 00:00:43,395][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 55.57%, Block Peak % of device VRAM: 62.22%, ΔTime: 00:00:32 -[2025-11-10 00:00:44,022][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:00:44,023][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:00:44,025][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:00:44,847][__main__][INFO] - Iteration 388 took 1m 7s (46.48% Gen, 52.30% Train). Generation: 31s, Training: 35s. Estimated remaining time: 50h 8m 52s. Estimated total time: 56h 10m 40s. Time estimates for 10 more iterations: 11m 14s, 100 more iterations: 1h 52m 21s, 500 more iterations: 9h 21m 46s. -[2025-11-10 00:00:44,849][__main__][INFO] - Starting iteration 388. -[2025-11-10 00:00:45,335][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 38 and human policies 1. -[2025-11-10 00:00:45,336][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:01:15,618][__main__][INFO] - Number of regex retries in iteration 388: 0 -[2025-11-10 00:01:15,619][__main__][INFO] - agents played in iteration 388 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:01:16,439][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:01:16,471][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:01:16,497][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:01:16,521][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:01:16,522][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:01:16,523][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:01:17,303][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:01:17,766][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:01:18,271][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:01:18,778][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:01:19,276][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:01:19,774][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:01:20,276][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:01:20,783][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:01:21,283][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:01:21,789][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:01:22,288][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:01:22,788][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:01:23,290][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:01:23,788][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:01:24,297][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:01:24,792][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:01:25,291][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:01:25,800][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:01:26,299][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:01:26,799][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:01:27,299][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:01:27,794][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:01:28,289][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:01:28,784][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:01:29,283][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:01:29,782][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:01:30,286][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:01:30,808][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:01:31,309][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:01:31,812][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:01:32,317][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:01:32,815][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:01:33,344][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:01:33,844][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:01:34,344][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:01:34,843][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:01:35,339][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:01:35,841][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:01:36,342][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:01:36,839][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:01:37,343][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:01:37,839][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:01:38,336][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:01:38,837][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:01:39,335][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:01:39,832][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:01:40,329][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:01:40,825][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:01:41,334][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:01:41,830][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:01:42,325][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:01:42,823][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:01:43,322][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:01:43,819][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:01:44,315][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:01:44,812][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:01:45,311][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:01:45,811][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:01:46,311][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:01:46,810][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:01:47,310][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:01:47,824][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:01:48,324][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:01:48,824][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:01:49,327][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10310 tokens. -[2025-11-10 00:01:50,031][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.25%, Current % of VRAM taken: 55.75%, Block Peak % of device VRAM: 62.23%, ΔTime: 00:00:32 -[2025-11-10 00:01:50,788][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:01:50,790][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:01:50,791][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:01:51,699][__main__][INFO] - Iteration 389 took 1m 6s (45.63% Gen, 53.00% Train). Generation: 30s, Training: 35s. Estimated remaining time: 49h 15m 20s. Estimated total time: 55h 18m 14s. Time estimates for 10 more iterations: 11m 3s, 100 more iterations: 1h 50m 36s, 500 more iterations: 9h 13m 2s. -[2025-11-10 00:01:51,701][__main__][INFO] - Starting iteration 389. -[2025-11-10 00:01:52,169][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 38 and human policies 1. -[2025-11-10 00:01:52,170][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:02:05,830][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 00:02:16,851][__main__][INFO] - Number of regex retries in iteration 389: 1 -[2025-11-10 00:02:16,851][__main__][INFO] - agents played in iteration 389 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:02:17,734][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:02:17,795][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:02:17,832][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:02:17,865][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:02:17,865][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:02:17,867][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:02:18,694][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:02:19,244][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:02:19,756][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:02:20,261][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:02:20,764][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:02:21,269][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:02:21,771][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:02:22,269][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:02:22,771][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:02:23,275][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:02:23,779][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:02:24,287][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:02:24,792][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:02:25,296][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:02:25,805][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:02:26,303][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:02:26,807][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:02:27,317][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:02:27,821][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:02:28,331][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:02:28,831][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:02:29,329][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:02:29,832][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:02:30,334][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:02:30,831][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:02:31,331][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:02:31,828][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:02:32,344][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:02:32,844][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:02:33,340][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:02:33,843][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:02:34,340][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:02:34,845][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:02:35,339][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:02:35,839][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:02:36,346][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:02:36,843][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:02:37,344][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:02:37,842][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:02:38,340][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:02:38,848][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:02:39,344][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:02:39,846][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:02:40,344][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:02:40,842][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:02:41,337][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:02:41,836][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:02:42,336][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:02:42,842][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:02:43,336][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:02:43,831][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:02:44,330][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:02:44,825][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:02:45,340][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:02:45,836][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:02:46,332][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:02:46,828][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:02:47,322][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:02:47,831][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:02:48,326][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:02:48,822][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:02:49,320][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:02:49,814][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:02:50,313][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:02:50,809][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10183 tokens. -[2025-11-10 00:02:51,555][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.02%, Current % of VRAM taken: 55.52%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 -[2025-11-10 00:02:52,196][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:02:52,197][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:02:52,199][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:02:53,020][__main__][INFO] - Iteration 390 took 1m 0s (40.56% Gen, 58.09% Train). Generation: 24s, Training: 35s. Estimated remaining time: 44h 38m 37s. Estimated total time: 50h 42m 33s. Time estimates for 10 more iterations: 10m 8s, 100 more iterations: 1h 41m 25s, 500 more iterations: 8h 27m 5s. -[2025-11-10 00:02:53,022][__main__][INFO] - Starting iteration 390. -[2025-11-10 00:02:53,486][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 38 and human policies 1. -[2025-11-10 00:02:53,487][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:03:20,449][__main__][INFO] - Number of regex retries in iteration 390: 0 -[2025-11-10 00:03:20,451][__main__][INFO] - agents played in iteration 390 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:03:21,266][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:03:21,293][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:03:21,318][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:03:21,341][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:03:21,342][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:03:21,343][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:03:22,105][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:03:22,564][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:03:23,065][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:03:23,564][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:03:24,069][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:03:24,571][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:03:25,069][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:03:25,568][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:03:26,071][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:03:26,568][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:03:27,066][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:03:27,568][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:03:28,065][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:03:28,572][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:03:29,094][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:03:29,598][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:03:30,095][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:03:30,594][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:03:31,100][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:03:31,598][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:03:32,102][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:03:32,616][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:03:33,118][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:03:33,619][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:03:34,121][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:03:34,616][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:03:35,115][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:03:35,614][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:03:36,113][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:03:36,621][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:03:37,121][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:03:37,625][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:03:38,123][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:03:38,623][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:03:39,123][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:03:39,623][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:03:40,122][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:03:40,618][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:03:41,118][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:03:41,631][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:03:42,130][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:03:42,631][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:03:43,131][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:03:43,627][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:03:44,136][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:03:44,633][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:03:45,136][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:03:45,650][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:03:46,146][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:03:46,649][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:03:47,147][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:03:47,643][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:03:48,160][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:03:48,660][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:03:49,158][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:03:49,659][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:03:50,162][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:03:50,667][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:03:51,167][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:03:51,668][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:03:52,171][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:03:52,673][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:03:53,178][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:03:53,682][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:03:54,187][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10290 tokens. -[2025-11-10 00:03:55,022][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 55.65%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 -[2025-11-10 00:03:55,677][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:03:55,679][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:03:55,681][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:03:57,357][__main__][INFO] - Iteration 391 took 1m 3s (42.21% Gen, 55.16% Train). Generation: 26s, Training: 35s. Estimated remaining time: 47h 8m 35s. Estimated total time: 53h 13m 35s. Time estimates for 10 more iterations: 10m 38s, 100 more iterations: 1h 46m 27s, 500 more iterations: 8h 52m 15s. -[2025-11-10 00:03:57,360][__main__][INFO] - Starting iteration 391. -[2025-11-10 00:03:57,848][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 39 and human policies 1. -[2025-11-10 00:03:57,849][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:04:22,218][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 00:04:23,688][__main__][INFO] - Number of regex retries in iteration 391: 1 -[2025-11-10 00:04:23,689][__main__][INFO] - agents played in iteration 391 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:04:24,505][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:04:24,530][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:04:24,559][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:04:24,594][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:04:24,594][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:04:24,595][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:04:25,382][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:04:25,837][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:04:26,340][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:04:26,837][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:04:27,341][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:04:27,838][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:04:28,334][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:04:28,845][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:04:29,344][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:04:29,841][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:04:30,342][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:04:30,839][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:04:31,349][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:04:31,850][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:04:32,349][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:04:32,846][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:04:33,346][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:04:33,844][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:04:34,349][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:04:34,845][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:04:35,362][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:04:35,860][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:04:36,362][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:04:36,864][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:04:37,364][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:04:37,859][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:04:38,352][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:04:38,847][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:04:39,360][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:04:39,861][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:04:40,356][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:04:40,854][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:04:41,348][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:04:41,842][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:04:42,337][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:04:42,839][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:04:43,336][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:04:43,831][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:04:44,327][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:04:44,835][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:04:45,333][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:04:45,836][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:04:46,336][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:04:46,833][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:04:47,348][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:04:47,845][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:04:48,343][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:04:48,848][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:04:49,345][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:04:49,844][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:04:50,341][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:04:50,841][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:04:51,356][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:04:51,852][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:04:52,349][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:04:52,850][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:04:53,349][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:04:53,869][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:04:54,370][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:04:54,872][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:04:55,375][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:04:55,875][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:04:56,375][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:04:56,873][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:04:57,373][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10112 tokens. -[2025-11-10 00:04:58,213][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.09%, Current % of VRAM taken: 55.59%, Block Peak % of device VRAM: 62.20%, ΔTime: 00:00:32 -[2025-11-10 00:04:58,986][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:04:58,988][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:04:58,990][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:04:59,919][__main__][INFO] - Iteration 392 took 1m 2s (41.63% Gen, 56.87% Train). Generation: 25s, Training: 35s. Estimated remaining time: 45h 37m 28s. Estimated total time: 51h 43m 31s. Time estimates for 10 more iterations: 10m 20s, 100 more iterations: 1h 43m 27s, 500 more iterations: 8h 37m 15s. -[2025-11-10 00:04:59,921][__main__][INFO] - Starting iteration 392. -[2025-11-10 00:05:00,404][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 39 and human policies 1. -[2025-11-10 00:05:00,405][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:05:10,005][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 00:05:21,687][__main__][INFO] - Number of regex retries in iteration 392: 1 -[2025-11-10 00:05:21,688][__main__][INFO] - agents played in iteration 392 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:05:22,503][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:05:22,531][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:05:22,558][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:05:22,582][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:05:22,582][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:05:22,583][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:05:23,362][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:05:23,821][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:05:24,325][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:05:24,825][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:05:25,335][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:05:25,837][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:05:26,336][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:05:26,845][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:05:27,344][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:05:27,860][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:05:28,364][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:05:28,864][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:05:29,377][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:05:29,876][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:05:30,379][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:05:30,883][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:05:31,391][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:05:31,926][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:05:32,426][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:05:32,926][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:05:33,437][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:05:33,936][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:05:34,447][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:05:34,945][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:05:35,444][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:05:35,945][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:05:36,442][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:05:36,942][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:05:37,438][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:05:37,934][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:05:38,452][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:05:38,947][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:05:39,447][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:05:39,951][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:05:40,450][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:05:40,960][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:05:41,461][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:05:41,958][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:05:42,455][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:05:42,954][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:05:43,453][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:05:43,950][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:05:44,447][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:05:44,963][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:05:45,459][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:05:45,957][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:05:46,459][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:05:46,957][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:05:47,467][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:05:47,963][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:05:48,463][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:05:48,966][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:05:49,464][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:05:49,962][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:05:50,460][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:05:50,961][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:05:51,469][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:05:51,972][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:05:52,471][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:05:52,972][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:05:53,472][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:05:53,976][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:05:54,479][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:05:54,981][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:05:55,486][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10222 tokens. -[2025-11-10 00:05:56,297][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:32 -[2025-11-10 00:05:56,959][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:05:56,961][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:05:56,963][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:05:57,870][__main__][INFO] - Iteration 393 took 57s (37.04% Gen, 61.38% Train). Generation: 21s, Training: 35s. Estimated remaining time: 41h 46m 16s. Estimated total time: 47h 53m 17s. Time estimates for 10 more iterations: 9m 34s, 100 more iterations: 1h 35m 46s, 500 more iterations: 7h 58m 52s. -[2025-11-10 00:05:57,872][__main__][INFO] - Starting iteration 393. -[2025-11-10 00:05:58,344][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 39 and human policies 1. -[2025-11-10 00:05:58,345][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:06:23,729][__main__][INFO] - Number of regex retries in iteration 393: 0 -[2025-11-10 00:06:23,730][__main__][INFO] - agents played in iteration 393 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:06:24,569][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:06:24,595][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:06:24,620][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:06:24,643][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:06:24,643][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:06:24,644][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:06:25,474][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:06:25,938][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:06:26,448][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:06:26,948][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:06:27,450][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:06:27,951][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:06:28,455][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:06:28,955][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:06:29,454][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:06:29,969][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:06:30,471][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:06:30,970][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:06:31,488][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:06:31,993][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:06:32,496][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:06:32,996][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:06:33,498][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:06:34,003][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:06:34,506][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:06:35,011][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:06:35,517][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:06:36,022][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:06:36,524][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:06:37,026][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:06:37,529][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:06:38,027][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:06:38,528][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:06:39,026][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:06:39,529][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:06:40,027][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:06:40,526][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:06:41,025][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:06:41,524][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:06:42,021][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:06:42,523][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:06:43,020][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:06:43,518][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:06:44,015][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:06:44,518][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:06:45,019][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:06:45,516][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:06:46,016][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:06:46,516][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:06:47,014][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:06:47,515][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:06:48,013][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:06:48,512][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:06:49,015][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:06:49,517][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:06:50,018][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:06:50,517][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:06:51,017][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:06:51,518][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:06:52,023][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:06:52,527][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:06:53,031][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:06:53,536][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:06:54,031][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:06:54,530][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:06:55,028][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:06:55,527][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:06:56,023][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:06:56,520][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:06:57,016][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:06:57,516][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10220 tokens. -[2025-11-10 00:06:58,239][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 55.56%, Block Peak % of device VRAM: 62.20%, ΔTime: 00:00:32 -[2025-11-10 00:06:58,984][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:06:58,986][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:06:58,987][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:06:59,915][__main__][INFO] - Iteration 394 took 1m 1s (41.23% Gen, 57.26% Train). Generation: 25s, Training: 35s. Estimated remaining time: 45h 10m 31s. Estimated total time: 51h 18m 34s. Time estimates for 10 more iterations: 10m 15s, 100 more iterations: 1h 42m 37s, 500 more iterations: 8h 33m 5s. -[2025-11-10 00:06:59,918][__main__][INFO] - Starting iteration 394. -[2025-11-10 00:07:00,382][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 39 and human policies 1. -[2025-11-10 00:07:00,382][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:07:26,461][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 00:07:31,261][__main__][INFO] - Number of regex retries in iteration 394: 1 -[2025-11-10 00:07:31,261][__main__][INFO] - agents played in iteration 394 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:07:32,078][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:07:32,103][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:07:32,129][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:07:32,152][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:07:32,153][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:07:32,153][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:07:32,949][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:07:33,409][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:07:33,915][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:07:34,416][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:07:34,918][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:07:35,424][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:07:35,925][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:07:36,431][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:07:36,933][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:07:37,432][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:07:37,933][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:07:38,438][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:07:38,940][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:07:39,446][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:07:39,951][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:07:40,451][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:07:40,954][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:07:41,457][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:07:41,969][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:07:42,467][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:07:42,966][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:07:43,480][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:07:43,979][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:07:44,493][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:07:44,994][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:07:45,493][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:07:46,000][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:07:46,504][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:07:47,009][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:07:47,510][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:07:48,013][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:07:48,514][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:07:49,017][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:07:49,519][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:07:50,028][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:07:50,532][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:07:51,033][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:07:51,538][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:07:52,045][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:07:52,554][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:07:53,055][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:07:53,551][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:07:54,046][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:07:54,545][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:07:55,042][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:07:55,538][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:07:56,034][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:07:56,531][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:07:57,025][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:07:57,523][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:07:58,020][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:07:58,520][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:07:59,020][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:07:59,520][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:08:00,015][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:08:00,513][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:08:01,014][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:08:01,515][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:08:02,023][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:08:02,527][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:08:03,031][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:08:03,533][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:08:04,036][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:08:04,543][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:08:05,049][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10315 tokens. -[2025-11-10 00:08:05,867][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 55.70%, Block Peak % of device VRAM: 62.41%, ΔTime: 00:00:32 -[2025-11-10 00:08:06,502][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:08:06,503][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:08:06,505][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:08:07,327][__main__][INFO] - Iteration 395 took 1m 6s (46.12% Gen, 52.64% Train). Generation: 30s, Training: 35s. Estimated remaining time: 49h 38m 8s. Estimated total time: 55h 47m 18s. Time estimates for 10 more iterations: 11m 9s, 100 more iterations: 1h 51m 34s, 500 more iterations: 9h 17m 53s. -[2025-11-10 00:08:07,330][__main__][INFO] - Starting iteration 395. -[2025-11-10 00:08:07,807][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 39 and human policies 1. -[2025-11-10 00:08:07,808][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:08:28,821][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 00:08:29,035][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 00:08:33,024][__main__][INFO] - Number of regex retries in iteration 395: 2 -[2025-11-10 00:08:33,025][__main__][INFO] - agents played in iteration 395 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:08:33,832][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:08:33,861][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:08:33,888][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:08:33,912][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:08:33,912][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:08:33,914][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:08:34,706][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:08:35,162][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:08:35,667][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:08:36,170][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:08:36,672][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:08:37,177][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:08:37,675][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:08:38,173][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:08:38,686][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:08:39,185][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:08:39,695][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:08:40,193][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:08:40,691][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:08:41,190][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:08:41,691][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:08:42,190][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:08:42,694][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:08:43,194][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:08:43,712][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:08:44,214][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:08:44,720][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:08:45,218][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:08:45,722][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:08:46,231][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:08:46,737][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:08:47,239][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:08:47,745][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:08:48,246][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:08:48,747][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:08:49,250][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:08:49,753][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:08:50,260][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:08:50,764][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:08:51,266][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:08:51,766][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:08:52,269][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:08:52,767][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:08:53,267][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:08:53,769][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:08:54,272][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:08:54,776][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:08:55,276][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:08:55,788][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:08:56,287][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:08:56,795][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:08:57,293][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:08:57,796][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:08:58,295][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:08:58,796][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:08:59,300][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:08:59,798][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:09:00,294][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:09:00,808][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:09:01,308][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:09:01,809][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:09:02,312][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:09:02,812][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:09:03,307][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:09:03,807][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:09:04,307][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:09:04,818][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:09:05,314][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:09:05,811][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:09:06,308][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:09:06,805][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10192 tokens. -[2025-11-10 00:09:07,559][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 55.57%, Block Peak % of device VRAM: 62.18%, ΔTime: 00:00:32 -[2025-11-10 00:09:08,196][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:09:08,198][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:09:08,200][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:09:09,010][__main__][INFO] - Iteration 396 took 1m 1s (41.20% Gen, 57.47% Train). Generation: 25s, Training: 35s. Estimated remaining time: 44h 49m 57s. Estimated total time: 51h 0m 9s. Time estimates for 10 more iterations: 10m 12s, 100 more iterations: 1h 42m 0s, 500 more iterations: 8h 30m 1s. -[2025-11-10 00:09:09,012][__main__][INFO] - Starting iteration 396. -[2025-11-10 00:09:09,480][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 39 and human policies 1. -[2025-11-10 00:09:09,480][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:09:40,515][__main__][INFO] - Number of regex retries in iteration 396: 0 -[2025-11-10 00:09:40,516][__main__][INFO] - agents played in iteration 396 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:09:41,293][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:09:41,322][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:09:41,348][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:09:41,372][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:09:41,372][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:09:41,374][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:09:42,073][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:09:42,527][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:09:43,032][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:09:43,534][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:09:44,034][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:09:44,532][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:09:45,029][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:09:45,535][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:09:46,032][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:09:46,530][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:09:47,033][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:09:47,532][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:09:48,033][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:09:48,555][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:09:49,055][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:09:49,557][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:09:50,058][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:09:50,562][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:09:51,065][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:09:51,566][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:09:52,072][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:09:52,574][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:09:53,074][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:09:53,582][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:09:54,085][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:09:54,587][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:09:55,085][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:09:55,587][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:09:56,090][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:09:56,592][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:09:57,091][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:09:57,591][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:09:58,091][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:09:58,599][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:09:59,098][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:09:59,599][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:10:00,113][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:10:00,613][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:10:01,127][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:10:01,629][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:10:02,131][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:10:02,644][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:10:03,144][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:10:03,648][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:10:04,150][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:10:04,652][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:10:05,157][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:10:05,658][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:10:06,155][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:10:06,657][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:10:07,161][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:10:07,665][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:10:08,162][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:10:08,665][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:10:09,169][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:10:09,667][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:10:10,168][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:10:10,665][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:10:11,167][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:10:11,668][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:10:12,166][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:10:12,667][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:10:13,167][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:10:13,665][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:10:14,166][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10233 tokens. -[2025-11-10 00:10:14,973][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 55.62%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:32 -[2025-11-10 00:10:15,728][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:10:15,729][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:10:15,731][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:10:16,654][__main__][INFO] - Iteration 397 took 1m 7s (46.20% Gen, 52.42% Train). Generation: 31s, Training: 35s. Estimated remaining time: 49h 47m 24s. Estimated total time: 55h 58m 44s. Time estimates for 10 more iterations: 11m 11s, 100 more iterations: 1h 51m 57s, 500 more iterations: 9h 19m 47s. -[2025-11-10 00:10:16,657][__main__][INFO] - Starting iteration 397. -[2025-11-10 00:10:17,124][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 39 and human policies 1. -[2025-11-10 00:10:17,125][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:10:45,930][__main__][INFO] - Number of regex retries in iteration 397: 0 -[2025-11-10 00:10:45,932][__main__][INFO] - agents played in iteration 397 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:10:46,787][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:10:46,815][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:10:46,842][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:10:46,865][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:10:46,866][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:10:46,867][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:10:47,594][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:10:48,052][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:10:48,558][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:10:49,056][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:10:49,558][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:10:50,058][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:10:50,559][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:10:51,056][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:10:51,559][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:10:52,058][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:10:52,554][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:10:53,050][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:10:53,551][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:10:54,055][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:10:54,554][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:10:55,050][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:10:55,548][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:10:56,053][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:10:56,552][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:10:57,054][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:10:57,559][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:10:58,059][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:10:58,557][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:10:59,060][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:10:59,567][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:11:00,074][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:11:00,577][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:11:01,078][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:11:01,582][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:11:02,084][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:11:02,588][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:11:03,090][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:11:03,591][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:11:04,097][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:11:04,597][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:11:05,098][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:11:05,617][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:11:06,122][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:11:06,644][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:11:07,144][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:11:07,647][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:11:08,155][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:11:08,655][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:11:09,159][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:11:09,666][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:11:10,166][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:11:10,668][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:11:11,171][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:11:11,670][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:11:12,173][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:11:12,676][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:11:13,175][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:11:13,677][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:11:14,179][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:11:14,681][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:11:15,185][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:11:15,695][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:11:16,210][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:11:16,710][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:11:17,216][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:11:17,716][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:11:18,216][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:11:18,734][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:11:19,234][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:11:19,734][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10211 tokens. -[2025-11-10 00:11:20,562][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 55.58%, Block Peak % of device VRAM: 62.27%, ΔTime: 00:00:32 -[2025-11-10 00:11:21,235][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:11:21,237][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:11:21,239][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:11:22,052][__main__][INFO] - Iteration 398 took 1m 4s (44.37% Gen, 54.38% Train). Generation: 28s, Training: 35s. Estimated remaining time: 47h 54m 0s. Estimated total time: 54h 6m 25s. Time estimates for 10 more iterations: 10m 49s, 100 more iterations: 1h 48m 12s, 500 more iterations: 9h 1m 4s. -[2025-11-10 00:11:22,054][__main__][INFO] - Starting iteration 398. -[2025-11-10 00:11:22,534][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 39 and human policies 1. -[2025-11-10 00:11:22,535][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:11:41,533][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 00:11:47,546][__main__][INFO] - Number of regex retries in iteration 398: 1 -[2025-11-10 00:11:47,546][__main__][INFO] - agents played in iteration 398 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:11:48,321][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:11:48,350][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:11:48,376][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:11:48,399][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:11:48,400][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:11:48,402][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:11:49,111][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:11:49,568][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:11:50,072][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:11:50,571][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:11:51,076][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:11:51,581][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:11:52,081][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:11:52,577][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:11:53,076][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:11:53,572][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:11:54,069][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:11:54,565][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:11:55,066][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:11:55,567][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:11:56,062][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:11:56,558][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:11:57,057][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:11:57,558][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:11:58,055][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:11:58,552][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:11:59,049][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:11:59,546][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:12:00,045][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:12:00,543][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:12:01,041][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:12:01,537][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:12:02,031][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:12:02,527][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:12:03,023][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:12:03,526][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:12:04,024][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:12:04,523][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:12:05,021][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:12:05,521][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:12:06,044][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:12:06,549][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:12:07,049][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:12:07,551][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:12:08,080][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:12:08,579][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:12:09,085][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:12:09,590][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:12:10,090][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:12:10,596][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:12:11,102][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:12:11,616][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:12:12,117][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:12:12,634][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:12:13,135][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:12:13,637][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:12:14,137][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:12:14,637][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:12:15,148][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:12:15,650][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:12:16,151][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:12:16,664][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:12:17,166][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:12:17,669][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:12:18,171][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:12:18,672][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:12:19,178][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:12:19,680][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:12:20,182][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:12:20,685][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:12:21,190][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10244 tokens. -[2025-11-10 00:12:22,047][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.37%, ΔTime: 00:00:32 -[2025-11-10 00:12:22,704][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:12:22,706][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:12:22,709][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:12:23,534][__main__][INFO] - Iteration 399 took 1m 1s (41.00% Gen, 57.64% Train). Generation: 25s, Training: 35s. Estimated remaining time: 44h 36m 33s. Estimated total time: 50h 50m 0s. Time estimates for 10 more iterations: 10m 10s, 100 more iterations: 1h 41m 40s, 500 more iterations: 8h 28m 20s. -[2025-11-10 00:12:23,536][__main__][INFO] - Starting iteration 399. -[2025-11-10 00:12:24,003][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 39 and human policies 1. -[2025-11-10 00:12:24,004][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:12:41,864][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 00:12:47,812][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 00:12:54,540][__main__][INFO] - Number of regex retries in iteration 399: 2 -[2025-11-10 00:12:54,541][__main__][INFO] - agents played in iteration 399 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:12:55,319][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:12:55,345][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:12:55,370][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:12:55,392][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:12:55,392][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:12:55,393][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:12:56,093][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:12:56,545][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:12:57,047][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:12:57,548][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:12:58,045][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:12:58,545][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:12:59,044][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:12:59,548][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:13:00,066][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:13:00,562][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:13:01,065][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:13:01,563][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:13:02,057][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:13:02,552][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:13:03,051][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:13:03,556][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:13:04,069][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:13:04,567][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:13:05,065][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:13:05,567][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:13:06,067][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:13:06,577][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:13:07,080][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:13:07,581][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:13:08,080][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:13:08,582][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:13:09,088][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:13:09,614][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:13:10,116][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:13:10,628][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:13:11,129][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:13:11,634][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:13:12,136][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:13:12,636][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:13:13,158][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:13:13,656][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:13:14,155][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:13:14,656][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:13:15,158][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:13:15,665][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:13:16,166][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:13:16,664][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:13:17,163][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:13:17,662][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:13:18,165][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:13:18,666][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:13:19,171][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:13:19,681][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:13:20,187][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:13:20,690][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:13:21,193][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:13:21,702][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:13:22,213][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:13:22,719][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:13:23,223][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:13:23,729][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:13:24,231][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:13:24,731][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:13:25,232][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:13:25,734][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:13:26,740][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:13:27,284][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:13:27,797][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:13:28,319][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:13:28,821][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10260 tokens. -[2025-11-10 00:13:29,697][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 55.59%, Block Peak % of device VRAM: 62.38%, ΔTime: 00:00:33 -[2025-11-10 00:13:30,331][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:13:30,333][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:13:30,334][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:13:31,300][__main__][INFO] - Iteration 400 took 1m 7s (45.37% Gen, 53.19% Train). Generation: 30s, Training: 35s. Estimated remaining time: 49h 50m 19s. Estimated total time: 56h 4m 53s. Time estimates for 10 more iterations: 11m 12s, 100 more iterations: 1h 52m 9s, 500 more iterations: 9h 20m 48s. -[2025-11-10 00:13:31,302][__main__][INFO] - Starting iteration 400. -[2025-11-10 00:13:31,774][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 39 and human policies 1. -[2025-11-10 00:13:31,774][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:13:53,812][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 00:14:01,066][__main__][INFO] - Number of regex retries in iteration 400: 1 -[2025-11-10 00:14:01,067][__main__][INFO] - agents played in iteration 400 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:14:01,850][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:14:01,874][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:14:01,899][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:14:01,921][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:14:01,922][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:14:01,922][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:14:02,624][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:14:03,083][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:14:03,594][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:14:04,092][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:14:04,589][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:14:05,087][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:14:05,590][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:14:06,088][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:14:06,585][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:14:07,088][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:14:07,586][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:14:08,086][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:14:08,585][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:14:09,082][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:14:09,594][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:14:10,091][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:14:10,591][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:14:11,094][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:14:11,589][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:14:12,092][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:14:12,588][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:14:13,085][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:14:13,582][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:14:14,082][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:14:14,578][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:14:15,074][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:14:15,571][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:14:16,087][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:14:16,585][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:14:17,081][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:14:17,579][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:14:18,073][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:14:18,569][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:14:19,066][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:14:19,561][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:14:20,081][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:14:20,577][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:14:21,074][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:14:21,578][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:14:22,078][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:14:22,585][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:14:23,086][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:14:23,587][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:14:24,097][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:14:24,602][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:14:25,106][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:14:25,609][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:14:26,114][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:14:26,620][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:14:27,120][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:14:27,621][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:14:28,124][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:14:28,626][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:14:29,134][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:14:29,639][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:14:30,140][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:14:30,654][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:14:31,160][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:14:31,678][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:14:32,186][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:14:32,690][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:14:33,191][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:14:33,693][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:14:34,204][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:14:34,706][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10137 tokens. -[2025-11-10 00:14:35,588][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 55.55%, Block Peak % of device VRAM: 62.15%, ΔTime: 00:00:32 -[2025-11-10 00:14:36,335][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:14:36,337][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:14:36,338][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:14:38,166][__main__][INFO] - Iteration 401 took 1m 6s (44.12% Gen, 53.13% Train). Generation: 29s, Training: 35s. Estimated remaining time: 49h 3m 56s. Estimated total time: 55h 19m 37s. Time estimates for 10 more iterations: 11m 3s, 100 more iterations: 1h 50m 39s, 500 more iterations: 9h 13m 16s. -[2025-11-10 00:14:38,168][__main__][INFO] - Starting iteration 401. -[2025-11-10 00:14:38,634][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 40 and human policies 1. -[2025-11-10 00:14:38,635][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:14:56,545][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 00:15:07,356][__main__][INFO] - Number of regex retries in iteration 401: 1 -[2025-11-10 00:15:07,357][__main__][INFO] - agents played in iteration 401 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:15:08,137][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:15:08,162][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:15:08,187][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:15:08,210][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:15:08,210][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:15:08,211][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:15:08,910][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:15:09,369][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:15:09,873][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:15:10,373][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:15:10,873][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:15:11,370][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:15:11,891][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:15:12,394][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:15:12,890][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:15:13,387][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:15:13,885][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:15:14,393][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:15:14,890][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:15:15,390][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:15:15,902][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:15:16,402][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:15:16,906][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:15:17,403][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:15:17,900][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:15:18,403][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:15:18,900][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:15:19,400][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:15:19,896][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:15:20,391][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:15:20,887][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:15:21,386][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:15:21,894][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:15:22,404][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:15:22,908][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:15:23,410][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:15:23,914][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:15:24,417][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:15:24,925][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:15:25,429][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:15:25,932][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:15:26,439][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:15:26,938][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:15:27,442][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:15:27,948][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:15:28,449][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:15:28,964][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:15:29,468][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:15:29,978][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:15:30,479][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:15:30,983][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:15:31,486][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:15:31,987][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:15:32,492][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:15:32,996][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:15:33,496][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:15:34,005][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:15:34,507][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:15:35,008][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:15:35,510][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:15:36,014][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:15:36,516][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:15:37,023][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:15:37,523][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:15:38,028][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:15:38,529][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:15:39,031][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:15:39,531][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:15:40,032][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:15:40,540][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:15:41,045][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10116 tokens. -[2025-11-10 00:15:41,882][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 55.69%, Block Peak % of device VRAM: 62.24%, ΔTime: 00:00:32 -[2025-11-10 00:15:42,516][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:15:42,518][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:15:42,520][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:15:43,330][__main__][INFO] - Iteration 402 took 1m 4s (44.39% Gen, 54.35% Train). Generation: 28s, Training: 35s. Estimated remaining time: 47h 38m 3s. Estimated total time: 53h 54m 49s. Time estimates for 10 more iterations: 10m 46s, 100 more iterations: 1h 47m 49s, 500 more iterations: 8h 59m 8s. -[2025-11-10 00:15:43,333][__main__][INFO] - Starting iteration 402. -[2025-11-10 00:15:43,799][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 40 and human policies 1. -[2025-11-10 00:15:43,800][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:16:06,756][__main__][INFO] - Number of regex retries in iteration 402: 0 -[2025-11-10 00:16:06,757][__main__][INFO] - agents played in iteration 402 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:16:07,553][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:16:07,577][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:16:07,600][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:16:07,622][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:16:07,623][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:16:07,623][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:16:08,385][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:16:08,842][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:16:09,345][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:16:09,847][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:16:10,344][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:16:10,844][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:16:11,346][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:16:11,846][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:16:12,345][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:16:12,842][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:16:13,339][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:16:13,833][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:16:14,327][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:16:14,827][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:16:15,324][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:16:15,821][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:16:16,316][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:16:16,812][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:16:17,313][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:16:17,808][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:16:18,304][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:16:18,805][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:16:19,301][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:16:19,803][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:16:20,300][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:16:20,796][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:16:21,310][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:16:21,807][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:16:22,304][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:16:22,802][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:16:23,302][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:16:23,801][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:16:24,297][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:16:24,795][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:16:25,302][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:16:25,801][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:16:26,296][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:16:26,814][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:16:27,309][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:16:27,806][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:16:28,304][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:16:28,804][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:16:29,319][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:16:29,817][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:16:30,318][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:16:30,816][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:16:31,322][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:16:31,824][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:16:32,325][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:16:32,828][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:16:33,328][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:16:33,828][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:16:34,331][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:16:34,831][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:16:35,335][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:16:35,838][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:16:36,338][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:16:36,837][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:16:37,345][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:16:37,846][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:16:38,349][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:16:38,853][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:16:39,357][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:16:39,857][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:16:40,364][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10130 tokens. -[2025-11-10 00:16:41,232][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 55.70%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 -[2025-11-10 00:16:43,183][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:16:43,185][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:16:43,187][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:16:44,177][__main__][INFO] - Iteration 403 took 1m 0s (38.02% Gen, 60.34% Train). Generation: 22s, Training: 36s. Estimated remaining time: 44h 1m 7s. Estimated total time: 50h 18m 54s. Time estimates for 10 more iterations: 10m 3s, 100 more iterations: 1h 40m 37s, 500 more iterations: 8h 23m 9s. -[2025-11-10 00:16:44,179][__main__][INFO] - Starting iteration 403. -[2025-11-10 00:16:44,649][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 40 and human policies 1. -[2025-11-10 00:16:44,649][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:17:11,056][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 00:17:11,905][__main__][INFO] - Number of regex retries in iteration 403: 1 -[2025-11-10 00:17:11,905][__main__][INFO] - agents played in iteration 403 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:17:12,691][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:17:12,714][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:17:12,737][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:17:12,759][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:17:12,760][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:17:12,761][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:17:13,518][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:17:13,972][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:17:14,472][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:17:14,973][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:17:15,475][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:17:15,977][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:17:16,484][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:17:16,986][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:17:17,486][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:17:18,006][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:17:18,505][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:17:19,006][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:17:19,512][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:17:20,013][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:17:20,520][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:17:21,021][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:17:21,520][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:17:22,022][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:17:22,523][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:17:23,021][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:17:23,524][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:17:24,023][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:17:24,526][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:17:25,026][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:17:25,525][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:17:26,027][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:17:26,527][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:17:27,025][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:17:27,523][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:17:28,022][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:17:28,528][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:17:29,026][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:17:29,528][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:17:30,024][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:17:30,523][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:17:31,028][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:17:31,533][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:17:32,034][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:17:32,536][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:17:33,037][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:17:33,530][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:17:34,029][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:17:34,532][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:17:35,038][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:17:35,542][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:17:36,044][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:17:36,547][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:17:37,046][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:17:37,545][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:17:38,047][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:17:38,549][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:17:39,053][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:17:39,553][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:17:40,057][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:17:40,562][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:17:41,067][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:17:41,575][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:17:42,077][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:17:42,580][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:17:43,085][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:17:43,590][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:17:44,096][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:17:44,597][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:17:45,104][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:17:45,626][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10110 tokens. -[2025-11-10 00:17:46,464][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 55.72%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:32 -[2025-11-10 00:17:47,289][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:17:47,291][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:17:47,293][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:17:48,289][__main__][INFO] - Iteration 404 took 1m 3s (42.83% Gen, 55.61% Train). Generation: 27s, Training: 35s. Estimated remaining time: 46h 43m 11s. Estimated total time: 53h 2m 2s. Time estimates for 10 more iterations: 10m 36s, 100 more iterations: 1h 46m 4s, 500 more iterations: 8h 50m 20s. -[2025-11-10 00:17:48,291][__main__][INFO] - Starting iteration 404. -[2025-11-10 00:17:48,758][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 40 and human policies 1. -[2025-11-10 00:17:48,759][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:18:16,502][__main__][INFO] - Number of regex retries in iteration 404: 0 -[2025-11-10 00:18:16,505][__main__][INFO] - agents played in iteration 404 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:18:17,285][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:18:17,313][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:18:17,339][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:18:17,362][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:18:17,363][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:18:17,364][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:18:18,077][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:18:18,530][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:18:19,037][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:18:19,537][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:18:20,035][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:18:20,533][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:18:21,032][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:18:21,527][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:18:22,026][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:18:22,521][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:18:23,021][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:18:23,534][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:18:24,030][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:18:24,526][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:18:25,025][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:18:25,522][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:18:26,018][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:18:26,512][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:18:27,005][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:18:27,516][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:18:28,009][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:18:28,505][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:18:29,001][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:18:29,498][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:18:29,999][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:18:30,496][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:18:30,994][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:18:31,500][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:18:32,000][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:18:32,497][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:18:32,992][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:18:33,487][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:18:33,983][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:18:34,478][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:18:34,973][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:18:35,470][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:18:35,967][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:18:36,463][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:18:36,963][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:18:37,462][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:18:37,958][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:18:38,453][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:18:38,952][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:18:39,453][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:18:39,952][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:18:40,458][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:18:40,960][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:18:41,458][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:18:41,963][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:18:42,463][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:18:42,990][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:18:43,493][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:18:44,000][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:18:44,503][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:18:45,005][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:18:45,507][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:18:46,017][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:18:46,517][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:18:47,028][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:18:47,531][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:18:48,034][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:18:48,547][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:18:49,049][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:18:49,553][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:18:50,053][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10089 tokens. -[2025-11-10 00:18:50,900][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 55.64%, Block Peak % of device VRAM: 62.12%, ΔTime: 00:00:32 -[2025-11-10 00:18:51,551][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:18:51,553][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:18:51,556][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:18:52,422][__main__][INFO] - Iteration 405 took 1m 3s (43.58% Gen, 55.06% Train). Generation: 27s, Training: 35s. Estimated remaining time: 46h 43m 18s. Estimated total time: 53h 3m 14s. Time estimates for 10 more iterations: 10m 36s, 100 more iterations: 1h 46m 6s, 500 more iterations: 8h 50m 32s. -[2025-11-10 00:18:52,425][__main__][INFO] - Starting iteration 405. -[2025-11-10 00:18:52,903][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 40 and human policies 1. -[2025-11-10 00:18:52,904][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:19:17,096][__main__][INFO] - Number of regex retries in iteration 405: 0 -[2025-11-10 00:19:17,097][__main__][INFO] - agents played in iteration 405 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:19:17,882][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:19:17,909][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:19:17,936][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:19:17,960][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:19:17,960][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:19:17,961][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:19:18,703][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:19:19,156][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:19:19,657][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:19:20,165][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:19:20,665][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:19:21,166][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:19:21,661][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:19:22,158][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:19:22,667][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:19:23,172][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:19:23,672][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:19:24,177][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:19:24,674][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:19:25,175][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:19:25,672][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:19:26,170][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:19:26,669][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:19:27,167][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:19:27,664][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:19:28,168][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:19:28,670][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:19:29,169][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:19:29,666][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:19:30,164][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:19:30,671][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:19:31,174][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:19:31,671][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:19:32,168][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:19:32,665][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:19:33,165][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:19:33,663][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:19:34,160][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:19:34,669][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:19:35,164][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:19:35,664][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:19:36,160][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:19:36,657][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:19:37,153][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:19:37,646][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:19:38,143][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:19:38,654][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:19:39,149][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:19:39,643][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:19:40,141][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:19:40,641][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:19:41,151][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:19:41,647][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:19:42,143][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:19:42,652][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:19:43,151][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:19:43,650][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:19:44,155][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:19:44,658][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:19:45,165][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:19:45,665][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:19:46,163][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:19:46,676][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:19:47,176][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:19:47,676][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:19:48,180][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:19:48,683][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:19:49,186][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:19:49,690][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:19:50,192][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:19:50,695][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10070 tokens. -[2025-11-10 00:19:51,535][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 13.97%, Current % of VRAM taken: 55.47%, Block Peak % of device VRAM: 62.14%, ΔTime: 00:00:32 -[2025-11-10 00:19:52,302][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:19:52,303][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:19:52,305][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:19:53,273][__main__][INFO] - Iteration 406 took 1m 0s (40.07% Gen, 58.32% Train). Generation: 24s, Training: 35s. Estimated remaining time: 43h 57m 36s. Estimated total time: 50h 18m 32s. Time estimates for 10 more iterations: 10m 3s, 100 more iterations: 1h 40m 37s, 500 more iterations: 8h 23m 5s. -[2025-11-10 00:19:53,275][__main__][INFO] - Starting iteration 406. -[2025-11-10 00:19:53,770][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 40 and human policies 1. -[2025-11-10 00:19:53,771][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:20:17,806][__main__][INFO] - Number of regex retries in iteration 406: 0 -[2025-11-10 00:20:17,808][__main__][INFO] - agents played in iteration 406 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:20:18,611][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.50%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:20:18,634][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.50%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:20:18,657][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.50%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:20:18,680][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.50%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:20:18,681][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:20:18,682][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:20:19,417][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:20:19,873][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:20:20,377][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:20:20,878][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:20:21,378][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:20:21,880][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:20:22,378][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:20:22,877][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:20:23,382][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:20:23,881][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:20:24,384][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:20:24,888][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:20:25,388][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:20:25,885][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:20:26,386][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:20:26,883][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:20:27,383][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:20:27,886][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:20:28,389][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:20:28,892][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:20:29,392][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:20:29,892][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:20:30,391][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:20:30,892][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:20:31,395][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:20:31,892][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:20:32,391][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:20:32,893][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:20:33,393][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:20:33,892][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:20:34,393][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:20:34,894][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:20:35,397][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:20:35,905][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:20:36,404][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:20:36,905][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:20:37,405][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:20:37,905][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:20:38,402][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:20:38,903][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:20:39,406][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:20:39,903][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:20:40,399][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:20:40,899][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:20:41,402][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:20:41,903][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:20:42,402][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:20:42,904][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:20:43,410][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:20:43,908][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:20:44,409][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:20:44,907][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:20:45,406][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:20:45,907][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:20:46,407][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:20:46,911][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:20:47,418][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:20:47,924][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:20:48,434][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:20:48,940][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:20:49,451][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:20:49,957][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:20:50,464][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:20:50,968][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:20:51,475][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10166 tokens. -[2025-11-10 00:20:52,299][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 55.67%, Block Peak % of device VRAM: 62.36%, ΔTime: 00:00:32 -[2025-11-10 00:20:52,941][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:20:52,943][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:20:52,945][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:20:53,780][__main__][INFO] - Iteration 407 took 1m 0s (40.05% Gen, 58.55% Train). Generation: 24s, Training: 35s. Estimated remaining time: 43h 38m 34s. Estimated total time: 50h 0m 30s. Time estimates for 10 more iterations: 10m 0s, 100 more iterations: 1h 40m 1s, 500 more iterations: 8h 20m 5s. -[2025-11-10 00:20:53,783][__main__][INFO] - Starting iteration 407. -[2025-11-10 00:20:54,251][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 40 and human policies 1. -[2025-11-10 00:20:54,252][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:21:08,378][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 00:21:20,818][__main__][INFO] - Number of regex retries in iteration 407: 1 -[2025-11-10 00:21:20,820][__main__][INFO] - agents played in iteration 407 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:21:21,622][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:21:21,648][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:21:21,673][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:21:21,697][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:21:21,697][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:21:21,698][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:21:22,480][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:21:22,954][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:21:23,462][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:21:23,963][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:21:24,469][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:21:24,971][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:21:25,487][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:21:25,987][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:21:26,488][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:21:26,997][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:21:27,497][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:21:27,997][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:21:28,499][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:21:28,999][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:21:29,499][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:21:29,999][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:21:30,502][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:21:31,005][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:21:31,505][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:21:32,007][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:21:32,506][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:21:33,006][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:21:33,507][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:21:34,010][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:21:34,510][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:21:35,011][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:21:35,508][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:21:36,012][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:21:36,520][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:21:37,020][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:21:37,521][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:21:38,022][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:21:38,518][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:21:39,016][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:21:39,515][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:21:40,015][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:21:40,515][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:21:41,012][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:21:41,514][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:21:42,012][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:21:42,512][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:21:43,008][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:21:43,504][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:21:44,009][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:21:44,507][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:21:45,007][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:21:45,506][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:21:46,003][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:21:46,504][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:21:47,005][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:21:47,504][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:21:48,005][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:21:48,504][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:21:49,003][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:21:49,502][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:21:49,999][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:21:50,504][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:21:51,002][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:21:51,505][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:21:52,016][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:21:52,519][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:21:53,024][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:21:53,526][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:21:54,028][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:21:54,530][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10194 tokens. -[2025-11-10 00:21:55,368][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.09%, Current % of VRAM taken: 55.59%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 -[2025-11-10 00:21:56,030][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:21:56,031][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:21:56,034][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:21:56,864][__main__][INFO] - Iteration 408 took 1m 2s (42.43% Gen, 56.24% Train). Generation: 26s, Training: 35s. Estimated remaining time: 45h 47m 40s. Estimated total time: 52h 10m 40s. Time estimates for 10 more iterations: 10m 26s, 100 more iterations: 1h 44m 21s, 500 more iterations: 8h 41m 46s. -[2025-11-10 00:21:56,867][__main__][INFO] - Starting iteration 408. -[2025-11-10 00:21:57,333][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 40 and human policies 1. -[2025-11-10 00:21:57,333][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:22:21,816][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 00:22:25,379][__main__][INFO] - Number of regex retries in iteration 408: 1 -[2025-11-10 00:22:25,380][__main__][INFO] - agents played in iteration 408 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:22:26,222][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:22:26,246][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:22:26,269][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:22:26,292][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:22:26,292][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:22:26,293][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:22:27,041][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:22:27,498][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:22:28,008][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:22:28,508][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:22:29,008][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:22:29,506][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:22:30,007][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:22:30,505][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:22:31,003][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:22:31,505][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:22:32,004][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:22:32,506][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:22:33,006][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:22:33,506][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:22:34,008][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:22:34,508][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:22:35,009][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:22:35,514][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:22:36,021][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:22:36,527][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:22:37,029][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:22:37,531][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:22:38,033][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:22:38,531][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:22:39,033][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:22:39,532][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:22:40,033][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:22:40,541][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:22:41,044][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:22:41,542][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:22:42,038][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:22:42,534][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:22:43,038][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:22:43,535][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:22:44,031][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:22:44,529][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:22:45,027][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:22:45,524][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:22:46,022][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:22:46,519][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:22:47,018][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:22:47,516][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:22:48,015][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:22:48,516][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:22:49,015][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:22:49,515][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:22:50,012][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:22:50,508][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:22:51,011][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:22:51,508][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:22:52,005][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:22:52,508][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:22:53,005][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:22:53,504][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:22:54,001][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:22:54,498][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:22:55,005][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:22:55,506][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:22:56,008][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:22:56,507][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:22:57,014][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:22:57,516][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:22:58,017][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:22:58,520][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:22:59,028][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10092 tokens. -[2025-11-10 00:22:59,841][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.25%, Current % of VRAM taken: 55.75%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:32 -[2025-11-10 00:23:00,621][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:23:00,623][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:23:00,625][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:23:01,581][__main__][INFO] - Iteration 409 took 1m 4s (43.65% Gen, 54.86% Train). Generation: 28s, Training: 35s. Estimated remaining time: 47h 8m 21s. Estimated total time: 53h 32m 26s. Time estimates for 10 more iterations: 10m 42s, 100 more iterations: 1h 47m 4s, 500 more iterations: 8h 55m 24s. -[2025-11-10 00:23:01,583][__main__][INFO] - Starting iteration 409. -[2025-11-10 00:23:02,049][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 40 and human policies 1. -[2025-11-10 00:23:02,050][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:23:30,058][__main__][INFO] - Number of regex retries in iteration 409: 0 -[2025-11-10 00:23:30,060][__main__][INFO] - agents played in iteration 409 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:23:30,868][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:23:30,893][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:23:30,918][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:23:30,942][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:23:30,942][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:23:30,943][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:23:31,740][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:23:32,199][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:23:32,703][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:23:33,212][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:23:33,711][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:23:34,219][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:23:34,719][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:23:35,218][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:23:35,720][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:23:36,218][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:23:36,723][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:23:37,220][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:23:37,719][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:23:38,222][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:23:38,725][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:23:39,225][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:23:39,726][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:23:40,226][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:23:40,725][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:23:41,224][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:23:41,723][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:23:42,228][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:23:42,731][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:23:43,233][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:23:43,731][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:23:44,234][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:23:44,741][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:23:45,244][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:23:45,748][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:23:46,249][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:23:46,750][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:23:47,251][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:23:47,753][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:23:48,251][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:23:48,749][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:23:49,246][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:23:49,746][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:23:50,246][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:23:50,742][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:23:51,242][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:23:51,743][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:23:52,239][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:23:52,740][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:23:53,234][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:23:53,729][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:23:54,225][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:23:54,727][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:23:55,231][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:23:55,730][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:23:56,232][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:23:56,731][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:23:57,230][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:23:57,737][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:23:58,240][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:23:58,744][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:23:59,245][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:23:59,749][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:24:00,249][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:24:00,753][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:24:01,255][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:24:01,758][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:24:02,261][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:24:02,762][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:24:03,267][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:24:03,770][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10181 tokens. -[2025-11-10 00:24:04,617][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 55.60%, Block Peak % of device VRAM: 62.23%, ΔTime: 00:00:32 -[2025-11-10 00:24:05,263][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:24:05,264][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:24:05,266][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:24:06,078][__main__][INFO] - Iteration 410 took 1m 4s (43.75% Gen, 54.98% Train). Generation: 28s, Training: 35s. Estimated remaining time: 46h 56m 18s. Estimated total time: 53h 21m 27s. Time estimates for 10 more iterations: 10m 40s, 100 more iterations: 1h 46m 42s, 500 more iterations: 8h 53m 34s. -[2025-11-10 00:24:06,081][__main__][INFO] - Starting iteration 410. -[2025-11-10 00:24:06,568][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 40 and human policies 1. -[2025-11-10 00:24:06,569][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:24:28,019][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 00:24:31,912][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 00:24:34,171][__main__][INFO] - Number of regex retries in iteration 410: 2 -[2025-11-10 00:24:34,171][__main__][INFO] - agents played in iteration 410 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:24:34,988][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:24:35,019][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:24:35,046][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:24:35,071][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:24:35,071][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:24:35,073][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:24:35,799][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:24:36,257][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:24:36,759][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:24:37,265][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:24:37,763][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:24:38,262][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:24:38,762][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:24:39,263][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:24:39,762][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:24:40,261][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:24:40,759][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:24:41,262][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:24:41,759][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:24:42,263][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:24:42,761][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:24:43,261][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:24:43,761][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:24:44,259][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:24:44,760][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:24:45,257][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:24:45,753][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:24:46,253][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:24:46,751][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:24:47,247][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:24:47,745][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:24:48,248][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:24:48,753][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:24:49,253][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:24:49,752][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:24:50,260][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:24:50,762][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:24:51,260][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:24:51,763][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:24:52,259][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:24:52,753][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:24:53,248][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:24:53,745][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:24:54,248][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:24:54,747][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:24:55,245][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:24:55,744][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:24:56,244][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:24:56,742][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:24:57,246][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:24:57,745][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:24:58,248][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:24:58,748][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:24:59,251][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:24:59,749][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:25:00,252][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:25:00,781][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:25:01,285][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:25:01,794][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:25:02,297][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:25:02,804][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:25:03,330][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:25:03,834][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:25:04,335][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:25:04,870][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:25:06,425][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:25:06,928][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:25:07,431][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:25:07,932][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:25:08,433][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:25:08,932][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10124 tokens. -[2025-11-10 00:25:09,755][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 55.56%, Block Peak % of device VRAM: 62.23%, ΔTime: 00:00:33 -[2025-11-10 00:25:10,418][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:25:10,419][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:25:10,422][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:25:12,197][__main__][INFO] - Iteration 411 took 1m 5s (42.06% Gen, 55.23% Train). Generation: 27s, Training: 36s. Estimated remaining time: 48h 15m 12s. Estimated total time: 54h 41m 28s. Time estimates for 10 more iterations: 10m 56s, 100 more iterations: 1h 49m 22s, 500 more iterations: 9h 6m 54s. -[2025-11-10 00:25:12,202][__main__][INFO] - Starting iteration 411. -[2025-11-10 00:25:12,677][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 41 and human policies 1. -[2025-11-10 00:25:12,678][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:25:41,302][__main__][INFO] - Number of regex retries in iteration 411: 0 -[2025-11-10 00:25:41,303][__main__][INFO] - agents played in iteration 411 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:25:42,101][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:25:42,130][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:25:42,154][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:25:42,177][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:25:42,178][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:25:42,179][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:25:42,917][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:25:43,372][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:25:43,876][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:25:44,378][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:25:44,884][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:25:45,383][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:25:45,884][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:25:46,389][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:25:46,888][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:25:47,388][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:25:47,886][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:25:48,387][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:25:48,886][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:25:49,387][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:25:49,885][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:25:50,387][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:25:50,894][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:25:51,395][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:25:51,897][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:25:52,396][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:25:52,893][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:25:53,409][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:25:53,909][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:25:54,405][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:25:54,913][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:25:55,416][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:25:55,929][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:25:56,435][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:25:56,934][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:25:57,435][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:25:57,931][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:25:58,431][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:25:58,929][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:25:59,426][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:25:59,939][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:26:00,436][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:26:00,932][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:26:01,430][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:26:01,928][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:26:02,436][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:26:02,934][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:26:03,432][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:26:03,941][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:26:04,436][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:26:04,933][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:26:05,429][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:26:05,924][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:26:06,424][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:26:06,923][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:26:07,424][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:26:07,925][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:26:08,422][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:26:08,922][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:26:09,422][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:26:09,921][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:26:10,436][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:26:10,938][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:26:11,438][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:26:11,938][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:26:12,439][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:26:12,947][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:26:13,447][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:26:13,951][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:26:14,456][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:26:14,958][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10152 tokens. -[2025-11-10 00:26:15,781][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 55.72%, Block Peak % of device VRAM: 62.18%, ΔTime: 00:00:32 -[2025-11-10 00:26:16,535][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:26:16,537][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:26:16,538][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:26:17,571][__main__][INFO] - Iteration 412 took 1m 4s (44.11% Gen, 54.30% Train). Generation: 28s, Training: 35s. Estimated remaining time: 47h 37m 24s. Estimated total time: 54h 4m 45s. Time estimates for 10 more iterations: 10m 48s, 100 more iterations: 1h 48m 9s, 500 more iterations: 9h 0m 47s. -[2025-11-10 00:26:17,573][__main__][INFO] - Starting iteration 412. -[2025-11-10 00:26:18,045][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 41 and human policies 1. -[2025-11-10 00:26:18,047][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:26:40,098][__main__][INFO] - Number of regex retries in iteration 412: 0 -[2025-11-10 00:26:40,100][__main__][INFO] - agents played in iteration 412 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:26:40,900][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:26:40,923][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:26:40,948][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:26:40,971][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:26:40,972][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:26:40,974][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:26:41,770][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:26:42,231][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:26:42,735][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:26:43,239][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:26:43,747][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:26:44,254][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:26:44,757][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:26:45,258][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:26:45,762][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:26:46,261][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:26:46,761][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:26:47,260][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:26:47,761][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:26:48,260][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:26:48,761][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:26:49,263][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:26:49,771][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:26:50,274][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:26:50,782][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:26:51,285][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:26:51,787][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:26:52,287][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:26:52,787][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:26:53,288][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:26:53,795][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:26:54,297][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:26:54,801][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:26:55,318][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:26:55,822][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:26:56,343][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:26:56,846][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:26:57,349][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:26:57,849][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:26:58,375][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:26:58,880][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:26:59,381][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:26:59,886][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:27:00,391][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:27:00,892][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:27:01,397][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:27:01,898][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:27:02,395][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:27:02,894][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:27:03,393][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:27:03,891][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:27:04,393][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:27:04,892][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:27:05,391][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:27:05,887][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:27:06,387][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:27:06,890][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:27:07,386][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:27:07,882][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:27:08,377][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:27:08,879][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:27:09,377][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:27:09,873][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:27:10,373][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:27:10,870][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:27:11,366][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:27:11,866][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:27:12,366][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:27:12,869][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:27:13,372][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:27:13,875][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10262 tokens. -[2025-11-10 00:27:14,675][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 55.65%, Block Peak % of device VRAM: 62.41%, ΔTime: 00:00:32 -[2025-11-10 00:27:15,328][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:27:15,331][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:27:15,333][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:27:16,174][__main__][INFO] - Iteration 413 took 58s (37.94% Gen, 60.61% Train). Generation: 22s, Training: 35s. Estimated remaining time: 41h 58m 11s. Estimated total time: 48h 26m 30s. Time estimates for 10 more iterations: 9m 41s, 100 more iterations: 1h 36m 53s, 500 more iterations: 8h 4m 25s. -[2025-11-10 00:27:16,177][__main__][INFO] - Starting iteration 413. -[2025-11-10 00:27:16,635][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 41 and human policies 1. -[2025-11-10 00:27:16,636][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:27:45,284][__main__][INFO] - Number of regex retries in iteration 413: 0 -[2025-11-10 00:27:45,285][__main__][INFO] - agents played in iteration 413 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:27:46,115][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:27:46,140][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:27:46,165][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:27:46,188][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:27:46,190][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:27:46,191][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:27:46,986][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:27:47,444][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:27:47,956][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:27:48,460][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:27:48,968][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:27:49,470][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:27:49,975][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:27:50,477][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:27:50,978][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:27:51,479][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:27:51,981][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:27:52,482][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:27:52,992][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:27:53,492][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:27:53,990][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:27:54,490][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:27:54,988][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:27:55,490][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:27:55,990][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:27:56,494][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:27:56,995][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:27:57,492][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:27:57,989][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:27:58,490][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:27:58,990][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:27:59,489][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:27:59,990][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:28:00,489][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:28:00,994][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:28:01,494][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:28:01,995][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:28:02,492][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:28:02,992][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:28:03,491][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:28:03,992][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:28:04,494][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:28:04,992][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:28:05,488][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:28:05,988][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:28:06,486][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:28:06,984][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:28:07,486][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:28:07,986][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:28:08,488][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:28:08,987][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:28:09,486][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:28:09,986][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:28:10,487][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:28:10,988][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:28:11,485][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:28:11,982][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:28:12,479][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:28:12,976][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:28:13,474][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:28:13,972][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:28:14,470][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:28:14,970][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:28:15,495][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:28:15,995][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:28:16,503][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:28:17,002][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:28:17,506][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:28:18,018][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:28:18,519][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:28:19,027][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10149 tokens. -[2025-11-10 00:28:19,813][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 55.67%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:32 -[2025-11-10 00:28:20,561][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:28:20,563][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:28:20,566][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:28:21,437][__main__][INFO] - Iteration 414 took 1m 4s (44.21% Gen, 54.44% Train). Generation: 28s, Training: 35s. Estimated remaining time: 47h 30m 43s. Estimated total time: 54h 0m 7s. Time estimates for 10 more iterations: 10m 48s, 100 more iterations: 1h 48m 0s, 500 more iterations: 9h 0m 1s. -[2025-11-10 00:28:21,440][__main__][INFO] - Starting iteration 414. -[2025-11-10 00:28:21,918][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 41 and human policies 1. -[2025-11-10 00:28:21,919][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:28:50,386][__main__][INFO] - Number of regex retries in iteration 414: 0 -[2025-11-10 00:28:50,389][__main__][INFO] - agents played in iteration 414 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:28:51,249][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:28:51,273][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:28:51,297][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:28:51,321][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:28:51,321][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:28:51,322][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:28:52,133][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:28:52,590][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:28:53,106][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:28:53,606][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:28:54,107][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:28:54,618][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:28:55,118][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:28:55,618][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:28:56,122][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:28:56,629][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:28:57,137][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:28:57,637][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:28:58,135][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:28:58,638][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:28:59,137][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:28:59,657][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:29:00,160][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:29:00,669][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:29:01,168][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:29:01,670][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:29:02,175][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:29:02,678][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:29:03,184][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:29:03,688][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:29:04,194][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:29:04,694][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:29:05,196][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:29:05,699][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:29:06,209][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:29:06,710][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:29:07,212][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:29:07,720][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:29:08,220][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:29:08,723][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:29:09,221][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:29:09,719][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:29:10,219][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:29:10,719][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:29:11,215][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:29:11,714][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:29:12,216][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:29:12,715][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:29:13,216][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:29:13,717][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:29:14,224][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:29:14,727][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:29:15,226][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:29:15,728][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:29:16,232][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:29:16,750][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:29:17,249][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:29:17,747][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:29:18,247][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:29:18,745][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:29:19,257][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:29:19,754][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:29:20,252][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:29:20,750][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:29:21,248][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:29:21,761][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:29:22,263][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:29:22,767][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:29:23,274][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:29:23,775][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:29:24,280][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10207 tokens. -[2025-11-10 00:29:25,117][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 55.65%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 -[2025-11-10 00:29:25,790][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:29:25,792][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:29:25,793][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:29:26,631][__main__][INFO] - Iteration 415 took 1m 4s (43.99% Gen, 54.71% Train). Generation: 28s, Training: 35s. Estimated remaining time: 47h 25m 11s. Estimated total time: 53h 55m 41s. Time estimates for 10 more iterations: 10m 47s, 100 more iterations: 1h 47m 51s, 500 more iterations: 8h 59m 16s. -[2025-11-10 00:29:26,634][__main__][INFO] - Starting iteration 415. -[2025-11-10 00:29:27,114][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 41 and human policies 1. -[2025-11-10 00:29:27,115][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:29:51,902][__main__][INFO] - Number of regex retries in iteration 415: 0 -[2025-11-10 00:29:51,905][__main__][INFO] - agents played in iteration 415 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:29:52,871][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:29:52,905][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:29:52,934][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:29:52,958][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:29:52,958][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:29:52,959][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:29:53,790][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:29:54,255][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:29:54,765][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:29:55,273][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:29:55,777][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:29:56,286][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:29:56,790][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:29:57,300][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:29:57,804][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:29:58,310][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:29:58,818][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:29:59,319][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:29:59,822][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:30:00,327][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:30:00,831][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:30:01,332][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:30:01,832][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:30:02,335][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:30:02,841][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:30:03,343][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:30:03,846][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:30:04,348][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:30:04,849][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:30:05,378][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:30:05,886][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:30:06,389][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:30:06,902][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:30:07,406][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:30:07,910][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:30:08,412][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:30:08,915][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:30:09,419][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:30:09,923][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:30:10,425][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:30:10,923][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:30:11,426][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:30:11,926][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:30:12,424][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:30:12,921][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:30:13,422][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:30:13,919][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:30:14,421][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:30:14,922][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:30:15,419][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:30:15,923][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:30:16,421][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:30:16,919][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:30:17,420][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:30:17,921][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:30:18,416][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:30:18,913][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:30:19,412][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:30:19,911][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:30:20,409][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:30:20,908][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:30:21,412][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:30:21,912][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:30:22,407][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:30:22,905][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:30:23,406][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:30:23,910][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:30:24,407][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:30:24,907][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:30:25,407][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:30:25,906][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10150 tokens. -[2025-11-10 00:30:26,706][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 55.57%, Block Peak % of device VRAM: 62.20%, ΔTime: 00:00:32 -[2025-11-10 00:30:27,346][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:30:27,348][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:30:27,350][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:30:28,181][__main__][INFO] - Iteration 416 took 1m 1s (40.59% Gen, 58.04% Train). Generation: 24s, Training: 35s. Estimated remaining time: 44h 21m 50s. Estimated total time: 50h 53m 21s. Time estimates for 10 more iterations: 10m 10s, 100 more iterations: 1h 41m 46s, 500 more iterations: 8h 28m 53s. -[2025-11-10 00:30:28,183][__main__][INFO] - Starting iteration 416. -[2025-11-10 00:30:28,683][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 41 and human policies 1. -[2025-11-10 00:30:28,684][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:31:00,864][__main__][INFO] - Number of regex retries in iteration 416: 0 -[2025-11-10 00:31:00,864][__main__][INFO] - agents played in iteration 416 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:31:01,714][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:31:01,737][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:31:01,761][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:31:01,787][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:31:01,788][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:31:01,789][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:31:02,609][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:31:03,066][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:31:03,571][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:31:04,070][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:31:04,579][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:31:05,087][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:31:05,587][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:31:06,102][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:31:06,605][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:31:07,107][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:31:07,611][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:31:08,111][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:31:08,612][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:31:09,112][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:31:09,611][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:31:10,113][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:31:12,254][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:31:12,740][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:31:13,245][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:31:13,746][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:31:14,250][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:31:14,747][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:31:15,250][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:31:15,748][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:31:16,250][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:31:16,748][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:31:17,250][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:31:17,749][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:31:18,258][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:31:18,758][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:31:19,257][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:31:19,753][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:31:20,250][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:31:20,761][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:31:21,258][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:31:21,754][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:31:22,251][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:31:22,749][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:31:23,245][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:31:23,743][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:31:24,238][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:31:24,740][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:31:25,236][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:31:25,738][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:31:26,237][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:31:26,732][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:31:27,231][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:31:27,732][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:31:28,231][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:31:28,731][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:31:29,235][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:31:29,733][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:31:30,230][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:31:30,726][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:31:31,223][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:31:31,721][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:31:32,218][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:31:32,713][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:31:33,211][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:31:33,710][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:31:34,205][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:31:34,701][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:31:35,195][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:31:35,689][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:31:36,193][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10091 tokens. -[2025-11-10 00:31:36,993][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.03%, Current % of VRAM taken: 55.52%, Block Peak % of device VRAM: 62.10%, ΔTime: 00:00:34 -[2025-11-10 00:31:37,646][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:31:37,647][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:31:37,649][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:31:38,450][__main__][INFO] - Iteration 417 took 1m 9s (46.12% Gen, 52.72% Train). Generation: 32s, Training: 36s. Estimated remaining time: 51h 35m 41s. Estimated total time: 58h 8m 23s. Time estimates for 10 more iterations: 11m 37s, 100 more iterations: 1h 56m 16s, 500 more iterations: 9h 41m 23s. -[2025-11-10 00:31:38,452][__main__][INFO] - Starting iteration 417. -[2025-11-10 00:31:38,955][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 41 and human policies 1. -[2025-11-10 00:31:38,956][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:32:10,427][__main__][INFO] - Number of regex retries in iteration 417: 0 -[2025-11-10 00:32:10,428][__main__][INFO] - agents played in iteration 417 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:32:11,225][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:32:11,248][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:32:11,271][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:32:11,295][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:32:11,296][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:32:11,296][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:32:12,081][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:32:12,540][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:32:13,046][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:32:13,549][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:32:14,055][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:32:14,557][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:32:15,062][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:32:15,566][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:32:16,077][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:32:16,577][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:32:17,082][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:32:17,585][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:32:18,089][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:32:18,630][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:32:19,136][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:32:19,642][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:32:20,142][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:32:20,641][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:32:21,141][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:32:21,640][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:32:22,145][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:32:22,646][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:32:23,147][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:32:23,644][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:32:24,141][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:32:24,637][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:32:25,138][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:32:25,639][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:32:26,139][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:32:26,640][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:32:27,140][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:32:27,642][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:32:28,139][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:32:28,640][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:32:29,139][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:32:29,637][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:32:30,136][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:32:30,634][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:32:31,131][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:32:31,629][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:32:32,128][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:32:32,625][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:32:33,126][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:32:33,628][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:32:34,128][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:32:34,624][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:32:35,123][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:32:35,619][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:32:36,114][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:32:36,614][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:32:37,112][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:32:37,608][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:32:38,110][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:32:38,606][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:32:39,102][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:32:39,599][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:32:40,095][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:32:40,593][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:32:41,090][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:32:41,591][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:32:42,105][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:32:42,604][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:32:43,106][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:32:43,608][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:32:44,108][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10246 tokens. -[2025-11-10 00:32:44,933][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 55.69%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 -[2025-11-10 00:32:45,655][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:32:45,657][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:32:45,659][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:32:46,559][__main__][INFO] - Iteration 418 took 1m 7s (46.55% Gen, 52.11% Train). Generation: 31s, Training: 35s. Estimated remaining time: 49h 46m 22s. Estimated total time: 56h 20m 11s. Time estimates for 10 more iterations: 11m 16s, 100 more iterations: 1h 52m 40s, 500 more iterations: 9h 23m 21s. -[2025-11-10 00:32:46,561][__main__][INFO] - Starting iteration 418. -[2025-11-10 00:32:47,023][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 41 and human policies 1. -[2025-11-10 00:32:47,024][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:33:14,212][__main__][INFO] - Number of regex retries in iteration 418: 0 -[2025-11-10 00:33:14,214][__main__][INFO] - agents played in iteration 418 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:33:15,022][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:33:15,046][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:33:15,070][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:33:15,093][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:33:15,094][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:33:15,095][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:33:15,884][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:33:16,342][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:33:16,849][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:33:17,352][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:33:17,859][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:33:18,358][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:33:18,862][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:33:19,361][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:33:19,867][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:33:20,371][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:33:20,869][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:33:21,374][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:33:21,888][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:33:22,392][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:33:22,896][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:33:23,396][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:33:23,895][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:33:24,397][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:33:24,895][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:33:25,395][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:33:25,898][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:33:26,397][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:33:26,903][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:33:27,403][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:33:27,903][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:33:28,407][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:33:28,906][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:33:29,405][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:33:29,906][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:33:30,408][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:33:30,916][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:33:31,416][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:33:31,912][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:33:32,433][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:33:32,933][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:33:33,435][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:33:33,934][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:33:34,435][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:33:34,937][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:33:35,435][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:33:35,937][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:33:36,441][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:33:36,939][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:33:37,439][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:33:37,935][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:33:38,434][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:33:38,939][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:33:39,444][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:33:39,944][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:33:40,446][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:33:40,945][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:33:41,445][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:33:41,947][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:33:42,448][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:33:42,947][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:33:43,446][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:33:43,947][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:33:44,447][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:33:44,947][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:33:45,446][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:33:45,946][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:33:46,444][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:33:46,948][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:33:47,447][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:33:47,946][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10228 tokens. -[2025-11-10 00:33:48,759][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 55.65%, Block Peak % of device VRAM: 62.46%, ΔTime: 00:00:32 -[2025-11-10 00:33:49,400][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:33:49,402][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:33:49,403][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:33:50,200][__main__][INFO] - Iteration 419 took 1m 3s (43.04% Gen, 55.70% Train). Generation: 27s, Training: 35s. Estimated remaining time: 46h 4m 0s. Estimated total time: 52h 38m 53s. Time estimates for 10 more iterations: 10m 31s, 100 more iterations: 1h 45m 17s, 500 more iterations: 8h 46m 28s. -[2025-11-10 00:33:50,203][__main__][INFO] - Starting iteration 419. -[2025-11-10 00:33:50,696][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 41 and human policies 1. -[2025-11-10 00:33:50,697][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:34:10,155][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 00:34:21,983][__main__][INFO] - Number of regex retries in iteration 419: 1 -[2025-11-10 00:34:21,984][__main__][INFO] - agents played in iteration 419 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:34:22,786][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:34:22,811][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:34:22,837][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:34:22,861][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:34:22,862][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:34:22,863][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:34:23,669][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:34:24,130][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:34:24,634][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:34:25,137][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:34:25,642][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:34:26,147][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:34:26,650][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:34:27,151][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:34:27,769][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:34:28,710][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:34:29,217][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:34:29,718][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:34:30,233][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:34:30,734][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:34:31,234][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:34:31,735][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:34:32,238][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:34:32,740][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:34:33,242][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:34:33,741][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:34:34,249][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:34:34,749][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:34:35,249][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:34:35,747][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:34:36,250][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:34:36,755][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:34:37,257][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:34:37,759][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:34:38,262][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:34:38,758][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:34:39,260][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:34:39,764][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:34:40,262][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:34:40,769][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:34:41,270][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:34:41,768][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:34:42,271][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:34:42,770][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:34:43,270][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:34:43,768][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:34:44,272][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:34:44,776][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:34:45,278][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:34:45,782][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:34:46,282][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:34:46,782][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:34:47,279][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:34:47,782][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:34:48,278][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:34:48,778][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:34:49,274][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:34:49,772][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:34:50,276][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:34:50,773][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:34:51,274][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:34:51,770][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:34:52,269][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:34:52,773][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:34:53,271][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:34:53,770][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:34:54,271][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:34:54,768][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:34:55,269][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:34:55,769][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:34:56,271][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10159 tokens. -[2025-11-10 00:34:57,091][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 55.63%, Block Peak % of device VRAM: 62.27%, ΔTime: 00:00:33 -[2025-11-10 00:34:57,733][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:34:57,735][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:34:57,737][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:34:58,590][__main__][INFO] - Iteration 420 took 1m 7s (46.08% Gen, 52.66% Train). Generation: 31s, Training: 35s. Estimated remaining time: 49h 58m 42s. Estimated total time: 56h 34m 43s. Time estimates for 10 more iterations: 11m 18s, 100 more iterations: 1h 53m 9s, 500 more iterations: 9h 25m 47s. -[2025-11-10 00:34:58,592][__main__][INFO] - Starting iteration 420. -[2025-11-10 00:34:59,065][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 41 and human policies 1. -[2025-11-10 00:34:59,066][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:35:27,943][__main__][INFO] - Number of regex retries in iteration 420: 0 -[2025-11-10 00:35:27,944][__main__][INFO] - agents played in iteration 420 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:35:28,781][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:35:28,805][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:35:28,828][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:35:28,851][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:35:28,852][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:35:28,852][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:35:29,644][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:35:30,108][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:35:30,618][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:35:31,136][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:35:31,643][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:35:32,154][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:35:32,660][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:35:33,166][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:35:33,671][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:35:34,176][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:35:34,674][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:35:35,175][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:35:35,674][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:35:36,179][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:35:36,678][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:35:37,185][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:35:37,688][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:35:38,188][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:35:38,688][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:35:39,189][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:35:39,685][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:35:40,183][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:35:40,684][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:35:41,180][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:35:41,679][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:35:42,175][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:35:42,670][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:35:43,166][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:35:43,664][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:35:44,160][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:35:44,657][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:35:45,163][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:35:45,658][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:35:46,153][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:35:46,652][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:35:47,148][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:35:47,645][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:35:48,144][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:35:48,640][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:35:49,156][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:35:49,654][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:35:50,154][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:35:50,655][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:35:51,154][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:35:51,651][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:35:52,149][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:35:52,647][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:35:53,163][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:35:53,664][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:35:54,164][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:35:54,661][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:35:55,158][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:35:55,662][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:35:56,159][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:35:56,656][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:35:57,170][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:35:57,672][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:35:58,175][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:35:58,672][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:35:59,176][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:35:59,686][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:36:00,191][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:36:00,694][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:36:01,201][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:36:01,704][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10157 tokens. -[2025-11-10 00:36:02,521][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 55.64%, Block Peak % of device VRAM: 62.48%, ΔTime: 00:00:32 -[2025-11-10 00:36:03,280][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:36:03,282][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:36:03,284][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:36:05,042][__main__][INFO] - Iteration 421 took 1m 5s (43.77% Gen, 53.57% Train). Generation: 28s, Training: 35s. Estimated remaining time: 48h 21m 42s. Estimated total time: 54h 58m 50s. Time estimates for 10 more iterations: 10m 59s, 100 more iterations: 1h 49m 57s, 500 more iterations: 9h 9m 48s. -[2025-11-10 00:36:05,044][__main__][INFO] - Starting iteration 421. -[2025-11-10 00:36:05,508][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 42 and human policies 1. -[2025-11-10 00:36:05,509][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:36:33,615][__main__][INFO] - Number of regex retries in iteration 421: 0 -[2025-11-10 00:36:33,617][__main__][INFO] - agents played in iteration 421 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:36:34,459][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:36:34,483][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:36:34,506][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:36:34,529][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:36:34,530][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:36:34,530][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:36:35,301][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:36:35,760][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:36:36,266][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:36:36,777][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:36:37,286][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:36:37,788][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:36:38,288][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:36:38,786][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:36:39,290][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:36:39,793][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:36:40,291][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:36:40,794][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:36:41,301][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:36:41,803][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:36:42,305][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:36:42,807][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:36:43,311][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:36:43,815][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:36:44,313][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:36:44,819][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:36:45,321][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:36:45,820][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:36:46,320][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:36:46,816][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:36:47,318][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:36:47,818][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:36:48,318][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:36:48,827][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:36:49,331][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:36:49,844][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:36:50,340][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:36:50,841][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:36:51,361][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:36:51,858][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:36:52,356][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:36:52,855][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:36:53,351][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:36:53,845][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:36:54,341][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:36:54,834][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:36:55,343][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:36:55,840][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:36:56,342][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:36:56,842][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:36:57,338][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:36:57,851][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:36:58,348][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:36:58,846][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:36:59,350][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:36:59,848][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:37:00,347][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:37:00,847][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:37:01,347][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:37:01,867][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:37:02,368][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:37:02,874][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:37:03,375][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:37:03,875][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:37:04,378][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:37:04,876][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:37:05,380][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:37:05,879][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:37:06,381][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:37:06,883][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:37:07,386][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10148 tokens. -[2025-11-10 00:37:08,215][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.05%, Current % of VRAM taken: 55.54%, Block Peak % of device VRAM: 62.05%, ΔTime: 00:00:32 -[2025-11-10 00:37:08,848][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:37:08,851][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:37:08,853][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:37:09,666][__main__][INFO] - Iteration 422 took 1m 4s (43.81% Gen, 54.92% Train). Generation: 28s, Training: 35s. Estimated remaining time: 46h 49m 42s. Estimated total time: 53h 27m 55s. Time estimates for 10 more iterations: 10m 41s, 100 more iterations: 1h 46m 55s, 500 more iterations: 8h 54m 39s. -[2025-11-10 00:37:09,669][__main__][INFO] - Starting iteration 422. -[2025-11-10 00:37:10,138][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 42 and human policies 1. -[2025-11-10 00:37:10,139][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:37:31,851][__main__][INFO] - Number of regex retries in iteration 422: 0 -[2025-11-10 00:37:31,852][__main__][INFO] - agents played in iteration 422 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:37:32,625][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:37:32,647][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:37:32,673][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:37:32,695][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:37:32,696][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:37:32,697][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:37:33,435][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:37:33,892][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:37:34,394][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:37:34,890][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:37:35,396][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:37:35,898][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:37:36,398][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:37:36,900][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:37:37,408][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:37:37,912][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:37:38,419][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:37:38,922][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:37:39,425][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:37:39,933][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:37:40,432][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:37:40,934][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:37:41,437][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:37:41,939][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:37:42,450][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:37:42,973][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:37:43,480][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:37:43,988][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:37:44,492][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:37:44,997][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:37:45,499][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:37:46,001][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:37:46,506][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:37:47,011][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:37:47,511][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:37:48,017][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:37:48,516][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:37:49,016][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:37:49,516][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:37:50,016][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:37:50,871][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:37:51,603][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:37:52,111][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:37:52,611][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:37:53,119][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:37:53,621][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:37:54,120][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:37:54,620][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:37:55,120][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:37:55,623][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:37:56,128][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:37:56,629][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:37:57,130][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:37:57,631][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:37:58,128][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:37:58,629][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:37:59,128][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:37:59,630][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:38:00,132][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:38:00,632][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:38:01,131][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:38:01,629][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:38:02,134][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:38:02,632][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:38:03,128][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:38:03,629][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:38:04,133][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:38:04,634][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:38:05,133][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:38:05,632][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:38:06,132][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10252 tokens. -[2025-11-10 00:38:06,935][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 55.60%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:33 -[2025-11-10 00:38:07,604][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:38:07,606][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:38:07,607][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:38:08,428][__main__][INFO] - Iteration 423 took 58s (37.25% Gen, 61.34% Train). Generation: 21s, Training: 35s. Estimated remaining time: 41h 55m 19s. Estimated total time: 48h 34m 30s. Time estimates for 10 more iterations: 9m 42s, 100 more iterations: 1h 37m 9s, 500 more iterations: 8h 5m 45s. -[2025-11-10 00:38:08,430][__main__][INFO] - Starting iteration 423. -[2025-11-10 00:38:08,897][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 42 and human policies 1. -[2025-11-10 00:38:08,898][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:38:25,813][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 00:38:37,137][__main__][INFO] - Number of regex retries in iteration 423: 1 -[2025-11-10 00:38:37,138][__main__][INFO] - agents played in iteration 423 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:38:37,931][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:38:37,956][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:38:37,982][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:38:38,004][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:38:38,004][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:38:38,006][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:38:38,833][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:38:39,291][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:38:39,798][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:38:40,300][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:38:40,808][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:38:41,312][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:38:41,813][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:38:42,319][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:38:42,821][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:38:43,321][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:38:43,820][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:38:44,325][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:38:44,826][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:38:45,326][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:38:45,830][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:38:46,329][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:38:46,829][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:38:47,330][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:38:47,831][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:38:48,342][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:38:48,843][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:38:49,344][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:38:49,845][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:38:50,345][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:38:50,848][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:38:51,348][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:38:51,849][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:38:52,364][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:38:52,869][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:38:53,380][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:38:53,879][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:38:54,380][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:38:54,886][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:38:55,386][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:38:55,887][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:38:56,388][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:38:56,890][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:38:57,396][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:38:57,897][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:38:58,393][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:38:58,895][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:38:59,392][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:38:59,893][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:39:00,395][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:39:00,894][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:39:01,393][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:39:01,891][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:39:02,393][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:39:02,893][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:39:03,391][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:39:03,890][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:39:04,390][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:39:04,888][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:39:05,391][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:39:05,892][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:39:06,392][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:39:06,889][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:39:07,390][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:39:07,893][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:39:08,389][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:39:08,885][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:39:09,390][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:39:09,890][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:39:10,388][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:39:10,889][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10128 tokens. -[2025-11-10 00:39:11,689][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 55.56%, Block Peak % of device VRAM: 62.16%, ΔTime: 00:00:32 -[2025-11-10 00:39:12,463][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:39:12,465][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:39:12,466][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:39:13,396][__main__][INFO] - Iteration 424 took 1m 4s (43.78% Gen, 54.77% Train). Generation: 28s, Training: 35s. Estimated remaining time: 47h 4m 43s. Estimated total time: 53h 44m 59s. Time estimates for 10 more iterations: 10m 44s, 100 more iterations: 1h 47m 29s, 500 more iterations: 8h 57m 29s. -[2025-11-10 00:39:13,398][__main__][INFO] - Starting iteration 424. -[2025-11-10 00:39:13,884][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 42 and human policies 1. -[2025-11-10 00:39:13,885][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:39:37,182][__main__][INFO] - Number of regex retries in iteration 424: 0 -[2025-11-10 00:39:37,185][__main__][INFO] - agents played in iteration 424 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:39:38,039][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:39:38,064][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:39:38,088][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:39:38,123][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:39:38,124][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:39:38,125][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:39:38,933][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:39:39,395][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:39:39,905][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:39:40,406][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:39:40,909][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:39:41,410][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:39:41,920][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:39:42,426][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:39:42,926][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:39:43,427][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:39:43,930][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:39:44,431][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:39:44,934][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:39:45,440][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:39:45,941][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:39:46,441][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:39:46,943][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:39:47,443][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:39:47,944][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:39:48,444][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:39:48,948][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:39:49,452][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:39:49,952][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:39:50,452][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:39:50,950][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:39:51,451][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:39:51,952][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:39:52,453][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:39:52,967][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:39:53,469][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:39:53,986][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:39:54,487][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:39:54,989][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:39:55,492][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:39:55,992][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:39:56,494][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:39:56,994][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:39:57,495][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:39:58,011][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:39:58,516][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:39:59,025][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:39:59,527][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:40:00,026][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:40:00,530][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:40:01,030][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:40:01,531][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:40:02,036][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:40:02,535][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:40:03,038][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:40:03,542][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:40:04,044][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:40:04,555][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:40:05,058][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:40:05,558][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:40:06,058][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:40:06,557][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:40:07,066][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:40:07,570][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:40:08,074][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:40:08,573][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:40:09,070][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:40:09,573][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:40:10,070][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:40:10,567][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:40:11,085][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10104 tokens. -[2025-11-10 00:40:11,828][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.26%, Current % of VRAM taken: 55.76%, Block Peak % of device VRAM: 62.28%, ΔTime: 00:00:32 -[2025-11-10 00:40:12,474][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:40:12,475][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:40:12,477][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:40:13,275][__main__][INFO] - Iteration 425 took 59s (39.23% Gen, 59.42% Train). Generation: 23s, Training: 35s. Estimated remaining time: 42h 48m 19s. Estimated total time: 49h 29m 36s. Time estimates for 10 more iterations: 9m 53s, 100 more iterations: 1h 38m 59s, 500 more iterations: 8h 14m 56s. -[2025-11-10 00:40:13,277][__main__][INFO] - Starting iteration 425. -[2025-11-10 00:40:13,745][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 42 and human policies 1. -[2025-11-10 00:40:13,745][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:40:44,743][__main__][INFO] - Number of regex retries in iteration 425: 0 -[2025-11-10 00:40:44,744][__main__][INFO] - agents played in iteration 425 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:40:45,544][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:40:45,573][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:40:45,601][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:40:45,624][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:40:45,625][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:40:45,626][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:40:46,431][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:40:46,888][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:40:47,399][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:40:47,902][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:40:48,405][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:40:48,912][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:40:49,415][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:40:49,915][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:40:50,417][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:40:50,917][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:40:51,419][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:40:51,920][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:40:52,427][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:40:52,936][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:40:53,439][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:40:53,938][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:40:54,438][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:40:54,933][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:40:55,430][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:40:55,929][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:40:56,433][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:40:56,936][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:40:57,435][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:40:57,936][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:40:58,438][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:40:58,937][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:40:59,447][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:40:59,948][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:41:00,451][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:41:00,953][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:41:01,458][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:41:01,959][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:41:02,464][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:41:02,963][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:41:03,465][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:41:03,966][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:41:04,469][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:41:04,975][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:41:05,474][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:41:05,977][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:41:06,485][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:41:06,991][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:41:08,106][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:41:09,432][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:41:09,938][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:41:10,442][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:41:10,944][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:41:11,447][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:41:11,949][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:41:12,461][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:41:12,967][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:41:13,466][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:41:13,978][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:41:14,480][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:41:14,980][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:41:15,483][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:41:15,985][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:41:16,494][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:41:16,994][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:41:17,500][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:41:18,001][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:41:18,502][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:41:19,033][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:41:19,538][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:41:20,039][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10221 tokens. -[2025-11-10 00:41:20,836][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 55.71%, Block Peak % of device VRAM: 62.42%, ΔTime: 00:00:34 -[2025-11-10 00:41:21,474][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:41:21,476][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:41:21,477][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:41:22,303][__main__][INFO] - Iteration 426 took 1m 8s (45.21% Gen, 53.58% Train). Generation: 30s, Training: 36s. Estimated remaining time: 50h 25m 32s. Estimated total time: 57h 7m 57s. Time estimates for 10 more iterations: 11m 25s, 100 more iterations: 1h 54m 15s, 500 more iterations: 9h 31m 19s. -[2025-11-10 00:41:22,306][__main__][INFO] - Starting iteration 426. -[2025-11-10 00:41:22,777][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 42 and human policies 1. -[2025-11-10 00:41:22,777][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:41:42,564][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 00:41:53,538][__main__][INFO] - Number of regex retries in iteration 426: 1 -[2025-11-10 00:41:53,538][__main__][INFO] - agents played in iteration 426 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:41:54,349][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:41:54,377][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:41:54,403][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:41:54,427][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:41:54,427][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:41:54,429][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:41:55,247][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:41:55,707][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:41:56,215][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:41:56,719][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:41:57,244][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:41:57,747][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:41:58,251][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:41:58,758][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:41:59,262][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:41:59,767][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:42:00,269][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:42:00,775][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:42:01,280][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:42:01,781][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:42:02,288][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:42:02,793][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:42:03,295][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:42:03,803][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:42:04,311][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:42:04,827][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:42:05,342][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:42:05,849][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:42:06,357][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:42:06,862][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:42:07,367][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:42:07,872][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:42:08,374][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:42:08,874][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:42:09,376][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:42:09,879][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:42:10,382][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:42:10,886][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:42:11,392][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:42:11,894][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:42:12,399][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:42:12,913][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:42:13,420][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:42:13,921][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:42:14,436][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:42:14,939][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:42:15,441][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:42:15,948][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:42:16,449][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:42:16,954][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:42:17,459][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:42:17,960][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:42:18,462][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:42:18,963][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:42:19,464][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:42:19,968][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:42:20,467][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:42:20,969][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:42:21,470][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:42:21,969][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:42:22,473][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:42:22,972][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:42:23,469][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:42:23,969][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:42:24,470][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:42:24,969][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:42:25,469][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:42:25,969][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:42:26,469][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:42:26,967][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:42:27,468][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10145 tokens. -[2025-11-10 00:42:28,271][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 55.67%, Block Peak % of device VRAM: 62.48%, ΔTime: 00:00:33 -[2025-11-10 00:42:29,021][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:42:29,023][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:42:29,025][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:42:29,960][__main__][INFO] - Iteration 427 took 1m 7s (45.78% Gen, 52.82% Train). Generation: 30s, Training: 35s. Estimated remaining time: 49h 15m 38s. Estimated total time: 55h 59m 11s. Time estimates for 10 more iterations: 11m 11s, 100 more iterations: 1h 51m 58s, 500 more iterations: 9h 19m 51s. -[2025-11-10 00:42:29,962][__main__][INFO] - Starting iteration 427. -[2025-11-10 00:42:30,429][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 42 and human policies 1. -[2025-11-10 00:42:30,429][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:42:42,445][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 00:42:55,518][__main__][INFO] - Number of regex retries in iteration 427: 1 -[2025-11-10 00:42:55,521][__main__][INFO] - agents played in iteration 427 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:42:56,351][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:42:56,396][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:42:56,428][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:42:56,453][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:42:56,453][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:42:56,454][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:42:57,226][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:42:57,683][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:42:58,185][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:42:58,688][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:42:59,188][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:42:59,685][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:43:00,184][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:43:00,683][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:43:01,179][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:43:01,677][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:43:02,176][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:43:02,678][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:43:03,178][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:43:03,682][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:43:04,181][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:43:04,686][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:43:05,187][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:43:05,692][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:43:06,191][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:43:06,690][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:43:07,198][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:43:07,704][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:43:08,210][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:43:08,726][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:43:09,229][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:43:09,743][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:43:10,243][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:43:10,745][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:43:11,265][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:43:11,766][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:43:12,272][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:43:12,774][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:43:13,276][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:43:13,780][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:43:14,283][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:43:14,786][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:43:15,291][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:43:15,796][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:43:16,302][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:43:16,805][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:43:17,311][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:43:17,843][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:43:18,369][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:43:18,880][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:43:19,383][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:43:19,885][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:43:20,393][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:43:20,900][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:43:21,406][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:43:21,914][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:43:22,415][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:43:22,939][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:43:23,439][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:43:23,941][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:43:24,444][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:43:24,949][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:43:25,455][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:43:25,957][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:43:26,459][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:43:26,968][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:43:27,474][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:43:27,981][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:43:28,488][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:43:28,989][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:43:29,494][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10207 tokens. -[2025-11-10 00:43:30,347][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 55.74%, Block Peak % of device VRAM: 62.36%, ΔTime: 00:00:33 -[2025-11-10 00:43:31,013][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:43:31,014][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:43:31,016][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:43:31,834][__main__][INFO] - Iteration 428 took 1m 1s (40.86% Gen, 57.81% Train). Generation: 25s, Training: 35s. Estimated remaining time: 44h 25m 41s. Estimated total time: 51h 10m 16s. Time estimates for 10 more iterations: 10m 14s, 100 more iterations: 1h 42m 20s, 500 more iterations: 8h 31m 42s. -[2025-11-10 00:43:31,836][__main__][INFO] - Starting iteration 428. -[2025-11-10 00:43:32,300][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 42 and human policies 1. -[2025-11-10 00:43:32,300][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:44:03,518][__main__][INFO] - Number of regex retries in iteration 428: 0 -[2025-11-10 00:44:03,519][__main__][INFO] - agents played in iteration 428 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:44:04,305][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:44:04,333][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:44:04,360][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:44:04,383][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:44:04,384][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:44:04,385][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:44:05,144][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:44:05,600][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:44:06,102][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:44:06,606][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:44:07,107][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:44:07,608][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:44:08,106][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:44:08,604][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:44:09,104][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:44:09,603][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:44:10,099][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:44:10,600][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:44:11,098][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:44:11,600][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:44:12,099][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:44:12,597][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:44:13,096][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:44:13,598][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:44:14,101][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:44:14,601][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:44:15,100][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:44:15,603][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:44:16,103][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:44:16,610][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:44:17,118][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:44:17,617][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:44:18,115][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:44:18,614][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:44:19,112][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:44:19,610][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:44:20,108][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:44:20,613][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:44:21,115][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:44:21,624][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:44:22,137][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:44:22,661][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:44:23,165][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:44:23,679][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:44:24,187][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:44:24,693][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:44:25,200][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:44:25,704][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:44:26,207][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:44:26,710][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:44:27,212][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:44:27,717][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:44:28,221][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:44:28,729][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:44:29,248][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:44:31,044][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:44:31,575][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:44:32,079][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:44:32,586][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:44:33,091][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:44:33,595][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:44:34,101][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:44:34,606][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:44:35,113][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:44:35,622][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:44:36,123][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:44:36,630][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:44:37,135][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:44:37,648][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:44:38,159][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:44:38,659][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10114 tokens. -[2025-11-10 00:44:39,488][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 55.60%, Block Peak % of device VRAM: 62.35%, ΔTime: 00:00:34 -[2025-11-10 00:44:40,134][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:44:40,136][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:44:40,138][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:44:41,068][__main__][INFO] - Iteration 429 took 1m 8s (45.40% Gen, 53.25% Train). Generation: 31s, Training: 36s. Estimated remaining time: 50h 32m 43s. Estimated total time: 57h 18m 27s. Time estimates for 10 more iterations: 11m 27s, 100 more iterations: 1h 54m 36s, 500 more iterations: 9h 33m 4s. -[2025-11-10 00:44:41,070][__main__][INFO] - Starting iteration 429. -[2025-11-10 00:44:41,541][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 42 and human policies 1. -[2025-11-10 00:44:41,541][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:45:12,767][__main__][INFO] - Number of regex retries in iteration 429: 0 -[2025-11-10 00:45:12,767][__main__][INFO] - agents played in iteration 429 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:45:13,586][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:45:13,609][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:45:13,631][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:45:13,654][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:45:13,654][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:45:13,655][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:45:14,379][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:45:14,836][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:45:15,341][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:45:15,843][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:45:16,349][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:45:16,846][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:45:17,343][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:45:17,839][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:45:18,339][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:45:18,838][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:45:19,336][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:45:19,840][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:45:20,341][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:45:20,844][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:45:21,344][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:45:21,844][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:45:22,346][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:45:22,856][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:45:23,353][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:45:23,858][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:45:24,355][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:45:24,856][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:45:25,361][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:45:25,862][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:45:26,364][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:45:26,866][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:45:27,361][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:45:27,864][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:45:28,363][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:45:28,863][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:45:29,363][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:45:29,862][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:45:30,358][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:45:30,858][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:45:31,364][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:45:31,880][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:45:32,380][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:45:32,887][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:45:33,388][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:45:33,891][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:45:34,399][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:45:34,924][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:45:35,431][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:45:35,936][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:45:36,439][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:45:36,944][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:45:37,449][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:45:37,954][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:45:38,456][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:45:38,958][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:45:39,460][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:45:39,968][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:45:40,468][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:45:40,985][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:45:41,486][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:45:41,988][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:45:42,493][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:45:43,000][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:45:43,499][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:45:44,001][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:45:44,502][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:45:45,002][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:45:45,509][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:45:46,010][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:45:46,516][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10167 tokens. -[2025-11-10 00:45:47,363][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 55.62%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 -[2025-11-10 00:45:48,132][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:45:48,134][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:45:48,135][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:45:49,026][__main__][INFO] - Iteration 430 took 1m 7s (46.27% Gen, 52.41% Train). Generation: 31s, Training: 35s. Estimated remaining time: 49h 27m 25s. Estimated total time: 56h 14m 17s. Time estimates for 10 more iterations: 11m 14s, 100 more iterations: 1h 52m 28s, 500 more iterations: 9h 22m 22s. -[2025-11-10 00:45:49,028][__main__][INFO] - Starting iteration 430. -[2025-11-10 00:45:49,511][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 42 and human policies 1. -[2025-11-10 00:45:49,511][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:46:03,522][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 00:46:13,894][__main__][INFO] - Number of regex retries in iteration 430: 1 -[2025-11-10 00:46:13,895][__main__][INFO] - agents played in iteration 430 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:46:14,693][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:46:14,720][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:46:14,748][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:46:14,772][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:46:14,772][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:46:14,773][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:46:15,615][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:46:16,079][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:46:16,604][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:46:17,108][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:46:17,619][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:46:18,121][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:46:18,624][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:46:19,128][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:46:19,630][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:46:20,133][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:46:20,635][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:46:21,138][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:46:21,639][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:46:22,143][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:46:22,645][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:46:23,149][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:46:23,648][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:46:24,150][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:46:24,649][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:46:25,148][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:46:25,653][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:46:26,154][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:46:26,653][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:46:27,153][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:46:27,652][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:46:28,152][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:46:28,655][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:46:29,156][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:46:29,667][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:46:30,167][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:46:30,669][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:46:31,172][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:46:31,672][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:46:32,182][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:46:32,688][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:46:33,194][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:46:33,695][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:46:34,197][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:46:34,705][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:46:35,209][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:46:35,712][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:46:36,209][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:46:36,709][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:46:37,211][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:46:37,715][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:46:38,214][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:46:38,721][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:46:39,228][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:46:39,731][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:46:40,235][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:46:40,735][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:46:41,239][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:46:41,739][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:46:42,240][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:46:42,744][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:46:43,245][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:46:43,746][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:46:44,252][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:46:44,757][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:46:45,260][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:46:45,768][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:46:46,269][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:46:46,773][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:46:47,277][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:46:47,782][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10096 tokens. -[2025-11-10 00:46:48,599][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 55.62%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 -[2025-11-10 00:46:49,268][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:46:49,271][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:46:49,274][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:46:51,047][__main__][INFO] - Iteration 431 took 1m 1s (39.62% Gen, 57.49% Train). Generation: 24s, Training: 35s. Estimated remaining time: 44h 28m 58s. Estimated total time: 51h 16m 52s. Time estimates for 10 more iterations: 10m 15s, 100 more iterations: 1h 42m 33s, 500 more iterations: 8h 32m 48s. -[2025-11-10 00:46:51,050][__main__][INFO] - Starting iteration 431. -[2025-11-10 00:46:51,514][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 43 and human policies 1. -[2025-11-10 00:46:51,515][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:47:09,812][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 00:47:20,795][__main__][INFO] - Number of regex retries in iteration 431: 1 -[2025-11-10 00:47:20,795][__main__][INFO] - agents played in iteration 431 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:47:21,626][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:47:21,652][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:47:21,676][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:47:21,700][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:47:21,701][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:47:21,702][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:47:22,540][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:47:22,997][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:47:23,519][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:47:24,018][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:47:24,517][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:47:25,017][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:47:25,517][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:47:26,019][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:47:26,517][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:47:27,028][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:47:27,531][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:47:28,033][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:47:28,535][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:47:29,045][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:47:29,550][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:47:30,054][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:47:30,557][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:47:31,057][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:47:31,561][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:47:32,067][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:47:32,568][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:47:33,070][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:47:33,571][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:47:34,073][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:47:34,573][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:47:35,075][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:47:35,578][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:47:36,081][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:47:36,580][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:47:37,084][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:47:37,591][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:47:38,097][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:47:38,603][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:47:39,105][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:47:39,610][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:47:40,110][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:47:40,624][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:47:41,123][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:47:41,623][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:47:42,125][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:47:42,623][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:47:43,127][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:47:43,629][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:47:44,130][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:47:44,640][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:47:45,139][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:47:45,642][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:47:46,143][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:47:46,641][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:47:47,145][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:47:47,643][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:47:48,140][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:47:48,638][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:47:49,136][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:47:49,639][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:47:50,135][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:47:50,634][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:47:51,140][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:47:51,640][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:47:52,145][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:47:52,650][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:47:53,154][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:47:53,658][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:47:54,160][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:47:54,658][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10154 tokens. -[2025-11-10 00:47:56,337][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.03%, Current % of VRAM taken: 55.53%, Block Peak % of device VRAM: 62.49%, ΔTime: 00:00:33 -[2025-11-10 00:47:57,976][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:47:57,978][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:47:57,981][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:47:58,841][__main__][INFO] - Iteration 432 took 1m 7s (43.49% Gen, 55.23% Train). Generation: 29s, Training: 37s. Estimated remaining time: 49h 17m 21s. Estimated total time: 56h 6m 23s. Time estimates for 10 more iterations: 11m 13s, 100 more iterations: 1h 52m 12s, 500 more iterations: 9h 21m 3s. -[2025-11-10 00:47:58,844][__main__][INFO] - Starting iteration 432. -[2025-11-10 00:47:59,346][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 43 and human policies 1. -[2025-11-10 00:47:59,347][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:48:28,722][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 00:48:32,232][__main__][INFO] - Number of regex retries in iteration 432: 1 -[2025-11-10 00:48:32,232][__main__][INFO] - agents played in iteration 432 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:48:33,051][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:48:33,077][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:48:33,102][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:48:33,124][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:48:33,125][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:48:33,125][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:48:33,916][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:48:34,374][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:48:34,883][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:48:35,383][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:48:35,884][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:48:36,390][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:48:36,895][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:48:37,399][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:48:37,901][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:48:38,417][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:48:38,927][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:48:39,444][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:48:39,950][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:48:40,454][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:48:40,972][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:48:41,477][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:48:41,981][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:48:42,479][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:48:42,985][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:48:43,488][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:48:43,991][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:48:44,493][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:48:44,993][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:48:45,495][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:48:45,997][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:48:46,504][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:48:47,005][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:48:47,518][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:48:48,023][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:48:48,537][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:48:49,038][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:48:49,542][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:48:50,047][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:48:50,548][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:48:51,055][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:48:51,560][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:48:52,062][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:48:52,572][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:48:53,076][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:48:53,577][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:48:54,083][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:48:54,585][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:48:55,090][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:48:55,590][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:48:56,092][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:48:56,598][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:48:57,099][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:48:57,606][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:48:58,109][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:48:58,612][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:48:59,123][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:48:59,627][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:49:00,132][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:49:00,635][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:49:01,138][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:49:01,650][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:49:02,153][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:49:02,653][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:49:03,162][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:49:03,663][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:49:04,163][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:49:04,668][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:49:05,172][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:49:05,682][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:49:06,183][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10128 tokens. -[2025-11-10 00:49:07,046][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.05%, Current % of VRAM taken: 55.55%, Block Peak % of device VRAM: 62.20%, ΔTime: 00:00:33 -[2025-11-10 00:49:07,846][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:49:07,848][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:49:07,850][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:49:08,855][__main__][INFO] - Iteration 433 took 1m 9s (47.31% Gen, 51.24% Train). Generation: 32s, Training: 35s. Estimated remaining time: 51h 5m 16s. Estimated total time: 57h 55m 28s. Time estimates for 10 more iterations: 11m 35s, 100 more iterations: 1h 55m 50s, 500 more iterations: 9h 39m 14s. -[2025-11-10 00:49:08,857][__main__][INFO] - Starting iteration 433. -[2025-11-10 00:49:09,326][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 43 and human policies 1. -[2025-11-10 00:49:09,327][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:49:36,353][__main__][INFO] - Number of regex retries in iteration 433: 0 -[2025-11-10 00:49:36,356][__main__][INFO] - agents played in iteration 433 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:49:37,158][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:49:37,182][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:49:37,207][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:49:37,229][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:49:37,230][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:49:37,231][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:49:38,033][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:49:38,498][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:49:39,011][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:49:39,513][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:49:40,014][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:49:40,519][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:49:41,021][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:49:41,522][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:49:42,024][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:49:42,526][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:49:43,025][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:49:43,536][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:49:44,036][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:49:44,535][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:49:45,038][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:49:45,535][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:49:46,030][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:49:46,531][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:49:47,031][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:49:47,541][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:49:48,041][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:49:48,537][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:49:49,040][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:49:49,537][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:49:50,051][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:49:50,552][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:49:51,049][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:49:51,562][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:49:52,059][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:49:52,560][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:49:53,060][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:49:53,558][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:49:54,058][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:49:54,558][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:49:55,056][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:49:55,556][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:49:56,057][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:49:56,553][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:49:57,051][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:49:57,557][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:49:58,062][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:49:58,562][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:49:59,057][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:49:59,554][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:50:00,053][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:50:00,550][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:50:01,048][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:50:01,549][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:50:02,052][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:50:02,553][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:50:03,056][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:50:03,556][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:50:04,061][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:50:04,569][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:50:05,073][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:50:05,576][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:50:06,079][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:50:06,589][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:50:07,094][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:50:07,595][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:50:08,100][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:50:08,609][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:50:09,115][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:50:09,622][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:50:10,135][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10242 tokens. -[2025-11-10 00:50:11,002][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 55.71%, Block Peak % of device VRAM: 62.38%, ΔTime: 00:00:32 -[2025-11-10 00:50:11,636][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:50:11,639][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:50:11,640][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:50:12,474][__main__][INFO] - Iteration 434 took 1m 3s (42.80% Gen, 55.88% Train). Generation: 27s, Training: 35s. Estimated remaining time: 45h 46m 8s. Estimated total time: 52h 37m 24s. Time estimates for 10 more iterations: 10m 31s, 100 more iterations: 1h 45m 14s, 500 more iterations: 8h 46m 14s. -[2025-11-10 00:50:12,476][__main__][INFO] - Starting iteration 434. -[2025-11-10 00:50:12,946][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 43 and human policies 1. -[2025-11-10 00:50:12,947][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:50:28,663][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 00:50:34,432][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 00:50:40,361][__main__][INFO] - Number of regex retries in iteration 434: 2 -[2025-11-10 00:50:40,362][__main__][INFO] - agents played in iteration 434 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:50:41,159][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:50:41,186][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:50:41,224][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:50:41,248][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:50:41,249][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:50:41,250][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:50:42,072][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:50:42,526][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:50:43,036][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:50:43,538][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:50:44,044][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:50:44,545][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:50:45,045][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:50:45,547][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:50:46,047][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:50:46,550][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:50:47,048][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:50:47,550][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:50:48,056][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:50:48,554][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:50:49,052][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:50:49,564][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:50:50,062][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:50:50,576][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:50:51,079][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:50:51,586][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:50:52,091][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:50:52,590][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:50:53,093][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:50:53,594][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:50:54,099][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:50:54,607][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:50:55,135][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:50:55,636][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:50:56,137][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:50:56,645][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:50:57,147][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:50:57,648][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:50:58,147][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:50:58,672][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:50:59,172][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:50:59,672][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:51:00,174][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:51:00,677][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:51:01,183][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:51:01,684][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:51:02,185][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:51:02,691][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:51:03,191][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:51:03,689][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:51:04,189][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:51:04,687][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:51:05,189][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:51:05,685][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:51:06,183][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:51:06,694][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:51:07,196][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:51:07,694][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:51:08,195][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:51:08,693][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:51:09,205][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:51:09,709][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:51:10,211][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:51:10,724][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:51:11,227][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:51:11,734][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:51:12,236][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:51:12,739][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:51:14,163][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:51:14,687][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:51:15,194][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10182 tokens. -[2025-11-10 00:51:16,079][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 55.67%, Block Peak % of device VRAM: 62.27%, ΔTime: 00:00:34 -[2025-11-10 00:51:16,741][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:51:16,743][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:51:16,745][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:51:17,617][__main__][INFO] - Iteration 435 took 1m 4s (42.39% Gen, 56.26% Train). Generation: 27s, Training: 36s. Estimated remaining time: 47h 1m 14s. Estimated total time: 53h 53m 34s. Time estimates for 10 more iterations: 10m 46s, 100 more iterations: 1h 47m 47s, 500 more iterations: 8h 58m 55s. -[2025-11-10 00:51:17,620][__main__][INFO] - Starting iteration 435. -[2025-11-10 00:51:18,083][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 43 and human policies 1. -[2025-11-10 00:51:18,083][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:51:52,668][__main__][INFO] - Number of regex retries in iteration 435: 0 -[2025-11-10 00:51:52,669][__main__][INFO] - agents played in iteration 435 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:51:53,457][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:51:53,485][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:51:53,512][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:51:53,535][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:51:53,536][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:51:53,536][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:51:54,291][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:51:54,750][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:51:55,274][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:51:55,779][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:51:56,284][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:51:56,789][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:51:57,294][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:51:57,804][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:51:58,307][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:51:58,816][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:51:59,317][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:51:59,822][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:52:00,329][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:52:00,833][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:52:01,337][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:52:01,844][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:52:02,349][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:52:02,850][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:52:03,350][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:52:03,854][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:52:04,354][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:52:04,858][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:52:05,362][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:52:05,864][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:52:06,364][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:52:06,865][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:52:07,366][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:52:07,868][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:52:08,389][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:52:08,888][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:52:09,387][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:52:09,886][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:52:10,386][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:52:10,894][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:52:11,396][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:52:11,898][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:52:12,401][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:52:12,900][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:52:13,399][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:52:13,902][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:52:14,406][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:52:14,912][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:52:15,412][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:52:15,915][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:52:16,416][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:52:16,915][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:52:17,419][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:52:17,923][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:52:18,425][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:52:18,926][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:52:19,427][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:52:19,931][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:52:20,433][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:52:20,933][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:52:21,435][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:52:21,937][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:52:22,439][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:52:22,947][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:52:23,448][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:52:23,952][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:52:24,457][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:52:24,958][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:52:25,476][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:52:25,979][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:52:26,488][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10343 tokens. -[2025-11-10 00:52:27,294][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.34%, ΔTime: 00:00:33 -[2025-11-10 00:52:28,085][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:52:28,087][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:52:28,089][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:52:29,019][__main__][INFO] - Iteration 436 took 1m 10s (48.75% Gen, 49.93% Train). Generation: 34s, Training: 35s. Estimated remaining time: 52h 13m 18s. Estimated total time: 59h 6m 50s. Time estimates for 10 more iterations: 11m 49s, 100 more iterations: 1h 58m 13s, 500 more iterations: 9h 51m 8s. -[2025-11-10 00:52:29,021][__main__][INFO] - Starting iteration 436. -[2025-11-10 00:52:29,490][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 43 and human policies 1. -[2025-11-10 00:52:29,490][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:52:52,186][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 00:52:57,402][__main__][INFO] - Number of regex retries in iteration 436: 1 -[2025-11-10 00:52:57,403][__main__][INFO] - agents played in iteration 436 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:52:58,208][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:52:58,235][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:52:58,261][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:52:58,285][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:52:58,286][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:52:58,286][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:52:59,057][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:52:59,518][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:53:00,029][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:53:00,531][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:53:01,035][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:53:01,536][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:53:02,040][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:53:02,541][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:53:03,043][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:53:03,552][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:53:04,054][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:53:04,556][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:53:05,061][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:53:05,564][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:53:06,065][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:53:06,569][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:53:07,070][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:53:07,573][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:53:08,077][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:53:08,578][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:53:09,078][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:53:09,578][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:53:10,079][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:53:10,581][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:53:11,083][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:53:11,588][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:53:12,090][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:53:12,593][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:53:13,094][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:53:13,594][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:53:14,123][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:53:14,628][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:53:15,126][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:53:15,630][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:53:16,130][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:53:16,638][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:53:17,135][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:53:17,634][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:53:18,141][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:53:18,640][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:53:19,141][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:53:19,645][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:53:20,143][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:53:20,648][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:53:21,147][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:53:21,651][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:53:22,154][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:53:22,654][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:53:23,153][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:53:23,652][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:53:24,153][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:53:24,654][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:53:25,154][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:53:25,658][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:53:26,164][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:53:26,662][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:53:27,160][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:53:27,665][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:53:28,173][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:53:28,677][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:53:29,177][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:53:29,676][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:53:30,177][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:53:30,678][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:53:31,180][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10091 tokens. -[2025-11-10 00:53:32,052][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 55.66%, Block Peak % of device VRAM: 62.23%, ΔTime: 00:00:33 -[2025-11-10 00:53:32,706][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:53:32,708][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:53:32,710][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:53:33,544][__main__][INFO] - Iteration 437 took 1m 4s (43.58% Gen, 55.12% Train). Generation: 27s, Training: 35s. Estimated remaining time: 46h 28m 9s. Estimated total time: 53h 22m 45s. Time estimates for 10 more iterations: 10m 40s, 100 more iterations: 1h 46m 45s, 500 more iterations: 8h 53m 47s. -[2025-11-10 00:53:33,547][__main__][INFO] - Starting iteration 437. -[2025-11-10 00:53:34,017][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 43 and human policies 1. -[2025-11-10 00:53:34,018][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:54:01,145][__main__][INFO] - Number of regex retries in iteration 437: 0 -[2025-11-10 00:54:01,145][__main__][INFO] - agents played in iteration 437 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:54:01,929][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:54:01,957][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:54:01,983][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:54:02,006][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:54:02,007][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:54:02,008][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:54:02,760][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:54:03,220][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:54:03,723][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:54:04,223][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:54:04,722][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:54:05,243][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:54:05,744][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:54:06,244][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:54:06,744][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:54:07,249][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:54:07,751][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:54:08,251][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:54:08,751][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:54:09,255][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:54:09,759][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:54:10,260][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:54:10,758][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:54:11,257][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:54:11,762][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:54:12,262][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:54:12,768][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:54:13,268][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:54:13,768][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:54:14,268][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:54:14,767][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:54:15,264][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:54:15,760][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:54:16,256][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:54:16,753][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:54:17,257][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:54:17,753][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:54:18,253][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:54:18,750][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:54:19,247][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:54:19,744][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:54:20,242][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:54:20,739][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:54:21,237][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:54:21,733][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:54:22,239][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:54:22,742][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:54:23,239][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:54:23,742][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:54:24,242][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:54:24,740][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:54:25,238][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:54:25,735][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:54:26,244][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:54:26,742][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:54:27,238][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:54:27,736][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:54:28,238][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:54:28,748][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:54:29,248][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:54:29,749][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:54:30,250][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:54:30,747][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:54:31,250][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:54:31,755][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:54:32,254][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:54:32,755][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:54:33,256][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:54:33,811][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:54:35,198][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:54:35,844][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10199 tokens. -[2025-11-10 00:54:36,741][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 55.65%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:33 -[2025-11-10 00:54:37,401][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:54:37,403][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:54:37,404][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:54:38,282][__main__][INFO] - Iteration 438 took 1m 4s (42.21% Gen, 56.42% Train). Generation: 27s, Training: 36s. Estimated remaining time: 46h 37m 33s. Estimated total time: 53h 33m 14s. Time estimates for 10 more iterations: 10m 42s, 100 more iterations: 1h 47m 6s, 500 more iterations: 8h 55m 32s. -[2025-11-10 00:54:38,284][__main__][INFO] - Starting iteration 438. -[2025-11-10 00:54:38,759][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 43 and human policies 1. -[2025-11-10 00:54:38,760][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:54:58,269][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 00:54:58,474][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 00:55:10,456][__main__][INFO] - Number of regex retries in iteration 438: 2 -[2025-11-10 00:55:10,457][__main__][INFO] - agents played in iteration 438 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:55:11,290][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:55:11,318][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:55:11,345][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:55:11,368][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:55:11,368][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:55:11,369][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:55:12,125][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:55:12,586][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:55:13,092][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:55:13,593][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:55:14,099][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:55:14,599][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:55:15,102][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:55:15,601][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:55:16,102][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:55:16,607][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:55:17,107][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:55:17,624][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:55:18,125][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:55:18,625][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:55:19,121][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:55:19,621][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:55:20,144][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:55:20,643][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:55:21,141][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:55:21,642][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:55:22,146][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:55:22,647][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:55:23,148][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:55:23,647][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:55:24,157][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:55:24,653][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:55:25,152][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:55:25,651][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:55:26,150][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:55:26,647][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:55:27,141][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:55:27,637][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:55:28,154][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:55:28,648][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:55:29,148][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:55:29,644][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:55:30,140][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:55:30,649][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:55:31,144][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:55:31,645][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:55:32,160][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:55:32,660][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:55:33,162][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:55:33,658][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:55:34,151][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:55:34,656][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:55:35,152][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:55:35,650][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:55:36,163][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:55:36,666][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:55:37,164][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:55:37,668][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:55:38,176][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:55:38,677][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:55:39,178][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:55:39,680][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:55:40,179][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:55:40,679][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:55:41,182][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:55:41,681][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:55:42,183][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:55:42,686][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:55:43,188][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:55:43,689][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:55:44,193][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10195 tokens. -[2025-11-10 00:55:45,024][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 55.62%, Block Peak % of device VRAM: 62.14%, ΔTime: 00:00:32 -[2025-11-10 00:55:45,785][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:55:45,787][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:55:45,788][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:55:46,721][__main__][INFO] - Iteration 439 took 1m 7s (46.64% Gen, 51.99% Train). Generation: 31s, Training: 35s. Estimated remaining time: 49h 41m 18s. Estimated total time: 56h 38m 7s. Time estimates for 10 more iterations: 11m 19s, 100 more iterations: 1h 53m 16s, 500 more iterations: 9h 26m 21s. -[2025-11-10 00:55:46,724][__main__][INFO] - Starting iteration 439. -[2025-11-10 00:55:47,193][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 43 and human policies 1. -[2025-11-10 00:55:47,194][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:56:10,353][__main__][INFO] - Number of regex retries in iteration 439: 0 -[2025-11-10 00:56:10,355][__main__][INFO] - agents played in iteration 439 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:56:11,161][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:56:11,186][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:56:11,210][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:56:11,234][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:56:11,234][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:56:11,235][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:56:12,072][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:56:12,537][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:56:13,043][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:56:13,550][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:56:14,056][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:56:14,562][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:56:15,067][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:56:15,573][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:56:16,072][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:56:16,573][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:56:17,072][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:56:17,570][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:56:18,069][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:56:18,571][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:56:19,072][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:56:19,575][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:56:20,074][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:56:20,577][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:56:21,085][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:56:21,587][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:56:22,088][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:56:22,591][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:56:23,090][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:56:23,592][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:56:24,092][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:56:24,596][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:56:25,095][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:56:25,596][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:56:26,095][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:56:26,593][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:56:27,096][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:56:27,595][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:56:28,096][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:56:28,594][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:56:29,093][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:56:29,597][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:56:30,100][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:56:30,600][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:56:31,102][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:56:31,600][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:56:32,104][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:56:32,601][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:56:33,098][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:56:33,595][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:56:34,097][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:56:34,593][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:56:35,094][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:56:35,592][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:56:36,091][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:56:36,593][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:56:37,092][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:56:37,594][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:56:38,094][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:56:38,597][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:56:39,096][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:56:39,596][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:56:40,100][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:56:40,600][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:56:41,099][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:56:41,600][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:56:42,100][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:56:42,603][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:56:43,110][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:56:43,609][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:56:44,113][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10118 tokens. -[2025-11-10 00:56:44,965][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 55.63%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:32 -[2025-11-10 00:56:45,601][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:56:45,603][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:56:45,605][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:56:46,423][__main__][INFO] - Iteration 440 took 59s (39.10% Gen, 59.51% Train). Generation: 23s, Training: 35s. Estimated remaining time: 42h 23m 41s. Estimated total time: 49h 21m 31s. Time estimates for 10 more iterations: 9m 52s, 100 more iterations: 1h 38m 43s, 500 more iterations: 8h 13m 35s. -[2025-11-10 00:56:46,425][__main__][INFO] - Starting iteration 440. -[2025-11-10 00:56:46,895][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 43 and human policies 1. -[2025-11-10 00:56:46,896][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:57:19,318][__main__][INFO] - Number of regex retries in iteration 440: 0 -[2025-11-10 00:57:19,319][__main__][INFO] - agents played in iteration 440 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:57:20,150][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:57:20,173][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:57:20,198][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:57:20,221][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:57:20,222][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:57:20,222][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:57:21,027][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:57:21,484][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:57:21,993][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:57:22,495][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:57:22,996][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:57:23,494][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:57:23,992][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:57:24,491][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:57:24,989][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:57:25,487][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:57:25,987][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:57:26,484][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:57:26,982][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:57:27,478][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:57:27,975][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:57:28,477][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:57:28,975][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:57:29,480][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:57:29,980][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:57:30,484][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:57:30,986][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:57:31,487][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:57:31,987][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:57:32,489][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:57:32,993][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:57:33,491][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:57:33,992][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:57:34,490][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:57:34,993][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:57:35,495][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:57:35,995][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:57:36,493][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:57:36,993][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:57:37,496][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:57:37,995][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:57:38,493][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:57:38,996][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:57:39,496][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:57:39,993][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:57:40,490][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:57:40,989][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:57:41,490][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:57:41,993][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:57:42,494][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:57:42,997][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:57:43,502][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:57:44,005][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:57:44,509][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:57:45,010][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:57:45,514][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:57:46,020][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:57:46,527][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:57:47,030][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:57:47,534][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:57:48,045][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:57:48,662][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:57:50,236][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:57:50,749][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:57:51,251][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:57:51,753][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:57:52,254][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:57:52,755][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:57:53,267][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:57:53,768][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:57:54,272][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10203 tokens. -[2025-11-10 00:57:55,149][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 55.56%, Block Peak % of device VRAM: 62.41%, ΔTime: 00:00:34 -[2025-11-10 00:57:55,827][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:57:55,829][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:57:55,832][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:57:57,530][__main__][INFO] - Iteration 441 took 1m 10s (45.90% Gen, 51.69% Train). Generation: 32s, Training: 36s. Estimated remaining time: 51h 52m 45s. Estimated total time: 58h 51m 46s. Time estimates for 10 more iterations: 11m 46s, 100 more iterations: 1h 57m 43s, 500 more iterations: 9h 48m 37s. -[2025-11-10 00:57:57,535][__main__][INFO] - Starting iteration 441. -[2025-11-10 00:57:58,009][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 44 and human policies 1. -[2025-11-10 00:57:58,010][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:58:28,884][__main__][INFO] - Number of regex retries in iteration 441: 0 -[2025-11-10 00:58:28,885][__main__][INFO] - agents played in iteration 441 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:58:29,694][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:58:29,717][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:58:29,743][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:58:29,765][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:58:29,766][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:58:29,767][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:58:30,537][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:58:30,991][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:58:31,498][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:58:31,999][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:58:32,499][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:58:32,998][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:58:33,497][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:58:33,998][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:58:34,502][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:58:35,002][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:58:35,503][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:58:36,004][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:58:36,512][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:58:37,011][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:58:37,512][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:58:38,013][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:58:38,513][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:58:39,014][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:58:39,516][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:58:40,015][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:58:40,513][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:58:41,011][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:58:41,511][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:58:42,009][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:58:42,507][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:58:43,014][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:58:43,511][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:58:44,014][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:58:44,514][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:58:45,012][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:58:45,513][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:58:46,011][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:58:46,510][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:58:47,010][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:58:47,509][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:58:48,039][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:58:48,551][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:58:49,054][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:58:49,552][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:58:50,060][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:58:50,563][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:58:51,079][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:58:51,582][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:58:52,081][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:58:52,590][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:58:53,092][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:58:53,593][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:58:54,094][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:58:54,597][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 00:58:55,102][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 00:58:55,602][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 00:58:56,102][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 00:58:56,606][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 00:58:57,109][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 00:58:57,609][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 00:58:58,110][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 00:58:58,610][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 00:58:59,114][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 00:58:59,613][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 00:59:00,113][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 00:59:00,616][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 00:59:01,118][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 00:59:01,623][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 00:59:02,130][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 00:59:02,631][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10113 tokens. -[2025-11-10 00:59:03,440][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.09%, Current % of VRAM taken: 55.59%, Block Peak % of device VRAM: 62.38%, ΔTime: 00:00:32 -[2025-11-10 00:59:04,183][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 00:59:04,185][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 00:59:04,186][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 00:59:05,102][__main__][INFO] - Iteration 442 took 1m 7s (46.02% Gen, 52.61% Train). Generation: 30s, Training: 35s. Estimated remaining time: 48h 54m 34s. Estimated total time: 55h 54m 42s. Time estimates for 10 more iterations: 11m 10s, 100 more iterations: 1h 51m 49s, 500 more iterations: 9h 19m 7s. -[2025-11-10 00:59:05,105][__main__][INFO] - Starting iteration 442. -[2025-11-10 00:59:05,583][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 44 and human policies 1. -[2025-11-10 00:59:05,584][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 00:59:22,357][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 00:59:33,833][__main__][INFO] - Number of regex retries in iteration 442: 1 -[2025-11-10 00:59:33,836][__main__][INFO] - agents played in iteration 442 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 00:59:34,661][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:59:34,696][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:59:34,724][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:59:34,748][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 00:59:34,748][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 00:59:34,749][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 00:59:35,552][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 00:59:36,014][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 00:59:36,521][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 00:59:37,020][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 00:59:37,523][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 00:59:38,029][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 00:59:38,530][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 00:59:39,035][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 00:59:39,536][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 00:59:40,044][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 00:59:40,545][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 00:59:41,050][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 00:59:41,558][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 00:59:42,059][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 00:59:42,559][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 00:59:43,064][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 00:59:43,567][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 00:59:44,078][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 00:59:44,575][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 00:59:45,080][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 00:59:45,586][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 00:59:46,090][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 00:59:46,601][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 00:59:47,102][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 00:59:47,604][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 00:59:48,144][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 00:59:48,645][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 00:59:49,150][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 00:59:49,648][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 00:59:50,154][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 00:59:50,655][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 00:59:51,157][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 00:59:51,661][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 00:59:52,164][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 00:59:52,662][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 00:59:53,187][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 00:59:53,691][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 00:59:54,194][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 00:59:54,697][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 00:59:55,201][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 00:59:55,709][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 00:59:56,215][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 00:59:56,714][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 00:59:57,221][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 00:59:57,725][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 00:59:58,225][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 00:59:58,728][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 00:59:59,228][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 00:59:59,728][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:00:00,233][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:00:00,735][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:00:01,238][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:00:01,737][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:00:02,243][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:00:02,748][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:00:03,253][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:00:03,754][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:00:04,257][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:00:04,759][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:00:05,263][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:00:05,766][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:00:06,268][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:00:06,788][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:00:07,292][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:00:07,806][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10193 tokens. -[2025-11-10 01:00:08,643][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 55.71%, Block Peak % of device VRAM: 62.18%, ΔTime: 00:00:33 -[2025-11-10 01:00:09,307][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:00:09,309][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:00:09,311][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:00:10,144][__main__][INFO] - Iteration 443 took 1m 4s (43.76% Gen, 54.95% Train). Generation: 28s, Training: 35s. Estimated remaining time: 46h 46m 52s. Estimated total time: 53h 48m 5s. Time estimates for 10 more iterations: 10m 45s, 100 more iterations: 1h 47m 36s, 500 more iterations: 8h 58m 0s. -[2025-11-10 01:00:10,147][__main__][INFO] - Starting iteration 443. -[2025-11-10 01:00:10,614][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 44 and human policies 1. -[2025-11-10 01:00:10,615][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:00:43,593][__main__][INFO] - Number of regex retries in iteration 443: 0 -[2025-11-10 01:00:43,594][__main__][INFO] - agents played in iteration 443 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:00:44,403][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:00:44,430][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:00:44,465][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:00:44,488][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:00:44,489][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:00:44,490][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:00:45,202][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:00:45,658][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:00:46,162][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:00:46,661][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:00:47,161][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:00:47,656][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:00:48,153][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:00:48,650][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:00:49,145][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:00:49,647][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:00:50,143][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:00:50,642][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:00:51,146][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:00:51,648][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:00:52,149][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:00:52,648][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:00:53,147][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:00:53,648][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:00:54,143][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:00:54,640][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:00:55,145][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:00:55,644][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:00:56,139][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:00:56,633][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:00:57,127][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:00:57,635][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:00:58,133][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:00:58,634][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:00:59,131][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:00:59,629][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:01:00,131][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:01:00,654][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:01:01,152][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:01:01,652][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:01:02,151][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:01:02,653][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:01:03,153][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:01:03,660][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:01:04,163][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:01:04,665][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:01:05,166][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:01:05,667][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:01:06,170][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:01:06,670][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:01:07,172][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:01:07,671][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:01:08,176][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:01:08,678][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:01:09,178][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:01:09,685][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:01:10,189][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:01:10,693][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:01:11,195][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:01:11,696][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:01:12,202][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:01:12,703][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:01:13,203][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:01:13,704][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:01:14,207][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:01:14,713][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:01:15,217][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:01:15,719][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:01:16,223][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:01:16,727][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:01:17,236][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10189 tokens. -[2025-11-10 01:01:18,048][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 55.70%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 -[2025-11-10 01:01:18,790][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:01:18,792][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:01:18,793][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:01:19,717][__main__][INFO] - Iteration 444 took 1m 9s (47.72% Gen, 50.94% Train). Generation: 32s, Training: 35s. Estimated remaining time: 50h 32m 48s. Estimated total time: 57h 35m 11s. Time estimates for 10 more iterations: 11m 31s, 100 more iterations: 1h 55m 10s, 500 more iterations: 9h 35m 51s. -[2025-11-10 01:01:19,720][__main__][INFO] - Starting iteration 444. -[2025-11-10 01:01:20,200][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 44 and human policies 1. -[2025-11-10 01:01:20,205][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:01:50,811][__main__][INFO] - Number of regex retries in iteration 444: 0 -[2025-11-10 01:01:50,812][__main__][INFO] - agents played in iteration 444 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:01:51,594][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:01:51,623][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:01:51,650][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:01:51,674][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:01:51,675][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:01:51,676][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:01:52,453][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:01:52,910][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:01:53,423][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:01:53,922][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:01:54,423][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:01:54,926][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:01:55,434][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:01:55,934][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:01:56,438][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:01:56,936][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:01:57,434][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:01:57,939][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:01:58,444][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:01:58,945][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:01:59,448][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:01:59,946][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:02:00,446][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:02:00,946][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:02:01,447][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:02:01,950][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:02:02,448][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:02:02,950][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:02:03,459][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:02:03,964][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:02:04,466][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:02:04,969][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:02:05,468][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:02:05,966][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:02:06,467][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:02:06,967][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:02:07,467][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:02:07,966][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:02:08,465][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:02:08,969][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:02:09,470][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:02:09,980][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:02:10,484][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:02:10,984][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:02:11,487][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:02:11,988][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:02:12,490][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:02:12,994][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:02:13,494][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:02:14,008][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:02:14,510][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:02:15,010][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:02:15,515][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:02:16,015][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:02:16,515][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:02:17,016][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:02:17,516][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:02:18,018][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:02:18,520][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:02:19,023][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:02:19,523][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:02:20,022][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:02:20,538][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:02:21,044][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:02:21,546][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:02:22,048][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:02:22,552][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:02:23,055][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:02:23,556][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:02:24,058][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:02:24,559][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10116 tokens. -[2025-11-10 01:02:25,392][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 55.65%, Block Peak % of device VRAM: 62.18%, ΔTime: 00:00:32 -[2025-11-10 01:02:26,017][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:02:26,019][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:02:26,021][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:02:26,889][__main__][INFO] - Iteration 445 took 1m 6s (45.89% Gen, 52.79% Train). Generation: 30s, Training: 35s. Estimated remaining time: 48h 30m 59s. Estimated total time: 55h 34m 29s. Time estimates for 10 more iterations: 11m 6s, 100 more iterations: 1h 51m 8s, 500 more iterations: 9h 15m 44s. -[2025-11-10 01:02:26,891][__main__][INFO] - Starting iteration 445. -[2025-11-10 01:02:27,355][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 44 and human policies 1. -[2025-11-10 01:02:27,356][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:02:38,902][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 01:02:50,380][__main__][INFO] - Number of regex retries in iteration 445: 1 -[2025-11-10 01:02:50,381][__main__][INFO] - agents played in iteration 445 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:02:51,188][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:02:51,218][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:02:51,247][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:02:51,271][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:02:51,271][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:02:51,272][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:02:52,084][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:02:52,542][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:02:53,056][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:02:53,557][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:02:54,061][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:02:54,565][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:02:55,066][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:02:55,572][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:02:56,072][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:02:56,571][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:02:57,076][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:02:57,576][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:02:58,076][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:02:58,578][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:02:59,081][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:02:59,580][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:03:01,212][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:03:01,775][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:03:02,277][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:03:02,775][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:03:03,276][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:03:03,787][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:03:04,288][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:03:04,788][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:03:05,293][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:03:05,793][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:03:06,294][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:03:06,795][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:03:07,297][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:03:07,798][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:03:08,303][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:03:08,805][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:03:09,303][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:03:09,804][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:03:10,302][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:03:10,806][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:03:11,308][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:03:11,806][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:03:12,307][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:03:12,807][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:03:13,308][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:03:13,809][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:03:14,324][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:03:14,825][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:03:15,329][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:03:15,833][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:03:16,339][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:03:16,839][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:03:17,339][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:03:17,846][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:03:18,348][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:03:18,852][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:03:19,359][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:03:19,867][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:03:20,366][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:03:20,872][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:03:21,376][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:03:21,874][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:03:22,378][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:03:22,877][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:03:23,380][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:03:23,880][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:03:24,385][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:03:24,882][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:03:25,382][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10153 tokens. -[2025-11-10 01:03:26,242][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 55.55%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:34 -[2025-11-10 01:03:26,882][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:03:26,883][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:03:26,885][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:03:27,746][__main__][INFO] - Iteration 446 took 1m 0s (38.13% Gen, 60.45% Train). Generation: 23s, Training: 36s. Estimated remaining time: 43h 15m 2s. Estimated total time: 50h 19m 33s. Time estimates for 10 more iterations: 10m 3s, 100 more iterations: 1h 40m 39s, 500 more iterations: 8h 23m 15s. -[2025-11-10 01:03:27,748][__main__][INFO] - Starting iteration 446. -[2025-11-10 01:03:28,240][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 44 and human policies 1. -[2025-11-10 01:03:28,241][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:03:49,168][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 01:03:59,408][__main__][INFO] - Number of regex retries in iteration 446: 1 -[2025-11-10 01:03:59,408][__main__][INFO] - agents played in iteration 446 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:04:00,284][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:04:00,312][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:04:00,338][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:04:00,362][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:04:00,362][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:04:00,363][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:04:01,168][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:04:01,631][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:04:02,142][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:04:02,645][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:04:03,148][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:04:03,653][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:04:04,157][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:04:04,654][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:04:05,151][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:04:05,667][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:04:06,166][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:04:06,665][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:04:07,160][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:04:07,659][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:04:08,162][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:04:08,661][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:04:09,164][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:04:09,687][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:04:10,188][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:04:10,685][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:04:11,187][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:04:11,690][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:04:12,191][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:04:12,693][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:04:13,191][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:04:13,688][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:04:14,187][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:04:14,686][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:04:15,184][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:04:15,684][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:04:16,184][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:04:16,679][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:04:17,176][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:04:17,671][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:04:18,171][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:04:18,675][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:04:19,174][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:04:19,670][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:04:20,169][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:04:20,671][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:04:21,172][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:04:21,669][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:04:22,167][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:04:22,676][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:04:23,176][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:04:23,676][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:04:24,174][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:04:24,675][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:04:25,176][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:04:25,675][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:04:26,176][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:04:26,677][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:04:27,178][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:04:27,680][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:04:28,182][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:04:28,684][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:04:29,189][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:04:29,692][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:04:30,198][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:04:30,700][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:04:31,203][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:04:31,710][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:04:32,213][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:04:32,716][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:04:33,221][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10268 tokens. -[2025-11-10 01:04:34,071][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 55.71%, Block Peak % of device VRAM: 62.36%, ΔTime: 00:00:32 -[2025-11-10 01:04:34,824][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:04:34,826][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:04:34,828][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:04:35,759][__main__][INFO] - Iteration 447 took 1m 7s (46.16% Gen, 52.46% Train). Generation: 31s, Training: 35s. Estimated remaining time: 49h 10m 18s. Estimated total time: 56h 15m 57s. Time estimates for 10 more iterations: 11m 15s, 100 more iterations: 1h 52m 31s, 500 more iterations: 9h 22m 39s. -[2025-11-10 01:04:35,761][__main__][INFO] - Starting iteration 447. -[2025-11-10 01:04:36,227][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 44 and human policies 1. -[2025-11-10 01:04:36,228][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:05:01,856][__main__][INFO] - Number of regex retries in iteration 447: 0 -[2025-11-10 01:05:01,859][__main__][INFO] - agents played in iteration 447 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:05:02,677][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:05:02,707][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:05:02,734][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:05:02,758][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:05:02,759][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:05:02,760][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:05:03,580][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:05:04,037][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:05:04,569][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:05:05,070][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:05:05,570][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:05:06,076][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:05:06,581][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:05:07,083][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:05:07,581][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:05:08,080][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:05:08,580][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:05:09,084][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:05:09,588][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:05:10,087][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:05:10,591][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:05:11,094][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:05:11,592][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:05:12,091][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:05:12,590][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:05:13,089][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:05:13,588][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:05:14,088][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:05:14,586][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:05:15,089][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:05:15,586][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:05:16,086][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:05:16,581][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:05:17,082][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:05:17,586][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:05:18,086][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:05:18,581][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:05:19,082][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:05:19,580][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:05:20,076][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:05:20,571][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:05:21,070][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:05:21,574][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:05:22,069][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:05:22,566][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:05:23,076][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:05:23,573][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:05:24,077][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:05:24,578][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:05:25,082][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:05:25,587][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:05:26,093][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:05:26,598][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:05:27,099][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:05:27,606][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:05:28,109][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:05:28,610][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:05:29,109][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:05:29,616][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:05:30,118][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:05:30,617][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:05:31,117][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:05:31,617][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:05:32,122][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:05:32,629][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:05:33,132][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:05:33,646][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:05:34,149][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:05:34,653][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:05:35,153][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:05:35,654][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10252 tokens. -[2025-11-10 01:05:36,509][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 55.56%, Block Peak % of device VRAM: 62.36%, ΔTime: 00:00:32 -[2025-11-10 01:05:37,170][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:05:37,171][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:05:37,173][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:05:37,997][__main__][INFO] - Iteration 448 took 1m 1s (41.49% Gen, 57.17% Train). Generation: 25s, Training: 35s. Estimated remaining time: 44h 21m 52s. Estimated total time: 51h 28m 33s. Time estimates for 10 more iterations: 10m 17s, 100 more iterations: 1h 42m 57s, 500 more iterations: 8h 34m 45s. -[2025-11-10 01:05:38,000][__main__][INFO] - Starting iteration 448. -[2025-11-10 01:05:38,477][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 44 and human policies 1. -[2025-11-10 01:05:38,478][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:06:03,503][__main__][INFO] - Number of regex retries in iteration 448: 0 -[2025-11-10 01:06:03,504][__main__][INFO] - agents played in iteration 448 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:06:04,275][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:06:04,303][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:06:04,329][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:06:04,353][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:06:04,353][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:06:04,354][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:06:05,078][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:06:05,531][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:06:06,031][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:06:06,531][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:06:07,031][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:06:07,527][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:06:08,030][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:06:08,526][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:06:09,023][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:06:09,519][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:06:10,016][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:06:10,521][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:06:11,017][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:06:11,515][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:06:12,014][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:06:12,511][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:06:13,009][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:06:13,533][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:06:14,032][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:06:14,528][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:06:15,029][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:06:15,529][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:06:16,028][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:06:16,526][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:06:17,027][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:06:17,523][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:06:18,024][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:06:18,519][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:06:19,017][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:06:19,515][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:06:20,011][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:06:20,509][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:06:21,008][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:06:21,506][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:06:22,009][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:06:22,507][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:06:23,010][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:06:23,511][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:06:24,007][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:06:24,508][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:06:25,007][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:06:25,506][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:06:26,006][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:06:26,505][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:06:27,004][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:06:27,504][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:06:28,008][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:06:28,513][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:06:29,014][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:06:29,512][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:06:30,015][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:06:30,515][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:06:31,015][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:06:31,520][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:06:32,025][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:06:32,525][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:06:33,026][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:06:33,530][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:06:34,034][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:06:34,535][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:06:35,038][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:06:35,544][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:06:36,046][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:06:36,547][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:06:37,046][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10149 tokens. -[2025-11-10 01:06:37,913][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.09%, Current % of VRAM taken: 55.58%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 -[2025-11-10 01:06:38,636][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:06:38,637][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:06:38,639][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:06:39,530][__main__][INFO] - Iteration 449 took 1m 1s (40.99% Gen, 57.55% Train). Generation: 25s, Training: 35s. Estimated remaining time: 43h 44m 59s. Estimated total time: 50h 52m 41s. Time estimates for 10 more iterations: 10m 10s, 100 more iterations: 1h 41m 45s, 500 more iterations: 8h 28m 46s. -[2025-11-10 01:06:39,532][__main__][INFO] - Starting iteration 449. -[2025-11-10 01:06:40,316][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 44 and human policies 1. -[2025-11-10 01:06:40,324][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:07:00,810][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 01:07:08,574][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 01:07:09,493][__main__][INFO] - Number of regex retries in iteration 449: 2 -[2025-11-10 01:07:09,494][__main__][INFO] - agents played in iteration 449 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:07:10,285][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:07:10,307][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:07:10,330][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:07:10,353][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:07:10,353][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:07:10,355][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:07:11,117][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:07:11,572][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:07:12,098][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:07:12,599][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:07:13,110][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:07:13,613][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:07:14,112][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:07:14,612][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:07:15,117][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:07:15,617][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:07:16,118][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:07:16,616][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:07:17,119][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:07:17,617][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:07:18,118][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:07:18,625][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:07:19,126][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:07:19,627][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:07:20,125][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:07:20,623][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:07:21,120][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:07:21,620][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:07:22,116][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:07:22,615][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:07:23,114][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:07:23,612][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:07:24,120][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:07:24,617][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:07:25,120][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:07:25,617][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:07:26,119][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:07:26,622][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:07:27,122][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:07:27,620][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:07:28,119][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:07:28,616][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:07:29,116][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:07:29,612][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:07:30,108][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:07:30,610][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:07:31,109][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:07:31,610][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:07:32,111][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:07:32,610][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:07:33,114][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:07:33,615][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:07:34,121][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:07:34,623][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:07:35,121][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:07:35,624][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:07:36,126][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:07:36,630][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:07:37,137][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:07:37,642][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:07:38,146][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:07:38,654][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:07:39,154][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:07:39,662][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:07:40,163][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:07:40,663][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:07:41,168][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:07:41,667][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:07:42,165][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:07:42,672][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:07:43,173][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10078 tokens. -[2025-11-10 01:07:44,045][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 13.99%, Current % of VRAM taken: 55.49%, Block Peak % of device VRAM: 62.10%, ΔTime: 00:00:32 -[2025-11-10 01:07:44,698][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:07:44,699][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:07:44,701][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:07:45,575][__main__][INFO] - Iteration 450 took 1m 5s (44.70% Gen, 53.95% Train). Generation: 29s, Training: 35s. Estimated remaining time: 47h 14m 10s. Estimated total time: 54h 22m 59s. Time estimates for 10 more iterations: 10m 52s, 100 more iterations: 1h 48m 45s, 500 more iterations: 9h 3m 49s. -[2025-11-10 01:07:45,577][__main__][INFO] - Starting iteration 450. -[2025-11-10 01:07:46,059][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 44 and human policies 1. -[2025-11-10 01:07:46,060][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:08:07,968][__main__][INFO] - Number of regex retries in iteration 450: 0 -[2025-11-10 01:08:07,968][__main__][INFO] - agents played in iteration 450 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:08:08,742][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:08:08,769][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:08:08,795][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:08:08,818][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:08:08,819][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:08:08,820][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:08:09,562][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:08:10,023][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:08:10,530][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:08:11,029][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:08:11,534][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:08:12,035][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:08:12,536][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:08:13,038][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:08:13,540][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:08:14,042][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:08:14,538][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:08:15,035][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:08:15,531][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:08:16,031][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:08:16,533][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:08:17,033][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:08:17,533][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:08:18,038][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:08:18,541][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:08:19,041][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:08:19,546][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:08:20,049][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:08:20,553][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:08:21,060][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:08:21,562][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:08:22,063][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:08:22,570][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:08:23,086][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:08:23,584][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:08:24,085][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:08:24,591][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:08:25,517][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:08:26,016][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:08:26,517][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:08:27,021][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:08:27,524][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:08:28,026][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:08:28,526][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:08:29,027][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:08:29,531][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:08:30,034][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:08:30,537][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:08:31,042][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:08:31,539][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:08:32,039][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:08:32,538][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:08:33,039][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:08:33,539][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:08:34,040][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:08:34,539][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:08:35,044][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:08:35,551][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:08:36,053][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:08:36,555][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:08:37,055][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:08:37,557][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:08:38,061][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:08:38,559][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:08:39,058][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:08:39,559][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:08:40,060][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:08:40,560][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:08:41,060][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:08:41,560][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:08:42,060][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10192 tokens. -[2025-11-10 01:08:42,906][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 55.62%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:33 -[2025-11-10 01:08:43,553][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:08:43,554][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:08:43,556][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:08:45,231][__main__][INFO] - Iteration 451 took 59s (37.02% Gen, 60.14% Train). Generation: 21s, Training: 35s. Estimated remaining time: 42h 8m 49s. Estimated total time: 49h 18m 37s. Time estimates for 10 more iterations: 9m 51s, 100 more iterations: 1h 38m 37s, 500 more iterations: 8h 13m 6s. -[2025-11-10 01:08:45,234][__main__][INFO] - Starting iteration 451. -[2025-11-10 01:08:45,712][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 45 and human policies 1. -[2025-11-10 01:08:45,713][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:09:00,486][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 01:09:04,477][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 01:09:12,254][__main__][INFO] - Number of regex retries in iteration 451: 2 -[2025-11-10 01:09:12,254][__main__][INFO] - agents played in iteration 451 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:09:13,042][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:09:13,070][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:09:13,096][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:09:13,119][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:09:13,120][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:09:13,121][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:09:13,860][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:09:14,316][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:09:14,822][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:09:15,330][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:09:15,837][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:09:16,336][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:09:16,837][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:09:17,338][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:09:17,847][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:09:18,350][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:09:18,854][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:09:19,360][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:09:19,860][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:09:20,359][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:09:20,857][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:09:21,355][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:09:21,857][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:09:22,357][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:09:22,859][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:09:23,366][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:09:23,868][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:09:24,368][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:09:24,873][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:09:25,375][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:09:25,874][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:09:26,373][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:09:26,872][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:09:27,372][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:09:27,871][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:09:28,367][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:09:28,866][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:09:29,361][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:09:29,859][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:09:30,359][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:09:30,857][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:09:31,357][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:09:31,852][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:09:32,350][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:09:32,850][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:09:33,349][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:09:33,854][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:09:34,354][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:09:34,855][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:09:35,355][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:09:35,851][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:09:36,349][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:09:36,850][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:09:37,347][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:09:37,845][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:09:38,344][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:09:38,842][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:09:39,340][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:09:39,835][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:09:40,335][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:09:40,832][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:09:41,330][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:09:41,828][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:09:42,323][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:09:42,818][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:09:43,335][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:09:43,834][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:09:44,337][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:09:44,837][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:09:45,336][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:09:45,842][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10205 tokens. -[2025-11-10 01:09:46,627][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 55.60%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 -[2025-11-10 01:09:47,362][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:09:47,365][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:09:47,367][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:09:48,220][__main__][INFO] - Iteration 452 took 1m 2s (42.46% Gen, 56.17% Train). Generation: 26s, Training: 35s. Estimated remaining time: 44h 54m 33s. Estimated total time: 52h 5m 24s. Time estimates for 10 more iterations: 10m 25s, 100 more iterations: 1h 44m 10s, 500 more iterations: 8h 40m 54s. -[2025-11-10 01:09:48,222][__main__][INFO] - Starting iteration 452. -[2025-11-10 01:09:48,715][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 45 and human policies 1. -[2025-11-10 01:09:48,716][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:10:09,864][__main__][INFO] - Number of regex retries in iteration 452: 0 -[2025-11-10 01:10:09,867][__main__][INFO] - agents played in iteration 452 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:10:10,825][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:10:10,848][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:10:10,871][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:10:10,894][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:10:10,895][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:10:10,896][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:10:11,774][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:10:12,302][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:10:12,814][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:10:13,319][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:10:13,821][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:10:14,322][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:10:14,822][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:10:15,324][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:10:15,825][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:10:16,328][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:10:16,831][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:10:17,333][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:10:17,832][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:10:18,341][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:10:18,840][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:10:19,343][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:10:19,843][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:10:20,354][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:10:20,861][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:10:21,363][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:10:21,866][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:10:22,369][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:10:22,872][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:10:23,376][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:10:23,886][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:10:24,386][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:10:24,891][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:10:25,390][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:10:25,891][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:10:26,396][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:10:26,898][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:10:27,402][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:10:27,903][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:10:28,403][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:10:28,904][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:10:29,404][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:10:29,906][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:10:30,409][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:10:30,912][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:10:31,415][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:10:31,914][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:10:32,415][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:10:32,914][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:10:33,414][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:10:33,931][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:10:34,431][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:10:34,935][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:10:35,435][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:10:35,934][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:10:36,440][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:10:36,937][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:10:37,437][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:10:37,940][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:10:38,441][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:10:38,945][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:10:39,443][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:10:39,944][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:10:40,450][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:10:40,951][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:10:41,452][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:10:41,953][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:10:42,450][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:10:42,959][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:10:43,455][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:10:43,955][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10108 tokens. -[2025-11-10 01:10:44,814][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.03%, Current % of VRAM taken: 55.53%, Block Peak % of device VRAM: 62.36%, ΔTime: 00:00:33 -[2025-11-10 01:10:45,458][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:10:45,460][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:10:45,462][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:10:46,313][__main__][INFO] - Iteration 453 took 57s (36.72% Gen, 61.80% Train). Generation: 21s, Training: 35s. Estimated remaining time: 40h 48m 7s. Estimated total time: 47h 59m 56s. Time estimates for 10 more iterations: 9m 35s, 100 more iterations: 1h 35m 59s, 500 more iterations: 7h 59m 59s. -[2025-11-10 01:10:46,317][__main__][INFO] - Starting iteration 453. -[2025-11-10 01:10:46,810][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 45 and human policies 1. -[2025-11-10 01:10:46,810][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:11:18,350][__main__][INFO] - Number of regex retries in iteration 453: 0 -[2025-11-10 01:11:18,350][__main__][INFO] - agents played in iteration 453 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:11:19,120][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:11:19,147][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:11:19,173][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:11:19,196][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:11:19,196][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:11:19,197][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:11:19,900][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:11:20,356][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:11:20,863][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:11:21,370][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:11:21,872][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:11:22,370][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:11:22,869][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:11:23,367][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:11:23,866][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:11:24,361][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:11:24,860][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:11:25,357][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:11:25,853][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:11:26,349][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:11:26,845][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:11:27,341][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:11:27,838][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:11:28,335][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:11:28,848][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:11:29,345][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:11:29,841][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:11:30,338][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:11:30,836][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:11:31,333][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:11:31,829][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:11:32,326][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:11:32,822][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:11:33,318][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:11:33,812][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:11:34,320][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:11:34,817][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:11:35,317][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:11:35,814][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:11:36,310][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:11:36,808][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:11:37,304][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:11:37,802][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:11:38,312][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:11:38,810][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:11:39,306][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:11:39,805][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:11:40,300][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:11:40,800][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:11:41,296][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:11:41,794][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:11:42,289][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:11:42,785][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:11:43,280][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:11:43,777][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:11:44,272][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:11:44,771][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:11:45,267][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:11:45,763][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:11:46,265][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:11:46,760][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:11:47,254][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:11:47,749][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:11:48,248][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:11:48,749][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:11:49,248][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:11:49,750][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:11:50,263][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:11:50,759][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:11:51,261][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:11:51,761][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10250 tokens. -[2025-11-10 01:11:52,579][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 55.61%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 -[2025-11-10 01:11:53,322][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:11:53,324][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:11:53,326][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:11:54,258][__main__][INFO] - Iteration 454 took 1m 7s (46.76% Gen, 51.86% Train). Generation: 31s, Training: 34s. Estimated remaining time: 48h 59m 30s. Estimated total time: 56h 12m 27s. Time estimates for 10 more iterations: 11m 14s, 100 more iterations: 1h 52m 24s, 500 more iterations: 9h 22m 4s. -[2025-11-10 01:11:54,260][__main__][INFO] - Starting iteration 454. -[2025-11-10 01:11:54,726][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 45 and human policies 1. -[2025-11-10 01:11:54,726][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:12:19,571][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 01:12:20,475][__main__][INFO] - Number of regex retries in iteration 454: 1 -[2025-11-10 01:12:20,476][__main__][INFO] - agents played in iteration 454 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:12:21,262][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:12:21,287][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:12:21,312][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:12:21,334][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:12:21,335][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:12:21,336][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:12:22,066][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:12:22,519][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:12:23,023][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:12:23,525][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:12:24,027][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:12:24,536][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:12:25,038][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:12:25,548][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:12:26,044][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:12:26,548][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:12:27,046][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:12:27,545][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:12:28,047][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:12:28,551][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:12:29,050][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:12:29,563][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:12:30,061][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:12:30,559][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:12:31,057][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:12:31,555][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:12:32,061][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:12:32,559][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:12:33,078][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:12:33,580][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:12:34,081][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:12:34,575][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:12:35,070][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:12:35,565][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:12:36,070][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:12:36,569][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:12:37,064][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:12:37,566][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:12:38,065][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:12:38,567][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:12:39,062][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:12:39,557][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:12:40,056][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:12:40,559][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:12:41,055][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:12:41,551][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:12:42,051][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:12:42,552][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:12:43,046][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:12:43,546][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:12:44,045][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:12:44,546][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:12:45,058][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:12:45,555][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:12:46,049][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:12:46,548][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:12:47,044][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:12:47,539][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:12:48,042][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:12:48,538][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:12:49,061][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:12:49,558][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:12:50,056][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:12:50,556][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:12:51,053][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:12:51,563][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:12:52,063][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:12:52,561][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:12:53,062][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:12:53,562][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:12:54,063][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10278 tokens. -[2025-11-10 01:12:54,886][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 55.74%, Block Peak % of device VRAM: 62.48%, ΔTime: 00:00:32 -[2025-11-10 01:12:55,544][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:12:55,547][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:12:55,548][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:12:56,348][__main__][INFO] - Iteration 455 took 1m 1s (41.79% Gen, 56.91% Train). Generation: 25s, Training: 35s. Estimated remaining time: 44h 7m 10s. Estimated total time: 51h 21m 9s. Time estimates for 10 more iterations: 10m 16s, 100 more iterations: 1h 42m 42s, 500 more iterations: 8h 33m 31s. -[2025-11-10 01:12:56,351][__main__][INFO] - Starting iteration 455. -[2025-11-10 01:12:56,819][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 45 and human policies 1. -[2025-11-10 01:12:56,819][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:13:20,903][__main__][INFO] - Number of regex retries in iteration 455: 0 -[2025-11-10 01:13:20,904][__main__][INFO] - agents played in iteration 455 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:13:21,682][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:13:21,719][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:13:21,743][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:13:21,767][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:13:21,768][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:13:21,768][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:13:22,536][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:13:22,988][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:13:23,494][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:13:23,995][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:13:24,499][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:13:25,000][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:13:25,500][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:13:26,004][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:13:26,499][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:13:26,996][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:13:27,490][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:13:27,985][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:13:28,487][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:13:28,981][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:13:29,475][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:13:29,969][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:13:30,463][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:13:30,962][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:13:31,461][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:13:31,956][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:13:32,477][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:13:32,975][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:13:33,469][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:13:33,970][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:13:34,465][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:13:34,971][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:13:35,464][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:13:35,956][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:13:36,452][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:13:36,945][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:13:37,438][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:13:37,943][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:13:38,440][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:13:38,934][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:13:39,431][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:13:39,927][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:13:40,422][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:13:40,917][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:13:41,414][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:13:41,908][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:13:42,402][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:13:42,896][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:13:43,391][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:13:43,888][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:13:44,393][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:13:44,888][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:13:45,382][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:13:45,876][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:13:46,369][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:13:46,874][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:13:47,369][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:13:47,864][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:13:48,375][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:13:48,870][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:13:49,369][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:13:49,868][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:13:50,362][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:13:50,878][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:13:51,373][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:13:51,866][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:13:52,363][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:13:52,856][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:13:53,358][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:13:53,856][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:13:54,354][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10085 tokens. -[2025-11-10 01:13:55,145][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 55.57%, Block Peak % of device VRAM: 62.13%, ΔTime: 00:00:32 -[2025-11-10 01:13:55,931][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:13:55,933][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:13:55,934][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:13:56,816][__main__][INFO] - Iteration 456 took 59s (40.14% Gen, 58.39% Train). Generation: 24s, Training: 35s. Estimated remaining time: 42h 44m 53s. Estimated total time: 49h 59m 53s. Time estimates for 10 more iterations: 9m 59s, 100 more iterations: 1h 39m 59s, 500 more iterations: 8h 19m 58s. -[2025-11-10 01:13:56,819][__main__][INFO] - Starting iteration 456. -[2025-11-10 01:13:57,292][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 45 and human policies 1. -[2025-11-10 01:13:57,293][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:14:16,350][__main__][INFO] - Number of regex retries in iteration 456: 0 -[2025-11-10 01:14:16,351][__main__][INFO] - agents played in iteration 456 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:14:17,146][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:14:17,174][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:14:17,200][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:14:17,223][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:14:17,223][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:14:17,225][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:14:17,974][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:14:18,433][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:14:18,938][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:14:19,437][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:14:19,933][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:14:20,436][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:14:20,937][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:14:21,441][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:14:21,941][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:14:22,447][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:14:22,945][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:14:23,440][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:14:23,934][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:14:24,433][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:14:24,928][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:14:25,424][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:14:25,922][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:14:26,423][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:14:26,919][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:14:27,417][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:14:27,917][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:14:28,412][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:14:28,910][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:14:29,411][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:14:29,909][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:14:30,410][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:14:30,909][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:14:31,410][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:14:31,909][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:14:32,410][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:14:32,908][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:14:33,409][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:14:33,907][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:14:34,400][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:14:34,895][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:14:35,387][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:14:35,881][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:14:36,373][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:14:36,871][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:14:37,365][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:14:37,867][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:14:38,366][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:14:38,863][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:14:39,358][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:14:39,852][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:14:40,346][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:14:40,840][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:14:41,335][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:14:41,833][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:14:42,330][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:14:42,826][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:14:43,323][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:14:43,838][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:14:44,338][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:14:44,833][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:14:45,334][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:14:45,836][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:14:46,328][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:14:46,820][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:14:47,313][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:14:47,808][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:14:48,305][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:14:48,798][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:14:49,295][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:14:49,799][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10128 tokens. -[2025-11-10 01:14:50,592][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 55.60%, Block Peak % of device VRAM: 62.33%, ΔTime: 00:00:32 -[2025-11-10 01:14:51,342][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:14:51,343][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:14:51,345][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:14:52,226][__main__][INFO] - Iteration 457 took 54s (34.69% Gen, 63.70% Train). Generation: 19s, Training: 34s. Estimated remaining time: 38h 30m 49s. Estimated total time: 45h 46m 44s. Time estimates for 10 more iterations: 9m 9s, 100 more iterations: 1h 31m 33s, 500 more iterations: 7h 37m 47s. -[2025-11-10 01:14:52,228][__main__][INFO] - Starting iteration 457. -[2025-11-10 01:14:52,710][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 45 and human policies 1. -[2025-11-10 01:14:52,711][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:15:17,311][__main__][INFO] - Number of regex retries in iteration 457: 0 -[2025-11-10 01:15:17,313][__main__][INFO] - agents played in iteration 457 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:15:18,205][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:15:18,267][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:15:18,304][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:15:18,337][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:15:18,337][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:15:18,338][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:15:19,103][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:15:19,611][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:15:20,130][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:15:20,630][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:15:21,130][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:15:21,630][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:15:22,131][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:15:22,632][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:15:23,136][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:15:23,633][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:15:24,138][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:15:24,639][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:15:25,141][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:15:25,646][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:15:26,146][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:15:26,644][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:15:27,141][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:15:27,640][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:15:28,140][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:15:28,637][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:15:29,141][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:15:29,641][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:15:30,139][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:15:30,637][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:15:31,135][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:15:31,631][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:15:32,127][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:15:32,622][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:15:33,123][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:15:33,617][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:15:34,119][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:15:34,612][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:15:35,108][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:15:35,613][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:15:36,110][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:15:36,607][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:15:37,108][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:15:37,606][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:15:38,107][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:15:38,605][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:15:39,102][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:15:39,597][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:15:40,095][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:15:40,591][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:15:41,091][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:15:41,586][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:15:42,090][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:15:42,587][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:15:43,085][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:15:43,585][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:15:44,081][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:15:44,583][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:15:45,078][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:15:45,572][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:15:46,067][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:15:46,563][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:15:47,059][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:15:47,555][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:15:48,054][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:15:48,552][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:15:49,048][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:15:49,542][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:15:50,039][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:15:50,537][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:15:51,037][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10224 tokens. -[2025-11-10 01:15:51,785][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 55.72%, Block Peak % of device VRAM: 62.39%, ΔTime: 00:00:32 -[2025-11-10 01:15:52,430][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:15:52,432][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:15:52,434][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:15:53,239][__main__][INFO] - Iteration 458 took 1m 0s (40.65% Gen, 58.02% Train). Generation: 24s, Training: 35s. Estimated remaining time: 43h 9m 31s. Estimated total time: 50h 26m 28s. Time estimates for 10 more iterations: 10m 5s, 100 more iterations: 1h 40m 52s, 500 more iterations: 8h 24m 24s. -[2025-11-10 01:15:53,241][__main__][INFO] - Starting iteration 458. -[2025-11-10 01:15:53,708][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 45 and human policies 1. -[2025-11-10 01:15:53,710][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:16:18,166][__main__][INFO] - Number of regex retries in iteration 458: 0 -[2025-11-10 01:16:18,167][__main__][INFO] - agents played in iteration 458 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:16:18,959][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:16:18,981][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:16:19,003][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:16:19,025][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:16:19,027][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:16:19,027][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:16:19,750][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:16:20,220][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:16:20,730][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:16:21,231][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:16:21,734][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:16:22,234][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:16:22,735][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:16:23,236][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:16:23,734][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:16:24,230][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:16:24,729][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:16:25,239][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:16:25,734][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:16:26,228][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:16:26,722][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:16:27,219][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:16:27,717][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:16:28,211][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:16:28,705][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:16:29,203][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:16:29,700][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:16:30,196][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:16:30,689][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:16:31,182][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:16:31,676][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:16:32,176][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:16:32,673][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:16:33,171][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:16:33,669][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:16:34,168][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:16:34,671][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:16:35,169][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:16:35,669][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:16:36,164][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:16:36,658][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:16:37,153][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:16:37,648][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:16:38,143][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:16:38,637][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:16:39,132][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:16:39,634][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:16:40,130][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:16:40,625][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:16:41,119][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:16:41,614][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:16:42,119][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:16:42,614][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:16:43,112][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:16:43,609][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:16:44,108][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:16:44,607][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:16:45,107][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:16:45,603][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:16:46,110][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:16:46,607][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:16:47,105][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:16:47,622][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:16:48,118][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:16:48,615][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:16:49,119][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:16:49,616][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:16:50,114][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:16:50,614][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:16:51,110][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:16:51,617][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10201 tokens. -[2025-11-10 01:16:52,343][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 55.70%, Block Peak % of device VRAM: 62.38%, ΔTime: 00:00:32 -[2025-11-10 01:16:53,073][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:16:53,075][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:16:53,076][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:16:54,019][__main__][INFO] - Iteration 459 took 1m 0s (40.55% Gen, 57.88% Train). Generation: 24s, Training: 34s. Estimated remaining time: 42h 57m 37s. Estimated total time: 50h 15m 34s. Time estimates for 10 more iterations: 10m 3s, 100 more iterations: 1h 40m 31s, 500 more iterations: 8h 22m 35s. -[2025-11-10 01:16:54,021][__main__][INFO] - Starting iteration 459. -[2025-11-10 01:16:54,487][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 45 and human policies 1. -[2025-11-10 01:16:54,488][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:17:17,172][__main__][INFO] - Number of regex retries in iteration 459: 0 -[2025-11-10 01:17:17,174][__main__][INFO] - agents played in iteration 459 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:17:18,013][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:17:18,053][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:17:18,084][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:17:18,111][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:17:18,111][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:17:18,112][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:17:18,855][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:17:19,351][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:17:19,859][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:17:20,356][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:17:20,853][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:17:21,353][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:17:21,852][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:17:22,349][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:17:22,844][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:17:23,348][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:17:23,852][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:17:24,355][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:17:24,861][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:17:25,361][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:17:25,864][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:17:26,360][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:17:26,854][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:17:27,356][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:17:27,853][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:17:28,354][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:17:28,854][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:17:29,350][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:17:29,850][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:17:30,348][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:17:30,842][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:17:31,346][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:17:31,843][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:17:32,337][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:17:32,835][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:17:33,330][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:17:33,826][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:17:34,324][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:17:34,818][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:17:35,331][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:17:35,829][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:17:36,326][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:17:36,824][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:17:37,321][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:17:37,827][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:17:38,324][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:17:38,820][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:17:39,335][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:17:39,836][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:17:40,335][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:17:40,831][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:17:41,326][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:17:41,838][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:17:42,332][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:17:42,828][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:17:43,327][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:17:43,826][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:17:44,334][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:17:44,835][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:17:45,334][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:17:45,834][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:17:46,334][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:17:46,833][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:17:47,332][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:17:47,827][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:17:48,357][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:17:48,857][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:17:49,357][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:17:49,857][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:17:50,357][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:17:50,860][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10229 tokens. -[2025-11-10 01:17:51,602][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.30%, Current % of VRAM taken: 55.79%, Block Peak % of device VRAM: 62.38%, ΔTime: 00:00:32 -[2025-11-10 01:17:52,263][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:17:52,265][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:17:52,267][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:17:53,074][__main__][INFO] - Iteration 460 took 58s (38.72% Gen, 59.90% Train). Generation: 22s, Training: 35s. Estimated remaining time: 41h 30m 26s. Estimated total time: 48h 49m 23s. Time estimates for 10 more iterations: 9m 45s, 100 more iterations: 1h 37m 38s, 500 more iterations: 8h 8m 13s. -[2025-11-10 01:17:53,076][__main__][INFO] - Starting iteration 460. -[2025-11-10 01:17:53,561][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 45 and human policies 1. -[2025-11-10 01:17:53,561][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:18:20,013][__main__][INFO] - Number of regex retries in iteration 460: 0 -[2025-11-10 01:18:20,014][__main__][INFO] - agents played in iteration 460 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:18:20,822][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:18:20,849][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:18:20,875][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:18:20,898][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.66%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:18:20,898][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:18:20,900][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:18:21,659][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:18:22,119][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:18:22,643][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:18:23,143][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:18:23,651][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:18:24,151][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:18:24,654][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:18:25,162][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:18:25,667][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:18:26,165][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:18:26,665][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:18:27,166][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:18:27,668][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:18:28,174][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:18:28,674][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:18:29,177][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:18:29,678][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:18:30,174][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:18:30,670][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:18:31,170][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:18:31,668][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:18:32,162][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:18:32,660][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:18:33,158][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:18:33,654][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:18:34,155][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:18:34,652][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:18:35,153][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:18:35,655][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:18:36,152][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:18:36,650][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:18:37,151][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:18:37,647][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:18:38,149][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:18:38,646][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:18:39,141][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:18:39,649][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:18:40,142][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:18:40,638][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:18:41,133][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:18:41,630][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:18:42,149][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:18:42,644][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:18:43,142][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:18:43,637][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:18:44,133][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:18:44,643][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:18:45,140][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:18:45,635][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:18:46,148][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:18:46,642][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:18:47,141][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:18:47,636][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:18:48,133][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:18:48,651][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:18:49,146][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:18:49,642][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:18:50,141][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:18:50,634][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:18:51,143][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:18:51,644][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:18:52,138][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:18:52,656][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:18:53,152][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:18:53,648][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10101 tokens. -[2025-11-10 01:18:54,369][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 55.61%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 -[2025-11-10 01:18:55,139][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:18:55,141][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:18:55,143][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:18:56,874][__main__][INFO] - Iteration 461 took 1m 3s (41.78% Gen, 55.48% Train). Generation: 26s, Training: 35s. Estimated remaining time: 45h 25m 42s. Estimated total time: 52h 45m 42s. Time estimates for 10 more iterations: 10m 33s, 100 more iterations: 1h 45m 31s, 500 more iterations: 8h 47m 37s. -[2025-11-10 01:18:56,876][__main__][INFO] - Starting iteration 461. -[2025-11-10 01:18:57,349][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 46 and human policies 1. -[2025-11-10 01:18:57,349][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:19:23,077][__main__][INFO] - Number of regex retries in iteration 461: 0 -[2025-11-10 01:19:23,078][__main__][INFO] - agents played in iteration 461 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:19:23,897][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:19:23,925][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:19:23,951][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:19:23,974][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:19:23,975][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:19:23,975][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:19:24,755][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:19:25,212][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:19:25,717][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:19:26,217][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:19:26,720][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:19:27,218][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:19:27,713][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:19:28,215][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:19:28,712][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:19:29,220][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:19:29,825][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:19:30,337][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:19:30,847][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:19:31,348][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:19:31,857][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:19:32,357][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:19:32,855][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:19:33,356][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:19:33,853][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:19:34,349][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:19:34,851][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:19:35,347][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:19:35,867][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:19:36,368][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:19:36,885][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:19:37,384][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:19:37,879][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:19:38,382][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:19:38,880][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:19:39,374][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:19:39,878][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:19:40,373][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:19:40,868][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:19:41,365][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:19:41,860][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:19:42,373][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:19:42,869][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:19:43,365][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:19:43,860][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:19:44,359][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:19:44,863][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:19:45,361][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:19:45,856][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:19:46,364][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:19:46,859][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:19:47,354][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:19:47,851][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:19:48,347][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:19:48,860][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:19:49,357][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:19:49,857][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:19:50,354][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:19:50,850][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:19:51,360][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:19:51,857][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:19:52,354][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:19:52,873][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:19:53,369][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:19:53,863][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:19:54,362][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:19:54,856][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:19:55,363][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:19:55,857][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:19:56,357][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:19:56,874][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10241 tokens. -[2025-11-10 01:19:57,648][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 55.71%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:32 -[2025-11-10 01:19:58,423][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:19:58,425][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:19:58,426][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:19:59,321][__main__][INFO] - Iteration 462 took 1m 1s (41.52% Gen, 57.04% Train). Generation: 25s, Training: 35s. Estimated remaining time: 44h 17m 36s. Estimated total time: 51h 38m 39s. Time estimates for 10 more iterations: 10m 19s, 100 more iterations: 1h 43m 17s, 500 more iterations: 8h 36m 26s. -[2025-11-10 01:19:59,323][__main__][INFO] - Starting iteration 462. -[2025-11-10 01:19:59,790][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 46 and human policies 1. -[2025-11-10 01:19:59,791][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:20:29,407][__main__][INFO] - Number of regex retries in iteration 462: 0 -[2025-11-10 01:20:29,410][__main__][INFO] - agents played in iteration 462 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:20:30,195][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:20:30,222][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:20:30,248][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:20:30,271][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:20:30,272][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:20:30,273][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:20:30,996][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:20:31,451][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:20:31,970][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:20:32,470][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:20:32,966][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:20:33,464][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:20:33,961][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:20:34,472][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:20:34,970][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:20:35,469][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:20:35,970][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:20:36,470][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:20:36,976][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:20:37,479][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:20:37,978][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:20:38,484][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:20:38,980][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:20:39,479][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:20:39,974][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:20:40,471][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:20:40,974][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:20:41,471][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:20:41,971][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:20:42,466][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:20:42,962][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:20:43,458][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:20:43,953][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:20:44,449][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:20:44,961][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:20:45,460][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:20:45,956][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:20:46,456][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:20:46,954][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:20:47,464][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:20:47,956][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:20:48,450][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:20:48,948][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:20:49,445][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:20:49,938][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:20:50,448][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:20:50,942][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:20:51,447][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:20:51,941][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:20:52,436][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:20:52,941][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:20:53,436][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:20:53,933][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:20:54,430][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:20:54,928][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:20:55,427][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:20:55,923][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:20:56,421][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:20:56,926][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:20:57,427][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:20:57,930][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:20:58,431][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:20:58,931][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:20:59,436][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:20:59,956][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:21:00,458][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:21:00,961][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:21:01,464][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:21:01,977][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:21:02,482][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:21:02,988][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10228 tokens. -[2025-11-10 01:21:03,839][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 -[2025-11-10 01:21:04,505][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:21:04,508][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:21:04,510][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:21:05,315][__main__][INFO] - Iteration 463 took 1m 5s (45.20% Gen, 53.57% Train). Generation: 29s, Training: 35s. Estimated remaining time: 47h 14m 8s. Estimated total time: 54h 36m 16s. Time estimates for 10 more iterations: 10m 55s, 100 more iterations: 1h 49m 12s, 500 more iterations: 9h 6m 2s. -[2025-11-10 01:21:05,318][__main__][INFO] - Starting iteration 463. -[2025-11-10 01:21:05,783][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 46 and human policies 1. -[2025-11-10 01:21:05,783][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:21:29,244][__main__][INFO] - Number of regex retries in iteration 463: 0 -[2025-11-10 01:21:29,244][__main__][INFO] - agents played in iteration 463 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:21:30,311][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:21:30,339][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:21:30,366][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:21:30,390][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:21:30,390][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:21:30,391][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:21:31,126][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:21:31,582][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:21:32,097][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:21:32,597][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:21:33,097][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:21:33,600][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:21:34,098][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:21:34,598][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:21:35,099][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:21:35,595][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:21:36,095][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:21:36,593][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:21:37,094][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:21:37,593][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:21:38,091][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:21:38,591][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:21:39,090][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:21:39,588][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:21:40,098][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:21:40,596][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:21:41,091][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:21:41,589][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:21:42,083][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:21:42,579][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:21:43,073][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:21:43,571][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:21:44,070][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:21:44,567][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:21:45,066][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:21:45,561][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:21:46,060][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:21:46,577][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:21:47,075][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:21:47,573][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:21:48,067][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:21:48,565][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:21:49,062][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:21:49,560][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:21:50,055][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:21:50,553][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:21:51,048][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:21:51,544][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:21:52,045][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:21:52,544][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:21:53,039][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:21:53,535][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:21:54,035][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:21:54,535][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:21:55,033][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:21:55,532][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:21:56,032][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:21:56,528][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:21:57,025][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:21:57,522][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:21:58,018][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:21:58,514][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:21:59,011][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:21:59,509][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:22:00,010][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:22:00,512][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:22:01,012][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:22:01,515][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:22:02,015][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:22:02,517][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:22:03,019][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10285 tokens. -[2025-11-10 01:22:03,856][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 55.62%, Block Peak % of device VRAM: 62.23%, ΔTime: 00:00:32 -[2025-11-10 01:22:04,552][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:22:04,553][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:22:04,555][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:22:05,456][__main__][INFO] - Iteration 464 took 59s (39.31% Gen, 59.17% Train). Generation: 23s, Training: 35s. Estimated remaining time: 42h 20m 34s. Estimated total time: 49h 43m 42s. Time estimates for 10 more iterations: 9m 56s, 100 more iterations: 1h 39m 27s, 500 more iterations: 8h 17m 17s. -[2025-11-10 01:22:05,459][__main__][INFO] - Starting iteration 464. -[2025-11-10 01:22:05,929][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 46 and human policies 1. -[2025-11-10 01:22:05,929][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:22:29,978][__main__][INFO] - Number of regex retries in iteration 464: 0 -[2025-11-10 01:22:29,980][__main__][INFO] - agents played in iteration 464 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:22:30,775][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:22:30,803][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:22:30,829][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:22:30,853][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:22:30,854][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:22:30,854][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:22:31,645][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:22:32,104][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:22:32,614][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:22:33,118][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:22:33,627][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:22:34,133][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:22:34,631][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:22:35,135][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:22:35,635][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:22:36,134][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:22:36,630][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:22:37,125][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:22:37,621][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:22:38,120][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:22:38,619][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:22:39,114][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:22:39,613][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:22:40,114][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:22:40,615][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:22:41,114][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:22:41,613][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:22:42,115][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:22:42,612][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:22:43,111][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:22:43,612][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:22:44,117][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:22:44,617][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:22:45,112][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:22:45,610][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:22:46,117][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:22:46,619][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:22:47,121][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:22:47,620][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:22:48,119][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:22:48,620][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:22:49,118][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:22:49,618][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:22:50,117][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:22:50,618][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:22:51,119][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:22:51,618][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:22:52,116][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:22:52,620][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:22:53,118][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:22:53,617][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:22:54,115][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:22:54,613][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:22:55,110][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:22:55,608][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:22:56,105][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:22:56,616][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:22:57,114][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:22:57,610][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:22:58,110][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:22:58,610][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:22:59,110][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:22:59,606][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:23:00,104][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:23:00,618][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:23:01,117][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:23:01,620][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:23:02,127][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:23:02,628][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:23:03,130][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:23:03,634][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10254 tokens. -[2025-11-10 01:23:04,474][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.29%, Current % of VRAM taken: 55.78%, Block Peak % of device VRAM: 62.39%, ΔTime: 00:00:32 -[2025-11-10 01:23:05,119][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:23:05,122][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:23:05,124][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:23:05,944][__main__][INFO] - Iteration 465 took 1m 0s (40.07% Gen, 58.56% Train). Generation: 24s, Training: 35s. Estimated remaining time: 42h 36m 39s. Estimated total time: 50h 0m 48s. Time estimates for 10 more iterations: 10m 0s, 100 more iterations: 1h 40m 1s, 500 more iterations: 8h 20m 8s. -[2025-11-10 01:23:05,946][__main__][INFO] - Starting iteration 465. -[2025-11-10 01:23:06,410][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 46 and human policies 1. -[2025-11-10 01:23:06,411][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:23:32,750][__main__][INFO] - Number of regex retries in iteration 465: 0 -[2025-11-10 01:23:32,751][__main__][INFO] - agents played in iteration 465 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:23:33,541][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:23:33,565][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:23:33,588][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:23:33,611][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:23:33,612][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:23:33,613][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:23:34,482][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:23:34,951][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:23:35,461][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:23:35,960][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:23:36,456][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:23:36,953][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:23:37,453][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:23:37,951][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:23:38,454][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:23:38,950][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:23:39,448][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:23:39,947][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:23:40,445][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:23:40,948][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:23:41,453][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:23:41,953][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:23:42,450][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:23:42,949][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:23:43,447][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:23:43,949][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:23:44,445][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:23:44,943][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:23:45,445][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:23:45,942][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:23:46,445][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:23:46,942][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:23:47,438][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:23:47,944][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:23:48,441][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:23:48,941][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:23:49,441][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:23:49,941][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:23:50,439][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:23:50,933][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:23:51,429][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:23:51,926][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:23:52,424][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:23:52,920][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:23:53,416][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:23:53,915][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:23:54,410][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:23:54,908][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:23:55,407][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:23:55,920][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:23:56,419][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:23:56,915][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:23:57,416][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:23:57,911][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:23:58,420][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:23:58,914][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:23:59,412][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:23:59,930][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:24:00,425][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:24:00,920][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:24:01,420][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:24:01,915][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:24:02,421][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:24:02,918][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:24:03,414][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:24:03,927][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:24:04,423][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:24:04,920][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:24:05,415][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:24:05,912][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:24:06,410][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10140 tokens. -[2025-11-10 01:24:07,268][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 55.58%, Block Peak % of device VRAM: 62.14%, ΔTime: 00:00:32 -[2025-11-10 01:24:08,038][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:24:08,040][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:24:08,042][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:24:08,955][__main__][INFO] - Iteration 466 took 1m 2s (42.11% Gen, 56.42% Train). Generation: 26s, Training: 35s. Estimated remaining time: 44h 42m 5s. Estimated total time: 52h 7m 17s. Time estimates for 10 more iterations: 10m 25s, 100 more iterations: 1h 44m 14s, 500 more iterations: 8h 41m 12s. -[2025-11-10 01:24:08,958][__main__][INFO] - Starting iteration 466. -[2025-11-10 01:24:09,425][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 46 and human policies 1. -[2025-11-10 01:24:09,426][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:24:31,563][__main__][INFO] - Number of regex retries in iteration 466: 0 -[2025-11-10 01:24:31,564][__main__][INFO] - agents played in iteration 466 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:24:32,358][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:24:32,381][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:24:32,405][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:24:32,428][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:24:32,429][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:24:32,430][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:24:33,640][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:24:34,212][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:24:34,720][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:24:35,224][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:24:35,732][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:24:36,239][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:24:36,748][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:24:37,247][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:24:37,746][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:24:38,249][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:24:38,745][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:24:39,242][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:24:39,740][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:24:40,239][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:24:40,747][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:24:41,244][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:24:41,741][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:24:42,251][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:24:42,754][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:24:43,252][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:24:43,749][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:24:44,249][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:24:44,752][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:24:45,253][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:24:45,751][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:24:46,247][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:24:46,744][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:24:47,250][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:24:47,748][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:24:48,244][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:24:48,742][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:24:49,240][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:24:49,743][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:24:50,243][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:24:50,738][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:24:51,245][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:24:51,740][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:24:52,236][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:24:52,731][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:24:53,226][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:24:53,721][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:24:54,216][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:24:54,711][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:24:55,207][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:24:55,700][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:24:56,193][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:24:56,689][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:24:57,184][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:24:57,677][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:24:58,175][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:24:58,674][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:24:59,179][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:24:59,674][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:25:00,170][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:25:00,668][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:25:01,163][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:25:01,658][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:25:02,153][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:25:02,649][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:25:03,144][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:25:03,639][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:25:04,136][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:25:04,633][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:25:05,128][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:25:05,635][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10035 tokens. -[2025-11-10 01:25:06,384][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 55.63%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:32 -[2025-11-10 01:25:07,050][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:25:07,051][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:25:07,053][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:25:07,858][__main__][INFO] - Iteration 467 took 58s (37.89% Gen, 60.73% Train). Generation: 22s, Training: 35s. Estimated remaining time: 41h 15m 29s. Estimated total time: 48h 41m 40s. Time estimates for 10 more iterations: 9m 44s, 100 more iterations: 1h 37m 23s, 500 more iterations: 8h 6m 56s. -[2025-11-10 01:25:07,860][__main__][INFO] - Starting iteration 467. -[2025-11-10 01:25:08,323][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 46 and human policies 1. -[2025-11-10 01:25:08,324][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:25:39,516][__main__][INFO] - Number of regex retries in iteration 467: 0 -[2025-11-10 01:25:39,517][__main__][INFO] - agents played in iteration 467 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:25:40,290][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:25:40,315][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:25:40,340][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:25:40,362][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:25:40,362][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:25:40,363][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:25:41,102][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:25:41,556][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:25:42,059][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:25:42,568][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:25:43,072][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:25:43,573][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:25:44,080][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:25:44,580][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:25:45,079][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:25:45,577][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:25:46,077][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:25:46,578][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:25:47,077][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:25:47,574][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:25:48,074][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:25:48,574][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:25:49,079][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:25:49,576][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:25:50,076][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:25:50,570][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:25:51,067][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:25:51,574][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:25:52,076][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:25:52,574][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:25:53,074][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:25:53,571][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:25:54,066][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:25:54,561][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:25:55,058][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:25:55,570][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:25:56,066][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:25:56,566][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:25:57,061][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:25:57,555][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:25:58,067][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:25:58,562][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:25:59,058][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:25:59,556][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:26:00,052][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:26:00,551][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:26:01,045][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:26:01,541][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:26:02,052][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:26:02,550][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:26:03,049][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:26:03,548][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:26:04,050][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:26:04,561][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:26:05,060][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:26:05,556][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:26:06,071][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:26:06,567][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:26:07,064][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:26:07,560][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:26:08,055][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:26:08,561][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:26:09,055][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:26:09,549][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:26:10,047][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:26:10,541][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:26:11,036][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:26:11,530][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:26:12,024][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:26:12,521][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:26:13,015][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10116 tokens. -[2025-11-10 01:26:13,755][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.00%, Current % of VRAM taken: 55.50%, Block Peak % of device VRAM: 62.12%, ΔTime: 00:00:32 -[2025-11-10 01:26:14,512][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:26:14,514][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:26:14,516][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:26:15,438][__main__][INFO] - Iteration 468 took 1m 7s (46.48% Gen, 52.15% Train). Generation: 31s, Training: 34s. Estimated remaining time: 48h 28m 27s. Estimated total time: 55h 55m 45s. Time estimates for 10 more iterations: 11m 11s, 100 more iterations: 1h 51m 51s, 500 more iterations: 9h 19m 17s. -[2025-11-10 01:26:15,440][__main__][INFO] - Starting iteration 468. -[2025-11-10 01:26:15,921][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 46 and human policies 1. -[2025-11-10 01:26:15,921][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:26:39,438][__main__][INFO] - Number of regex retries in iteration 468: 0 -[2025-11-10 01:26:39,439][__main__][INFO] - agents played in iteration 468 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:26:40,238][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:26:40,265][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:26:40,291][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:26:40,313][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:26:40,314][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:26:40,314][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:26:41,058][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:26:41,514][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:26:42,023][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:26:42,524][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:26:43,028][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:26:43,528][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:26:44,029][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:26:44,531][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:26:45,034][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:26:45,535][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:26:46,036][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:26:46,535][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:26:47,034][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:26:47,533][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:26:48,030][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:26:48,534][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:26:49,031][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:26:49,529][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:26:50,026][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:26:50,524][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:26:51,028][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:26:51,532][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:26:52,033][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:26:52,530][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:26:53,025][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:26:53,523][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:26:54,017][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:26:54,516][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:26:55,011][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:26:55,510][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:26:56,009][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:26:56,504][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:26:56,998][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:26:57,494][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:26:57,988][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:26:58,481][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:26:58,975][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:26:59,470][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:26:59,976][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:27:00,469][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:27:00,961][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:27:01,471][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:27:01,969][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:27:02,462][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:27:02,963][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:27:03,456][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:27:03,956][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:27:04,451][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:27:04,949][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:27:05,448][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:27:05,941][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:27:06,439][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:27:06,946][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:27:07,445][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:27:07,939][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:27:08,435][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:27:08,936][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:27:09,433][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:27:09,928][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:27:10,423][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:27:10,917][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:27:11,413][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:27:11,912][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:27:12,411][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:27:12,908][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10082 tokens. -[2025-11-10 01:27:13,657][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 55.63%, Block Peak % of device VRAM: 62.36%, ΔTime: 00:00:32 -[2025-11-10 01:27:14,454][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:27:14,456][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:27:14,458][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:27:15,376][__main__][INFO] - Iteration 469 took 59s (39.55% Gen, 58.90% Train). Generation: 23s, Training: 35s. Estimated remaining time: 42h 4m 28s. Estimated total time: 49h 32m 47s. Time estimates for 10 more iterations: 9m 54s, 100 more iterations: 1h 39m 5s, 500 more iterations: 8h 15m 27s. -[2025-11-10 01:27:15,378][__main__][INFO] - Starting iteration 469. -[2025-11-10 01:27:15,868][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 46 and human policies 1. -[2025-11-10 01:27:15,869][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:27:35,978][__main__][INFO] - Number of regex retries in iteration 469: 0 -[2025-11-10 01:27:35,981][__main__][INFO] - agents played in iteration 469 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:27:36,905][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:27:36,962][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:27:36,998][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:27:37,029][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:27:37,030][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:27:37,030][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:27:37,854][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:27:38,375][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:27:38,893][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:27:39,394][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:27:39,900][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:27:40,399][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:27:40,910][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:27:41,411][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:27:41,913][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:27:42,419][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:27:42,914][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:27:43,415][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:27:43,914][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:27:44,414][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:27:44,917][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:27:45,413][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:27:45,912][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:27:46,405][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:27:46,899][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:27:47,401][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:27:47,896][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:27:48,390][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:27:48,886][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:27:49,381][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:27:49,875][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:27:50,380][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:27:50,876][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:27:51,378][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:27:51,873][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:27:52,371][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:27:52,870][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:27:53,367][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:27:53,861][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:27:54,359][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:27:54,859][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:27:55,366][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:27:55,867][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:27:56,362][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:27:56,871][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:27:57,367][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:27:57,866][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:27:58,361][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:27:58,856][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:27:59,356][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:27:59,853][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:28:00,348][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:28:00,846][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:28:01,341][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:28:01,841][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:28:02,340][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:28:02,839][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:28:03,357][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:28:03,858][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:28:04,359][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:28:04,856][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:28:05,353][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:28:05,853][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:28:06,347][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:28:06,841][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:28:07,336][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:28:07,830][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:28:08,331][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:28:08,824][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:28:09,320][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:28:09,814][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10149 tokens. -[2025-11-10 01:28:10,563][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 55.60%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:32 -[2025-11-10 01:28:11,208][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:28:11,210][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:28:11,212][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:28:12,022][__main__][INFO] - Iteration 470 took 56s (35.81% Gen, 62.74% Train). Generation: 20s, Training: 35s. Estimated remaining time: 39h 18m 29s. Estimated total time: 46h 47m 44s. Time estimates for 10 more iterations: 9m 21s, 100 more iterations: 1h 33m 35s, 500 more iterations: 7h 47m 57s. -[2025-11-10 01:28:12,025][__main__][INFO] - Starting iteration 470. -[2025-11-10 01:28:12,494][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 46 and human policies 1. -[2025-11-10 01:28:12,495][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:28:41,085][__main__][INFO] - Number of regex retries in iteration 470: 0 -[2025-11-10 01:28:41,086][__main__][INFO] - agents played in iteration 470 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:28:41,878][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:28:41,901][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:28:41,924][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:28:41,961][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:28:41,962][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:28:41,962][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:28:42,740][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:28:43,199][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:28:43,712][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:28:44,209][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:28:44,716][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:28:45,212][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:28:45,708][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:28:46,208][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:28:46,709][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:28:47,206][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:28:47,708][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:28:48,209][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:28:48,731][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:28:49,230][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:28:49,733][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:28:50,235][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:28:50,735][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:28:51,234][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:28:51,731][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:28:52,229][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:28:52,743][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:28:53,243][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:28:53,740][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:28:54,236][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:28:54,735][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:28:55,242][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:28:55,744][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:28:56,240][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:28:56,737][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:28:57,236][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:28:57,732][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:28:58,227][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:28:58,726][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:28:59,240][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:28:59,736][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:29:00,233][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:29:00,735][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:29:01,232][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:29:01,730][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:29:02,227][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:29:02,725][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:29:03,240][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:29:03,737][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:29:04,233][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:29:04,731][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:29:05,227][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:29:05,732][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:29:06,228][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:29:06,728][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:29:07,240][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:29:07,738][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:29:08,233][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:29:08,729][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:29:09,226][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:29:09,741][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:29:10,241][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:29:10,745][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:29:11,241][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:29:11,742][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:29:12,246][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:29:12,750][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:29:13,252][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:29:13,756][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:29:14,256][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:29:14,765][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10274 tokens. -[2025-11-10 01:29:15,609][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 55.63%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 -[2025-11-10 01:29:16,366][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:29:16,367][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:29:16,369][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:29:18,128][__main__][INFO] - Iteration 471 took 1m 5s (43.56% Gen, 53.76% Train). Generation: 28s, Training: 35s. Estimated remaining time: 47h 11m 22s. Estimated total time: 54h 41m 44s. Time estimates for 10 more iterations: 10m 56s, 100 more iterations: 1h 49m 23s, 500 more iterations: 9h 6m 57s. -[2025-11-10 01:29:18,130][__main__][INFO] - Starting iteration 471. -[2025-11-10 01:29:18,633][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 47 and human policies 1. -[2025-11-10 01:29:18,633][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:29:44,387][__main__][INFO] - Number of regex retries in iteration 471: 0 -[2025-11-10 01:29:44,388][__main__][INFO] - agents played in iteration 471 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:29:45,184][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:29:45,209][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:29:45,233][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:29:45,255][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:29:45,256][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:29:45,257][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:29:45,980][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:29:46,448][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:29:46,953][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:29:47,450][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:29:47,946][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:29:48,442][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:29:48,946][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:29:49,443][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:29:49,940][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:29:50,452][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:29:50,950][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:29:51,475][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:29:51,977][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:29:52,477][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:29:52,974][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:29:53,472][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:29:53,978][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:29:54,483][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:29:54,983][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:29:55,482][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:29:55,979][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:29:56,473][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:29:56,967][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:29:57,462][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:29:57,955][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:29:58,452][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:29:58,949][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:29:59,455][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:29:59,951][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:30:00,448][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:30:00,944][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:30:01,445][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:30:01,941][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:30:02,437][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:30:02,933][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:30:03,444][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:30:03,942][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:30:04,437][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:30:04,937][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:30:05,432][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:30:05,938][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:30:06,435][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:30:06,931][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:30:07,444][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:30:07,943][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:30:08,439][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:30:08,937][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:30:09,432][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:30:09,940][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:30:10,435][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:30:10,930][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:30:11,429][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:30:11,927][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:30:12,421][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:30:12,927][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:30:13,421][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:30:13,917][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:30:14,415][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:30:14,913][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:30:15,427][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:30:15,926][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:30:16,426][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:30:16,926][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:30:17,428][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:30:17,931][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10076 tokens. -[2025-11-10 01:30:18,752][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 55.59%, Block Peak % of device VRAM: 62.11%, ΔTime: 00:00:32 -[2025-11-10 01:30:19,511][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:30:19,513][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:30:19,515][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:30:20,417][__main__][INFO] - Iteration 472 took 1m 1s (41.68% Gen, 56.85% Train). Generation: 25s, Training: 35s. Estimated remaining time: 43h 57m 51s. Estimated total time: 51h 29m 14s. Time estimates for 10 more iterations: 10m 17s, 100 more iterations: 1h 42m 58s, 500 more iterations: 8h 34m 52s. -[2025-11-10 01:30:20,419][__main__][INFO] - Starting iteration 472. -[2025-11-10 01:30:20,886][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 47 and human policies 1. -[2025-11-10 01:30:20,887][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:30:47,249][__main__][INFO] - Number of regex retries in iteration 472: 0 -[2025-11-10 01:30:47,251][__main__][INFO] - agents played in iteration 472 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:30:48,056][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:30:48,083][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:30:48,109][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:30:48,132][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:30:48,133][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:30:48,133][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:30:48,921][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:30:49,386][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:30:49,889][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:30:50,392][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:30:50,896][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:30:51,394][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:30:51,892][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:30:52,387][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:30:52,882][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:30:53,380][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:30:53,882][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:30:54,376][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:30:54,872][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:30:55,371][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:30:55,877][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:30:56,373][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:30:56,869][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:30:57,366][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:30:57,861][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:30:58,365][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:30:58,861][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:30:59,357][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:30:59,854][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:31:00,354][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:31:00,852][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:31:01,353][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:31:01,850][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:31:02,359][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:31:02,856][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:31:03,353][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:31:03,856][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:31:04,356][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:31:04,861][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:31:05,357][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:31:05,853][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:31:06,385][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:31:06,883][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:31:07,383][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:31:07,879][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:31:08,378][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:31:08,875][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:31:09,373][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:31:09,869][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:31:10,369][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:31:10,867][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:31:11,359][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:31:11,855][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:31:12,350][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:31:12,861][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:31:13,356][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:31:13,851][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:31:14,354][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:31:14,849][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:31:15,352][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:31:15,851][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:31:16,354][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:31:16,856][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:31:17,354][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:31:17,852][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:31:18,352][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:31:18,852][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:31:19,366][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:31:19,867][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:31:20,371][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:31:20,872][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10105 tokens. -[2025-11-10 01:31:21,711][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 55.61%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:32 -[2025-11-10 01:31:22,371][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:31:22,373][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:31:22,375][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:31:23,224][__main__][INFO] - Iteration 473 took 1m 2s (42.29% Gen, 56.34% Train). Generation: 26s, Training: 35s. Estimated remaining time: 44h 24m 30s. Estimated total time: 51h 56m 56s. Time estimates for 10 more iterations: 10m 23s, 100 more iterations: 1h 43m 53s, 500 more iterations: 8h 39m 29s. -[2025-11-10 01:31:23,226][__main__][INFO] - Starting iteration 473. -[2025-11-10 01:31:23,689][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 47 and human policies 1. -[2025-11-10 01:31:23,689][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:31:49,107][__main__][INFO] - Number of regex retries in iteration 473: 0 -[2025-11-10 01:31:49,107][__main__][INFO] - agents played in iteration 473 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:31:49,929][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:31:49,956][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:31:49,983][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:31:50,006][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:31:50,007][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:31:50,007][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:31:50,791][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:31:51,252][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:31:51,757][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:31:52,271][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:31:52,774][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:31:53,276][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:31:53,781][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:31:54,284][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:31:54,789][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:31:55,293][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:31:55,798][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:31:56,297][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:31:56,797][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:31:57,297][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:31:57,797][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:31:58,294][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:31:58,800][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:31:59,301][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:31:59,798][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:32:00,298][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:32:00,795][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:32:01,299][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:32:01,795][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:32:02,295][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:32:02,794][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:32:03,296][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:32:03,806][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:32:04,307][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:32:04,805][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:32:05,310][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:32:05,809][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:32:06,311][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:32:06,815][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:32:07,313][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:32:07,812][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:32:08,311][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:32:08,808][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:32:09,306][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:32:09,805][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:32:10,303][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:32:10,800][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:32:11,294][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:32:11,789][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:32:12,287][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:32:12,785][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:32:13,296][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:32:13,792][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:32:14,286][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:32:14,790][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:32:15,289][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:32:15,795][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:32:16,292][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:32:16,790][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:32:17,286][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:32:17,785][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:32:18,282][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:32:18,779][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:32:19,274][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:32:19,794][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:32:20,293][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:32:20,794][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:32:21,292][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:32:21,794][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:32:22,296][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:32:22,793][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10100 tokens. -[2025-11-10 01:32:23,637][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 55.60%, Block Peak % of device VRAM: 62.23%, ΔTime: 00:00:32 -[2025-11-10 01:32:24,375][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:32:24,376][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:32:24,378][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:32:25,269][__main__][INFO] - Iteration 474 took 1m 1s (41.28% Gen, 57.28% Train). Generation: 25s, Training: 35s. Estimated remaining time: 43h 45m 34s. Estimated total time: 51h 19m 3s. Time estimates for 10 more iterations: 10m 15s, 100 more iterations: 1h 42m 38s, 500 more iterations: 8h 33m 10s. -[2025-11-10 01:32:25,271][__main__][INFO] - Starting iteration 474. -[2025-11-10 01:32:25,736][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 47 and human policies 1. -[2025-11-10 01:32:25,736][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:32:48,809][__main__][INFO] - Number of regex retries in iteration 474: 0 -[2025-11-10 01:32:48,811][__main__][INFO] - agents played in iteration 474 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:32:49,613][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:32:49,642][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:32:49,667][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:32:49,693][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:32:49,693][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:32:49,694][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:32:50,525][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:32:50,989][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:32:51,510][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:32:52,017][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:32:52,522][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:32:53,026][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:32:53,532][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:32:54,039][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:32:54,543][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:32:55,044][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:32:55,547][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:32:56,048][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:32:56,548][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:32:57,052][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:32:57,552][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:32:58,058][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:32:58,560][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:32:59,059][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:32:59,571][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:33:00,068][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:33:00,569][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:33:01,069][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:33:01,565][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:33:02,066][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:33:02,562][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:33:03,062][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:33:03,573][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:33:04,071][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:33:04,569][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:33:05,078][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:33:05,574][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:33:06,096][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:33:06,593][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:33:07,091][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:33:07,592][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:33:08,090][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:33:08,598][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:33:09,097][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:33:09,597][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:33:10,097][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:33:10,595][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:33:11,092][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:33:11,596][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:33:12,096][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:33:12,604][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:33:13,103][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:33:13,602][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:33:14,097][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:33:14,596][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:33:15,101][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:33:15,597][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:33:16,093][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:33:16,591][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:33:17,093][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:33:17,592][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:33:18,094][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:33:18,591][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:33:19,110][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:33:19,613][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:33:20,116][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:33:20,618][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:33:21,123][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:33:21,628][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:33:22,128][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:33:22,629][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10228 tokens. -[2025-11-10 01:33:23,470][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 55.66%, Block Peak % of device VRAM: 62.37%, ΔTime: 00:00:32 -[2025-11-10 01:33:24,123][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:33:24,125][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:33:24,127][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:33:24,935][__main__][INFO] - Iteration 475 took 59s (38.98% Gen, 59.66% Train). Generation: 23s, Training: 35s. Estimated remaining time: 41h 45m 31s. Estimated total time: 49h 19m 59s. Time estimates for 10 more iterations: 9m 51s, 100 more iterations: 1h 38m 39s, 500 more iterations: 8h 13m 19s. -[2025-11-10 01:33:24,937][__main__][INFO] - Starting iteration 475. -[2025-11-10 01:33:25,428][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 47 and human policies 1. -[2025-11-10 01:33:25,429][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:33:57,820][__main__][INFO] - Number of regex retries in iteration 475: 0 -[2025-11-10 01:33:57,821][__main__][INFO] - agents played in iteration 475 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:33:58,638][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:33:58,666][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:33:58,693][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:33:58,717][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:33:58,717][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:33:58,718][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:33:59,447][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:33:59,916][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:34:00,424][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:34:00,928][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:34:01,429][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:34:01,927][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:34:02,427][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:34:02,924][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:34:03,425][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:34:03,923][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:34:04,420][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:34:04,917][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:34:05,415][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:34:05,912][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:34:06,427][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:34:06,924][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:34:07,425][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:34:07,922][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:34:08,418][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:34:08,921][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:34:09,421][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:34:09,922][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:34:10,428][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:34:10,926][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:34:11,421][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:34:11,917][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:34:12,413][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:34:12,912][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:34:13,409][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:34:13,906][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:34:14,403][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:34:14,900][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:34:15,396][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:34:15,895][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:34:16,391][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:34:16,909][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:34:17,404][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:34:17,902][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:34:18,399][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:34:18,895][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:34:19,401][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:34:19,898][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:34:20,394][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:34:20,892][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:34:21,388][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:34:21,885][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:34:22,389][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:34:22,889][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:34:23,397][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:34:23,894][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:34:24,391][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:34:24,890][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:34:25,385][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:34:25,882][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:34:26,376][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:34:26,877][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:34:27,389][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:34:27,890][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:34:28,390][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:34:28,891][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:34:29,390][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:34:29,885][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:34:30,384][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:34:30,882][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:34:31,398][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10079 tokens. -[2025-11-10 01:34:32,201][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.05%, Current % of VRAM taken: 55.54%, Block Peak % of device VRAM: 62.17%, ΔTime: 00:00:32 -[2025-11-10 01:34:32,964][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:34:32,967][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:34:32,968][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:34:33,872][__main__][INFO] - Iteration 476 took 1m 8s (47.32% Gen, 51.35% Train). Generation: 32s, Training: 35s. Estimated remaining time: 49h 26m 36s. Estimated total time: 57h 2m 13s. Time estimates for 10 more iterations: 11m 24s, 100 more iterations: 1h 54m 4s, 500 more iterations: 9h 30m 22s. -[2025-11-10 01:34:33,874][__main__][INFO] - Starting iteration 476. -[2025-11-10 01:34:34,347][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 47 and human policies 1. -[2025-11-10 01:34:34,348][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:34:58,477][__main__][INFO] - Number of regex retries in iteration 476: 0 -[2025-11-10 01:34:58,480][__main__][INFO] - agents played in iteration 476 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:34:59,338][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:34:59,376][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:34:59,403][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:34:59,426][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:34:59,427][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:34:59,428][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:35:00,170][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:35:00,624][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:35:01,127][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:35:01,628][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:35:02,123][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:35:02,621][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:35:03,114][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:35:03,611][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:35:04,108][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:35:04,606][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:35:05,103][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:35:05,604][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:35:06,100][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:35:06,597][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:35:07,092][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:35:07,590][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:35:08,113][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:35:08,643][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:35:09,140][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:35:09,639][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:35:10,136][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:35:10,636][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:35:11,137][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:35:11,634][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:35:12,135][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:35:12,633][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:35:13,130][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:35:13,626][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:35:14,124][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:35:14,627][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:35:15,126][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:35:15,622][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:35:16,126][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:35:16,627][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:35:17,125][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:35:17,618][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:35:18,113][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:35:18,625][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:35:19,120][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:35:19,615][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:35:20,114][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:35:20,612][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:35:21,109][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:35:21,614][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:35:22,113][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:35:22,610][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:35:23,108][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:35:23,606][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:35:24,104][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:35:24,602][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:35:25,096][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:35:25,593][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:35:26,091][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:35:26,595][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:35:27,090][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:35:27,585][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:35:28,093][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:35:28,589][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:35:29,091][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:35:29,590][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:35:30,088][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:35:30,591][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:35:31,087][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:35:31,588][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:35:32,092][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10030 tokens. -[2025-11-10 01:35:32,915][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 55.55%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:32 -[2025-11-10 01:35:33,554][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:35:33,556][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:35:33,558][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:35:34,370][__main__][INFO] - Iteration 477 took 1m 0s (40.20% Gen, 58.44% Train). Generation: 24s, Training: 35s. Estimated remaining time: 42h 24m 32s. Estimated total time: 50h 1m 9s. Time estimates for 10 more iterations: 10m 0s, 100 more iterations: 1h 40m 2s, 500 more iterations: 8h 20m 11s. -[2025-11-10 01:35:34,372][__main__][INFO] - Starting iteration 477. -[2025-11-10 01:35:34,836][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 47 and human policies 1. -[2025-11-10 01:35:34,837][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:36:05,125][__main__][INFO] - Number of regex retries in iteration 477: 0 -[2025-11-10 01:36:05,126][__main__][INFO] - agents played in iteration 477 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:36:05,897][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:36:05,922][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:36:05,947][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:36:05,970][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:36:05,970][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:36:05,971][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:36:06,700][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:36:07,158][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:36:07,665][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:36:08,164][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:36:08,665][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:36:09,159][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:36:09,658][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:36:10,160][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:36:10,658][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:36:11,159][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:36:11,657][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:36:12,158][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:36:12,654][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:36:13,155][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:36:13,653][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:36:14,154][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:36:14,652][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:36:15,153][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:36:15,657][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:36:16,157][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:36:16,656][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:36:17,157][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:36:17,652][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:36:18,149][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:36:18,648][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:36:19,143][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:36:19,640][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:36:20,136][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:36:20,632][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:36:21,126][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:36:21,621][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:36:22,127][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:36:22,626][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:36:23,122][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:36:23,619][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:36:24,116][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:36:24,615][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:36:25,112][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:36:25,609][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:36:26,105][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:36:26,601][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:36:27,099][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:36:27,596][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:36:28,097][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:36:28,594][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:36:29,092][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:36:29,593][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:36:30,109][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:36:30,608][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:36:31,103][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:36:31,605][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:36:32,100][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:36:32,609][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:36:33,104][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:36:33,599][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:36:34,097][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:36:34,596][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:36:35,097][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:36:35,605][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:36:36,108][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:36:36,620][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:36:37,121][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:36:37,621][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:36:38,123][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:36:38,623][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10194 tokens. -[2025-11-10 01:36:39,441][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 55.67%, Block Peak % of device VRAM: 62.34%, ΔTime: 00:00:32 -[2025-11-10 01:36:40,175][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:36:40,177][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:36:40,179][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:36:41,094][__main__][INFO] - Iteration 478 took 1m 6s (45.71% Gen, 52.90% Train). Generation: 30s, Training: 35s. Estimated remaining time: 47h 35m 12s. Estimated total time: 55h 12m 56s. Time estimates for 10 more iterations: 11m 2s, 100 more iterations: 1h 50m 25s, 500 more iterations: 9h 12m 9s. -[2025-11-10 01:36:41,097][__main__][INFO] - Starting iteration 478. -[2025-11-10 01:36:41,575][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 47 and human policies 1. -[2025-11-10 01:36:41,575][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:37:05,484][__main__][INFO] - Number of regex retries in iteration 478: 0 -[2025-11-10 01:37:05,484][__main__][INFO] - agents played in iteration 478 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:37:06,291][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:37:06,317][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:37:06,341][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:37:06,364][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:37:06,365][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:37:06,366][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:37:07,113][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:37:07,565][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:37:08,068][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:37:08,573][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:37:09,070][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:37:09,569][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:37:10,065][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:37:10,566][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:37:11,062][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:37:11,564][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:37:12,064][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:37:12,564][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:37:13,065][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:37:13,565][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:37:14,068][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:37:14,566][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:37:15,062][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:37:15,556][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:37:16,051][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:37:16,549][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:37:17,044][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:37:17,540][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:37:18,031][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:37:18,526][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:37:19,035][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:37:19,531][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:37:20,024][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:37:20,519][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:37:21,012][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:37:21,516][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:37:22,013][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:37:22,511][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:37:23,026][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:37:23,524][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:37:24,019][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:37:24,514][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:37:25,012][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:37:25,509][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:37:26,005][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:37:26,502][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:37:26,996][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:37:27,491][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:37:27,983][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:37:28,475][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:37:28,969][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:37:29,468][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:37:29,960][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:37:30,453][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:37:30,949][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:37:31,446][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:37:31,938][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:37:32,431][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:37:32,924][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:37:33,436][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:37:33,932][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:37:34,424][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:37:34,917][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:37:35,411][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:37:35,920][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:37:36,414][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:37:36,911][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:37:37,420][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:37:37,915][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:37:38,414][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:37:38,918][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10143 tokens. -[2025-11-10 01:37:39,712][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 55.60%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:32 -[2025-11-10 01:37:40,448][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:37:40,450][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:37:40,451][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:37:41,347][__main__][INFO] - Iteration 479 took 59s (40.00% Gen, 58.50% Train). Generation: 23s, Training: 34s. Estimated remaining time: 42h 9m 54s. Estimated total time: 49h 48m 39s. Time estimates for 10 more iterations: 9m 57s, 100 more iterations: 1h 39m 37s, 500 more iterations: 8h 18m 6s. -[2025-11-10 01:37:41,350][__main__][INFO] - Starting iteration 479. -[2025-11-10 01:37:41,819][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 47 and human policies 1. -[2025-11-10 01:37:41,820][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:38:02,947][__main__][INFO] - Number of regex retries in iteration 479: 0 -[2025-11-10 01:38:02,949][__main__][INFO] - agents played in iteration 479 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:38:03,745][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:38:03,768][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:38:03,791][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:38:03,814][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:38:03,815][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:38:03,815][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:38:04,635][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:38:05,093][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:38:05,597][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:38:06,099][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:38:06,605][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:38:07,105][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:38:07,608][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:38:08,120][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:38:08,622][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:38:09,135][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:38:09,635][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:38:10,134][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:38:10,632][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:38:11,128][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:38:11,639][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:38:12,140][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:38:12,635][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:38:13,151][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:38:13,649][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:38:14,146][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:38:14,648][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:38:15,145][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:38:15,649][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:38:16,146][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:38:16,641][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:38:17,153][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:38:17,649][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:38:18,144][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:38:18,641][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:38:19,140][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:38:19,644][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:38:20,139][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:38:20,634][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:38:21,141][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:38:21,643][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:38:22,169][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:38:22,671][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:38:23,169][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:38:23,668][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:38:24,163][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:38:24,661][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:38:25,160][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:38:25,656][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:38:26,159][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:38:26,656][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:38:27,153][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:38:27,656][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:38:28,155][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:38:28,655][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:38:29,153][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:38:29,654][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:38:30,152][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:38:30,650][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:38:31,148][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:38:31,647][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:38:32,145][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:38:32,644][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:38:33,158][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:38:33,655][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:38:34,157][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:38:34,660][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:38:35,162][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:38:35,667][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:38:36,162][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:38:36,666][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10061 tokens. -[2025-11-10 01:38:37,511][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 55.72%, Block Peak % of device VRAM: 62.34%, ΔTime: 00:00:32 -[2025-11-10 01:38:38,153][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:38:38,154][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:38:38,156][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:38:38,989][__main__][INFO] - Iteration 480 took 57s (36.96% Gen, 61.58% Train). Generation: 21s, Training: 35s. Estimated remaining time: 39h 58m 50s. Estimated total time: 47h 38m 32s. Time estimates for 10 more iterations: 9m 31s, 100 more iterations: 1h 35m 17s, 500 more iterations: 7h 56m 25s. -[2025-11-10 01:38:38,992][__main__][INFO] - Starting iteration 480. -[2025-11-10 01:38:39,476][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 47 and human policies 1. -[2025-11-10 01:38:39,477][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:39:10,245][__main__][INFO] - Number of regex retries in iteration 480: 0 -[2025-11-10 01:39:10,246][__main__][INFO] - agents played in iteration 480 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:39:11,033][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:39:11,059][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:39:11,085][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:39:11,108][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:39:11,109][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:39:11,110][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:39:11,840][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:39:12,295][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:39:12,798][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:39:13,294][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:39:13,794][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:39:14,295][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:39:14,793][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:39:15,289][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:39:15,785][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:39:16,287][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:39:16,782][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:39:17,277][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:39:17,779][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:39:18,273][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:39:18,771][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:39:19,263][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:39:19,761][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:39:20,261][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:39:20,757][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:39:21,252][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:39:21,764][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:39:22,258][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:39:22,755][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:39:23,251][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:39:23,749][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:39:24,262][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:39:24,756][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:39:25,249][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:39:25,744][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:39:26,236][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:39:26,733][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:39:27,230][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:39:27,723][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:39:28,220][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:39:28,719][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:39:29,218][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:39:29,854][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:39:30,354][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:39:30,868][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:39:31,364][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:39:31,860][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:39:32,359][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:39:32,854][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:39:33,357][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:39:33,853][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:39:34,356][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:39:34,868][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:39:35,363][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:39:35,890][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:39:36,383][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:39:36,878][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:39:37,379][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:39:37,875][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:39:38,375][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:39:38,872][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:39:39,368][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:39:39,868][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:39:40,364][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:39:40,860][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:39:41,357][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:39:41,852][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:39:42,351][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:39:42,851][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:39:43,349][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:39:43,867][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9962 tokens. -[2025-11-10 01:39:44,661][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 55.61%, Block Peak % of device VRAM: 62.06%, ΔTime: 00:00:32 -[2025-11-10 01:39:45,407][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:39:45,408][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:39:45,411][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:39:47,098][__main__][INFO] - Iteration 481 took 1m 7s (45.50% Gen, 52.00% Train). Generation: 30s, Training: 35s. Estimated remaining time: 48h 40m 19s. Estimated total time: 56h 21m 9s. Time estimates for 10 more iterations: 11m 16s, 100 more iterations: 1h 52m 42s, 500 more iterations: 9h 23m 31s. -[2025-11-10 01:39:47,101][__main__][INFO] - Starting iteration 481. -[2025-11-10 01:39:47,567][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 48 and human policies 1. -[2025-11-10 01:39:47,568][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:40:07,789][__main__][INFO] - Number of regex retries in iteration 481: 0 -[2025-11-10 01:40:07,791][__main__][INFO] - agents played in iteration 481 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:40:08,679][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:40:08,735][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:40:08,777][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:40:08,809][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:40:08,810][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:40:08,811][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:40:09,644][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:40:10,187][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:40:10,702][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:40:11,207][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:40:11,706][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:40:12,212][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:40:12,718][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:40:13,218][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:40:13,716][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:40:14,212][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:40:14,714][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:40:15,214][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:40:15,712][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:40:16,211][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:40:16,709][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:40:17,207][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:40:17,704][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:40:18,201][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:40:18,700][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:40:19,198][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:40:19,699][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:40:20,196][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:40:20,694][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:40:21,192][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:40:21,691][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:40:22,191][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:40:22,695][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:40:23,194][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:40:23,690][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:40:24,187][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:40:24,680][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:40:25,178][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:40:25,678][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:40:26,173][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:40:26,670][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:40:27,183][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:40:27,681][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:40:28,175][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:40:28,672][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:40:29,169][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:40:29,678][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:40:30,176][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:40:30,673][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:40:31,174][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:40:31,673][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:40:32,170][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:40:32,668][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:40:33,165][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:40:33,679][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:40:34,173][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:40:34,668][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:40:35,165][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:40:35,658][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:40:36,161][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:40:36,656][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:40:37,150][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:40:37,646][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:40:38,141][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:40:38,637][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:40:39,132][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:40:39,627][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:40:40,124][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:40:40,619][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:40:41,117][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:40:41,615][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10052 tokens. -[2025-11-10 01:40:42,378][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 55.59%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 -[2025-11-10 01:40:43,034][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:40:43,035][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:40:43,037][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:40:43,860][__main__][INFO] - Iteration 482 took 56s (35.92% Gen, 62.61% Train). Generation: 20s, Training: 35s. Estimated remaining time: 39h 12m 54s. Estimated total time: 46h 54m 41s. Time estimates for 10 more iterations: 9m 22s, 100 more iterations: 1h 33m 49s, 500 more iterations: 7h 49m 6s. -[2025-11-10 01:40:43,865][__main__][INFO] - Starting iteration 482. -[2025-11-10 01:40:44,342][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 48 and human policies 1. -[2025-11-10 01:40:44,342][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:41:17,542][__main__][INFO] - Number of regex retries in iteration 482: 0 -[2025-11-10 01:41:17,542][__main__][INFO] - agents played in iteration 482 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:41:18,337][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:41:18,362][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:41:18,387][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:41:18,409][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:41:18,410][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:41:18,411][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:41:19,140][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:41:19,593][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:41:20,097][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:41:20,592][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:41:21,085][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:41:21,579][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:41:22,073][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:41:22,573][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:41:23,072][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:41:23,572][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:41:24,073][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:41:24,578][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:41:25,087][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:41:25,589][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:41:26,091][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:41:26,600][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:41:27,100][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:41:27,599][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:41:28,101][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:41:28,603][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:41:29,107][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:41:29,605][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:41:30,103][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:41:30,600][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:41:31,097][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:41:31,602][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:41:32,097][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:41:32,594][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:41:33,095][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:41:33,599][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:41:34,098][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:41:34,605][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:41:35,105][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:41:35,601][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:41:36,093][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:41:36,590][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:41:37,102][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:41:37,596][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:41:38,094][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:41:38,620][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:41:39,115][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:41:39,614][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:41:40,109][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:41:40,604][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:41:41,102][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:41:41,597][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:41:42,093][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:41:42,593][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:41:43,090][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:41:43,590][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:41:44,090][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:41:44,588][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:41:45,098][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:41:45,597][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:41:46,094][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:41:46,601][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:41:47,098][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:41:47,592][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:41:48,090][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:41:48,587][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:41:49,084][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:41:49,579][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:41:50,083][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:41:50,606][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:41:51,110][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10151 tokens. -[2025-11-10 01:41:51,904][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 55.64%, Block Peak % of device VRAM: 62.34%, ΔTime: 00:00:32 -[2025-11-10 01:41:52,644][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:41:52,646][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:41:52,647][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:41:53,595][__main__][INFO] - Iteration 483 took 1m 9s (47.94% Gen, 50.69% Train). Generation: 33s, Training: 35s. Estimated remaining time: 49h 59m 43s. Estimated total time: 57h 42m 40s. Time estimates for 10 more iterations: 11m 32s, 100 more iterations: 1h 55m 25s, 500 more iterations: 9h 37m 6s. -[2025-11-10 01:41:53,598][__main__][INFO] - Starting iteration 483. -[2025-11-10 01:41:54,060][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 48 and human policies 1. -[2025-11-10 01:41:54,061][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:42:15,719][__main__][INFO] - Number of regex retries in iteration 483: 0 -[2025-11-10 01:42:15,720][__main__][INFO] - agents played in iteration 483 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:42:16,501][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:42:16,524][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:42:16,546][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:42:16,568][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:42:16,569][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:42:16,570][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:42:17,352][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:42:17,810][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:42:18,318][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:42:18,819][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:42:19,328][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:42:19,827][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:42:20,328][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:42:20,829][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:42:21,331][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:42:21,832][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:42:22,326][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:42:22,822][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:42:23,321][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:42:23,818][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:42:24,315][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:42:24,814][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:42:25,314][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:42:25,820][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:42:26,322][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:42:26,823][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:42:27,323][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:42:27,821][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:42:28,321][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:42:28,820][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:42:29,318][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:42:29,816][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:42:30,313][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:42:30,808][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:42:31,308][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:42:31,809][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:42:32,319][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:42:32,819][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:42:33,313][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:42:33,810][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:42:34,307][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:42:34,816][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:42:35,312][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:42:35,809][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:42:36,308][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:42:36,804][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:42:37,300][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:42:37,794][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:42:38,290][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:42:38,812][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:42:39,305][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:42:39,802][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:42:40,297][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:42:40,794][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:42:41,288][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:42:41,783][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:42:42,276][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:42:42,775][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:42:43,269][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:42:43,763][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:42:44,260][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:42:44,754][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:42:45,253][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:42:45,751][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:42:46,248][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:42:46,750][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:42:47,250][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:42:47,747][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:42:48,246][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:42:48,745][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:42:49,246][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10146 tokens. -[2025-11-10 01:42:49,992][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 55.68%, Block Peak % of device VRAM: 62.27%, ΔTime: 00:00:32 -[2025-11-10 01:42:50,767][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:42:50,769][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:42:50,770][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:42:51,675][__main__][INFO] - Iteration 484 took 57s (37.59% Gen, 60.84% Train). Generation: 21s, Training: 35s. Estimated remaining time: 40h 16m 50s. Estimated total time: 48h 0m 45s. Time estimates for 10 more iterations: 9m 36s, 100 more iterations: 1h 36m 1s, 500 more iterations: 8h 0m 7s. -[2025-11-10 01:42:51,677][__main__][INFO] - Starting iteration 484. -[2025-11-10 01:42:52,146][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 48 and human policies 1. -[2025-11-10 01:42:52,146][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:43:19,007][__main__][INFO] - Number of regex retries in iteration 484: 0 -[2025-11-10 01:43:19,010][__main__][INFO] - agents played in iteration 484 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:43:19,842][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:43:19,892][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:43:19,927][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:43:19,957][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:43:19,957][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:43:19,958][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:43:20,692][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:43:21,155][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:43:21,666][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:43:22,164][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:43:22,666][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:43:23,167][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:43:23,666][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:43:24,164][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:43:24,661][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:43:25,164][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:43:25,663][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:43:26,164][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:43:26,665][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:43:27,163][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:43:27,665][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:43:28,162][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:43:28,660][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:43:29,158][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:43:29,661][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:43:30,157][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:43:30,652][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:43:31,151][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:43:31,652][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:43:32,155][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:43:32,655][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:43:33,154][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:43:33,652][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:43:34,151][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:43:34,646][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:43:35,141][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:43:35,640][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:43:36,136][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:43:36,635][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:43:37,132][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:43:37,626][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:43:38,120][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:43:38,614][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:43:39,108][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:43:39,626][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:43:40,124][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:43:40,617][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:43:41,115][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:43:41,616][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:43:42,114][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:43:42,610][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:43:43,109][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:43:43,618][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:43:44,117][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:43:44,616][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:43:45,118][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:43:45,612][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:43:46,109][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:43:46,604][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:43:47,098][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:43:47,602][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:43:48,097][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:43:48,596][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:43:49,108][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:43:49,603][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:43:50,104][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:43:50,599][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:43:51,094][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:43:51,599][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:43:52,097][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:43:52,597][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10189 tokens. -[2025-11-10 01:43:53,342][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 -[2025-11-10 01:43:53,988][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:43:53,989][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:43:53,991][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:43:54,795][__main__][INFO] - Iteration 485 took 1m 2s (42.88% Gen, 55.84% Train). Generation: 26s, Training: 34s. Estimated remaining time: 44h 27m 31s. Estimated total time: 52h 12m 28s. Time estimates for 10 more iterations: 10m 26s, 100 more iterations: 1h 44m 24s, 500 more iterations: 8h 42m 4s. -[2025-11-10 01:43:54,797][__main__][INFO] - Starting iteration 485. -[2025-11-10 01:43:55,269][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 48 and human policies 1. -[2025-11-10 01:43:55,270][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:44:10,030][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 01:44:16,855][__main__][INFO] - Number of regex retries in iteration 485: 1 -[2025-11-10 01:44:16,856][__main__][INFO] - agents played in iteration 485 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:44:17,670][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:44:17,693][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:44:17,716][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:44:17,739][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:44:17,740][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:44:17,741][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:44:18,563][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:44:19,027][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:44:19,537][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:44:20,042][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:44:20,540][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:44:21,040][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:44:21,554][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:44:22,059][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:44:22,561][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:44:23,060][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:44:23,558][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:44:24,058][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:44:24,561][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:44:25,060][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:44:25,563][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:44:26,064][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:44:26,564][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:44:27,065][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:44:27,566][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:44:28,065][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:44:28,567][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:44:29,067][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:44:29,568][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:44:30,065][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:44:30,565][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:44:31,069][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:44:31,566][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:44:32,065][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:44:32,563][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:44:33,064][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:44:33,571][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:44:34,069][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:44:34,566][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:44:35,064][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:44:35,560][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:44:36,056][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:44:36,552][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:44:37,048][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:44:37,544][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:44:38,041][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:44:38,539][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:44:39,038][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:44:39,535][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:44:40,054][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:44:40,553][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:44:41,051][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:44:41,552][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:44:42,048][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:44:42,547][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:44:43,044][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:44:43,540][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:44:44,056][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:44:44,551][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:44:45,052][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:44:45,547][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:44:46,042][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:44:46,537][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:44:47,036][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:44:47,535][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:44:48,058][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:44:48,561][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:44:49,063][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:44:49,561][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:44:50,059][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:44:50,564][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10184 tokens. -[2025-11-10 01:44:51,304][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 55.72%, Block Peak % of device VRAM: 62.36%, ΔTime: 00:00:32 -[2025-11-10 01:44:52,066][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:44:52,068][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:44:52,070][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:44:52,995][__main__][INFO] - Iteration 486 took 57s (37.39% Gen, 61.00% Train). Generation: 21s, Training: 35s. Estimated remaining time: 40h 20m 23s. Estimated total time: 48h 6m 19s. Time estimates for 10 more iterations: 9m 37s, 100 more iterations: 1h 36m 12s, 500 more iterations: 8h 1m 3s. -[2025-11-10 01:44:52,997][__main__][INFO] - Starting iteration 486. -[2025-11-10 01:44:53,459][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 48 and human policies 1. -[2025-11-10 01:44:53,460][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:45:15,563][__main__][INFO] - Number of regex retries in iteration 486: 0 -[2025-11-10 01:45:15,566][__main__][INFO] - agents played in iteration 486 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:45:16,517][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:45:16,571][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:45:16,609][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:45:16,642][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:45:16,642][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:45:16,643][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:45:17,482][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:45:18,032][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:45:18,540][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:45:19,047][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:45:19,552][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:45:20,061][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:45:20,571][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:45:21,076][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:45:21,580][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:45:22,087][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:45:22,586][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:45:23,092][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:45:23,593][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:45:24,095][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:45:24,597][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:45:25,097][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:45:25,597][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:45:26,098][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:45:26,594][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:45:27,093][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:45:27,590][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:45:28,091][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:45:28,587][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:45:29,085][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:45:29,584][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:45:30,086][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:45:30,586][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:45:31,085][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:45:31,585][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:45:32,090][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:45:32,609][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:45:33,109][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:45:33,632][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:45:34,127][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:45:34,623][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:45:35,125][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:45:35,621][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:45:36,130][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:45:36,627][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:45:37,121][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:45:37,628][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:45:38,132][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:45:38,647][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:45:39,146][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:45:39,644][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:45:40,157][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:45:40,655][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:45:41,152][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:45:41,650][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:45:42,145][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:45:42,642][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:45:43,137][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:45:43,633][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:45:44,130][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:45:44,625][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:45:45,124][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:45:45,628][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:45:46,125][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:45:46,626][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:45:47,123][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:45:47,619][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:45:48,118][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:45:48,616][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:45:49,113][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:45:49,610][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10222 tokens. -[2025-11-10 01:45:50,387][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 55.69%, Block Peak % of device VRAM: 62.50%, ΔTime: 00:00:32 -[2025-11-10 01:45:51,028][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:45:51,030][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:45:51,032][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:45:51,854][__main__][INFO] - Iteration 487 took 58s (37.85% Gen, 60.73% Train). Generation: 22s, Training: 35s. Estimated remaining time: 40h 52m 51s. Estimated total time: 48h 39m 45s. Time estimates for 10 more iterations: 9m 43s, 100 more iterations: 1h 37m 19s, 500 more iterations: 8h 6m 37s. -[2025-11-10 01:45:51,857][__main__][INFO] - Starting iteration 487. -[2025-11-10 01:45:52,336][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 48 and human policies 1. -[2025-11-10 01:45:52,337][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:46:26,723][__main__][INFO] - Number of regex retries in iteration 487: 0 -[2025-11-10 01:46:26,724][__main__][INFO] - agents played in iteration 487 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:46:27,576][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:46:27,608][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:46:27,635][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:46:27,657][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:46:27,659][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:46:27,659][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:46:28,430][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:46:28,890][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:46:29,398][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:46:29,897][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:46:30,394][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:46:30,891][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:46:31,387][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:46:31,888][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:46:32,385][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:46:32,882][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:46:33,377][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:46:33,879][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:46:34,382][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:46:34,877][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:46:35,379][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:46:35,876][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:46:36,376][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:46:36,872][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:46:37,373][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:46:37,877][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:46:38,373][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:46:38,880][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:46:39,379][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:46:39,878][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:46:40,376][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:46:40,876][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:46:41,377][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:46:41,878][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:46:42,376][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:46:42,871][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:46:43,372][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:46:43,870][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:46:44,368][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:46:44,869][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:46:45,365][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:46:45,865][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:46:46,360][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:46:46,854][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:46:47,365][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:46:47,857][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:46:48,356][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:46:48,851][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:46:49,346][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:46:49,857][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:46:50,354][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:46:50,851][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:46:51,347][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:46:51,847][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:46:52,345][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:46:52,840][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:46:53,333][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:46:53,830][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:46:54,323][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:46:54,816][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:46:55,313][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:46:55,806][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:46:56,297][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:46:56,794][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:46:57,291][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:46:57,789][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:46:58,288][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:46:58,810][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:46:59,309][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:46:59,812][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:47:00,317][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10182 tokens. -[2025-11-10 01:47:01,107][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 55.71%, Block Peak % of device VRAM: 62.22%, ΔTime: 00:00:32 -[2025-11-10 01:47:01,851][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:47:01,855][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:47:01,857][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:47:02,726][__main__][INFO] - Iteration 488 took 1m 10s (48.85% Gen, 49.91% Train). Generation: 34s, Training: 35s. Estimated remaining time: 50h 51m 26s. Estimated total time: 58h 39m 32s. Time estimates for 10 more iterations: 11m 43s, 100 more iterations: 1h 57m 19s, 500 more iterations: 9h 46m 35s. -[2025-11-10 01:47:02,729][__main__][INFO] - Starting iteration 488. -[2025-11-10 01:47:03,200][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 48 and human policies 1. -[2025-11-10 01:47:03,201][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:47:16,815][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 01:47:28,283][__main__][INFO] - Number of regex retries in iteration 488: 1 -[2025-11-10 01:47:28,285][__main__][INFO] - agents played in iteration 488 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:47:29,148][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:47:29,193][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:47:29,222][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:47:29,248][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:47:29,249][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:47:29,250][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:47:30,069][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:47:30,579][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:47:31,083][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:47:31,585][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:47:32,086][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:47:32,596][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:47:33,101][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:47:33,604][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:47:34,111][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:47:34,610][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:47:35,111][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:47:35,615][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:47:36,116][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:47:36,617][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:47:37,114][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:47:37,615][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:47:38,116][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:47:38,616][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:47:39,114][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:47:39,613][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:47:40,111][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:47:40,611][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:47:41,109][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:47:41,608][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:47:42,106][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:47:42,607][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:47:43,111][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:47:43,615][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:47:44,120][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:47:44,619][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:47:45,121][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:47:45,622][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:47:46,125][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:47:46,624][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:47:47,120][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:47:47,618][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:47:48,114][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:47:48,612][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:47:49,109][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:47:49,607][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:47:50,104][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:47:50,601][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:47:51,096][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:47:51,592][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:47:52,088][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:47:52,583][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:47:53,077][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:47:53,573][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:47:54,068][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:47:54,563][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:47:55,056][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:47:55,552][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:47:56,048][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:47:56,541][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:47:57,059][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:47:57,552][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:47:58,047][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:47:58,543][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:47:59,039][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:47:59,534][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:48:00,057][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:48:00,552][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:48:01,047][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:48:01,543][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:48:02,038][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10110 tokens. -[2025-11-10 01:48:02,945][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 55.60%, Block Peak % of device VRAM: 62.42%, ΔTime: 00:00:32 -[2025-11-10 01:48:03,588][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:48:03,589][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:48:03,591][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:48:04,618][__main__][INFO] - Iteration 489 took 1m 1s (40.84% Gen, 57.48% Train). Generation: 25s, Training: 35s. Estimated remaining time: 43h 21m 47s. Estimated total time: 51h 10m 55s. Time estimates for 10 more iterations: 10m 14s, 100 more iterations: 1h 42m 21s, 500 more iterations: 8h 31m 49s. -[2025-11-10 01:48:04,620][__main__][INFO] - Starting iteration 489. -[2025-11-10 01:48:05,076][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 48 and human policies 1. -[2025-11-10 01:48:05,078][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:48:32,939][__main__][INFO] - Number of regex retries in iteration 489: 0 -[2025-11-10 01:48:32,940][__main__][INFO] - agents played in iteration 489 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:48:33,732][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:48:33,755][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:48:33,779][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:48:33,802][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:48:33,802][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:48:33,803][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:48:34,640][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:48:35,102][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:48:35,609][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:48:36,111][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:48:36,614][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:48:37,119][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:48:37,620][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:48:38,121][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:48:38,622][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:48:39,128][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:48:39,637][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:48:40,133][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:48:40,628][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:48:41,136][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:48:41,632][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:48:42,131][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:48:42,627][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:48:43,126][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:48:43,624][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:48:44,123][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:48:44,621][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:48:45,127][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:48:45,625][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:48:46,126][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:48:46,623][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:48:47,124][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:48:47,628][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:48:48,131][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:48:48,631][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:48:49,130][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:48:49,629][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:48:50,138][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:48:50,640][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:48:51,141][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:48:51,650][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:48:52,147][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:48:52,643][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:48:53,138][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:48:53,634][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:48:54,132][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:48:54,631][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:48:55,131][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:48:55,630][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:48:56,129][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:48:56,627][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:48:57,122][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:48:57,618][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:48:58,115][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:48:58,614][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:48:59,111][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:48:59,607][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:49:00,104][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:49:00,604][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:49:01,099][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:49:01,596][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:49:02,092][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:49:02,589][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:49:03,089][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:49:03,583][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:49:04,077][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:49:04,573][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:49:05,070][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:49:05,563][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:49:06,059][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:49:06,553][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10091 tokens. -[2025-11-10 01:49:07,310][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.04%, Current % of VRAM taken: 55.53%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:32 -[2025-11-10 01:49:08,046][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:49:08,047][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:49:08,049][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:49:08,955][__main__][INFO] - Iteration 490 took 1m 3s (43.62% Gen, 54.96% Train). Generation: 27s, Training: 35s. Estimated remaining time: 45h 23m 49s. Estimated total time: 53h 14m 1s. Time estimates for 10 more iterations: 10m 38s, 100 more iterations: 1h 46m 28s, 500 more iterations: 8h 52m 20s. -[2025-11-10 01:49:08,957][__main__][INFO] - Starting iteration 490. -[2025-11-10 01:49:09,428][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 48 and human policies 1. -[2025-11-10 01:49:09,428][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:49:30,016][__main__][INFO] - Number of regex retries in iteration 490: 0 -[2025-11-10 01:49:30,017][__main__][INFO] - agents played in iteration 490 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:49:30,820][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:49:30,843][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:49:30,868][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:49:30,891][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:49:30,892][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:49:30,892][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:49:31,704][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:49:32,165][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:49:32,672][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:49:33,182][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:49:33,682][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:49:34,183][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:49:34,686][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:49:35,194][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:49:35,696][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:49:36,203][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:49:36,705][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:49:37,210][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:49:37,714][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:49:38,214][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:49:38,714][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:49:39,213][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:49:39,711][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:49:40,208][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:49:40,709][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:49:41,213][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:49:41,711][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:49:42,210][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:49:42,710][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:49:43,208][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:49:43,706][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:49:44,204][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:49:44,701][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:49:45,201][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:49:45,704][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:49:46,201][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:49:46,698][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:49:47,195][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:49:47,698][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:49:48,199][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:49:48,726][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:49:49,230][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:49:49,728][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:49:50,232][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:49:50,730][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:49:51,230][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:49:51,729][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:49:52,225][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:49:52,721][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:49:53,216][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:49:53,711][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:49:54,209][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:49:54,706][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:49:55,207][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:49:55,703][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:49:56,198][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:49:56,696][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:49:57,192][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:49:57,688][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:49:58,186][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:49:58,680][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:49:59,179][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:49:59,674][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:50:00,170][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:50:00,686][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:50:01,181][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:50:01,681][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:50:02,176][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:50:02,674][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:50:03,172][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:50:03,669][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10119 tokens. -[2025-11-10 01:50:04,424][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 55.63%, Block Peak % of device VRAM: 62.17%, ΔTime: 00:00:32 -[2025-11-10 01:50:05,160][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:50:05,161][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:50:05,163][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:50:06,955][__main__][INFO] - Iteration 491 took 57s (35.79% Gen, 61.09% Train). Generation: 20s, Training: 35s. Estimated remaining time: 40h 5m 14s. Estimated total time: 47h 56m 24s. Time estimates for 10 more iterations: 9m 35s, 100 more iterations: 1h 35m 52s, 500 more iterations: 7h 59m 24s. -[2025-11-10 01:50:06,958][__main__][INFO] - Starting iteration 491. -[2025-11-10 01:50:07,437][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 49 and human policies 1. -[2025-11-10 01:50:07,437][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:50:24,254][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 01:50:36,188][__main__][INFO] - Number of regex retries in iteration 491: 1 -[2025-11-10 01:50:36,189][__main__][INFO] - agents played in iteration 491 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:50:36,997][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:50:37,026][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:50:37,054][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:50:37,078][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:50:37,078][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:50:37,079][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:50:37,908][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:50:38,372][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:50:38,882][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:50:39,389][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:50:39,892][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:50:40,394][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:50:40,901][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:50:41,413][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:50:41,917][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:50:42,417][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:50:42,918][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:50:43,416][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:50:43,914][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:50:44,409][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:50:44,904][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:50:45,400][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:50:45,898][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:50:46,400][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:50:46,899][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:50:47,397][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:50:47,897][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:50:48,397][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:50:48,897][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:50:49,398][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:50:49,897][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:50:50,396][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:50:50,895][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:50:51,397][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:50:51,895][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:50:52,394][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:50:52,899][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:50:53,397][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:50:53,897][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:50:54,405][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:50:54,907][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:50:55,410][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:50:55,912][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:50:56,414][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:50:56,913][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:50:57,414][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:50:57,912][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:50:58,416][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:50:58,913][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:50:59,407][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:50:59,903][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:51:00,397][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:51:00,902][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:51:01,399][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:51:01,893][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:51:02,404][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:51:02,903][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:51:03,397][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:51:03,894][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:51:04,390][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:51:04,886][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:51:05,381][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:51:05,879][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:51:06,376][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:51:06,874][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:51:07,369][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:51:07,868][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:51:08,362][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:51:08,861][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:51:09,360][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:51:09,859][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10185 tokens. -[2025-11-10 01:51:10,663][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 55.62%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:32 -[2025-11-10 01:51:11,311][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:51:11,312][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:51:11,314][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:51:12,076][__main__][INFO] - Iteration 492 took 1m 4s (44.48% Gen, 54.34% Train). Generation: 28s, Training: 35s. Estimated remaining time: 45h 59m 45s. Estimated total time: 53h 52m 0s. Time estimates for 10 more iterations: 10m 46s, 100 more iterations: 1h 47m 44s, 500 more iterations: 8h 58m 40s. -[2025-11-10 01:51:12,079][__main__][INFO] - Starting iteration 492. -[2025-11-10 01:51:12,550][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 49 and human policies 1. -[2025-11-10 01:51:12,551][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:51:40,987][__main__][INFO] - Number of regex retries in iteration 492: 0 -[2025-11-10 01:51:40,988][__main__][INFO] - agents played in iteration 492 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:51:41,803][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:51:41,831][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:51:41,854][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:51:41,876][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:51:41,877][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:51:41,878][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:51:42,686][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:51:43,154][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:51:43,662][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:51:44,166][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:51:44,680][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:51:45,183][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:51:45,694][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:51:46,197][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:51:46,699][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:51:47,197][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:51:47,695][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:51:48,192][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:51:48,690][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:51:49,190][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:51:49,704][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:51:50,200][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:51:50,702][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:51:51,197][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:51:51,693][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:51:52,203][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:51:52,697][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:51:53,193][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:51:53,697][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:51:54,196][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:51:54,702][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:51:55,198][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:51:55,696][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:51:56,196][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:51:56,693][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:51:57,195][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:51:57,704][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:51:58,199][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:51:58,715][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:51:59,213][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:51:59,712][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:52:00,215][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:52:00,720][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:52:01,226][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:52:01,721][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:52:02,217][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:52:02,728][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:52:03,224][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:52:03,724][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:52:04,223][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:52:04,721][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:52:05,232][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:52:05,727][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:52:06,222][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:52:06,720][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:52:07,216][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:52:07,712][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:52:08,207][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:52:08,702][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:52:09,197][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:52:09,693][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:52:10,191][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:52:10,687][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:52:11,186][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:52:11,698][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:52:12,200][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:52:12,702][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:52:13,204][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:52:13,708][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:52:14,221][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:52:14,722][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10239 tokens. -[2025-11-10 01:52:15,557][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 55.70%, Block Peak % of device VRAM: 62.37%, ΔTime: 00:00:32 -[2025-11-10 01:52:16,299][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:52:16,301][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:52:16,303][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:52:17,221][__main__][INFO] - Iteration 493 took 1m 4s (43.97% Gen, 54.61% Train). Generation: 28s, Training: 35s. Estimated remaining time: 46h 0m 13s. Estimated total time: 53h 53m 34s. Time estimates for 10 more iterations: 10m 46s, 100 more iterations: 1h 47m 47s, 500 more iterations: 8h 58m 55s. -[2025-11-10 01:52:17,224][__main__][INFO] - Starting iteration 493. -[2025-11-10 01:52:17,698][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 49 and human policies 1. -[2025-11-10 01:52:17,699][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:52:43,712][__main__][INFO] - Number of regex retries in iteration 493: 0 -[2025-11-10 01:52:43,714][__main__][INFO] - agents played in iteration 493 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:52:44,519][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:52:44,545][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:52:44,571][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:52:44,595][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:52:44,596][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:52:44,597][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:52:45,408][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:52:45,867][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:52:46,375][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:52:46,878][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:52:47,380][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:52:47,881][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:52:48,390][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:52:48,890][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:52:49,394][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:52:49,895][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:52:50,395][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:52:50,895][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:52:51,396][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:52:51,898][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:52:52,404][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:52:52,914][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:52:53,417][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:52:53,920][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:52:54,419][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:52:54,928][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:52:55,428][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:52:55,928][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:52:56,442][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:52:56,936][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:52:57,439][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:52:57,935][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:52:58,435][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:52:58,939][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:52:59,468][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:52:59,970][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:53:00,477][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:53:00,976][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:53:01,478][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:53:01,976][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:53:02,474][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:53:02,975][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:53:03,473][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:53:03,973][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:53:04,475][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:53:04,976][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:53:05,483][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:53:05,986][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:53:06,485][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:53:06,991][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:53:07,491][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:53:07,993][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:53:08,494][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:53:08,995][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:53:09,505][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:53:10,000][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:53:10,498][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:53:11,006][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:53:11,506][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:53:12,011][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:53:12,514][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:53:13,017][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:53:13,536][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:53:14,034][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:53:14,538][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:53:15,039][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:53:15,541][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:53:16,051][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:53:16,555][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:53:17,057][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:53:17,559][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10171 tokens. -[2025-11-10 01:53:18,385][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 55.63%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 -[2025-11-10 01:53:19,030][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:53:19,032][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:53:19,034][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:53:19,862][__main__][INFO] - Iteration 494 took 1m 2s (41.85% Gen, 56.82% Train). Generation: 26s, Training: 35s. Estimated remaining time: 43h 53m 49s. Estimated total time: 51h 48m 12s. Time estimates for 10 more iterations: 10m 21s, 100 more iterations: 1h 43m 36s, 500 more iterations: 8h 38m 2s. -[2025-11-10 01:53:19,864][__main__][INFO] - Starting iteration 494. -[2025-11-10 01:53:20,328][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 49 and human policies 1. -[2025-11-10 01:53:20,329][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:53:50,605][__main__][INFO] - Number of regex retries in iteration 494: 0 -[2025-11-10 01:53:50,605][__main__][INFO] - agents played in iteration 494 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:53:51,399][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:53:51,423][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:53:51,447][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:53:51,470][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:53:51,470][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:53:51,472][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:53:52,272][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:53:52,730][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:53:53,234][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:53:53,737][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:53:54,235][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:53:54,734][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:53:55,233][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:53:55,731][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:53:56,232][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:53:56,743][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:53:57,241][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:53:57,761][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:53:58,260][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:53:58,757][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:53:59,259][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:53:59,757][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:54:00,257][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:54:00,762][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:54:01,265][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:54:01,764][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:54:02,264][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:54:02,764][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:54:03,268][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:54:03,771][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:54:04,272][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:54:04,766][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:54:05,265][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:54:05,766][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:54:06,265][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:54:06,763][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:54:07,259][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:54:07,756][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:54:08,252][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:54:08,750][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:54:09,247][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:54:09,743][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:54:10,242][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:54:10,739][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:54:11,243][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:54:11,740][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:54:12,237][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:54:12,733][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:54:13,228][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:54:13,745][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:54:14,240][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:54:14,737][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:54:15,232][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:54:15,727][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:54:16,235][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:54:16,736][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:54:17,233][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:54:17,733][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:54:18,233][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:54:18,733][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:54:19,232][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:54:19,731][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:54:20,229][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:54:20,723][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:54:21,226][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:54:21,726][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:54:22,226][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:54:22,727][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:54:23,230][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:54:23,730][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:54:24,248][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10288 tokens. -[2025-11-10 01:54:25,080][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 55.72%, Block Peak % of device VRAM: 62.44%, ΔTime: 00:00:32 -[2025-11-10 01:54:25,811][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:54:25,813][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:54:25,815][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:54:26,728][__main__][INFO] - Iteration 495 took 1m 6s (45.60% Gen, 53.03% Train). Generation: 30s, Training: 35s. Estimated remaining time: 47h 24m 31s. Estimated total time: 55h 20m 1s. Time estimates for 10 more iterations: 11m 4s, 100 more iterations: 1h 50m 40s, 500 more iterations: 9h 13m 20s. -[2025-11-10 01:54:26,730][__main__][INFO] - Starting iteration 495. -[2025-11-10 01:54:27,222][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 49 and human policies 1. -[2025-11-10 01:54:27,223][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:54:48,593][__main__][INFO] - Number of regex retries in iteration 495: 0 -[2025-11-10 01:54:48,593][__main__][INFO] - agents played in iteration 495 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:54:49,397][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:54:49,423][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:54:49,448][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:54:49,471][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:54:49,472][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:54:49,473][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:54:50,274][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:54:50,734][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:54:51,237][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:54:51,748][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:54:52,250][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:54:52,763][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:54:53,261][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:54:53,759][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:54:54,257][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:54:54,754][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:54:55,251][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:54:55,748][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:54:56,246][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:54:56,759][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:54:57,262][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:54:57,761][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:54:58,260][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:54:58,760][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:54:59,265][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:54:59,766][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:55:00,264][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:55:00,768][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:55:01,266][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:55:01,768][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:55:02,265][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:55:02,763][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:55:03,269][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:55:03,768][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:55:04,269][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:55:04,768][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:55:05,267][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:55:05,767][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:55:06,269][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:55:06,769][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:55:07,277][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:55:07,776][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:55:08,271][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:55:08,768][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:55:09,269][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:55:09,771][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:55:10,267][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:55:10,762][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:55:11,258][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:55:11,753][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:55:12,258][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:55:12,756][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:55:13,252][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:55:13,764][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:55:14,261][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:55:14,758][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:55:15,254][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:55:15,749][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:55:16,262][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:55:16,759][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:55:17,259][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:55:17,762][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:55:18,263][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:55:18,766][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:55:19,682][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:55:20,174][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:55:20,673][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:55:21,173][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:55:21,673][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:55:22,174][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:55:22,678][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10191 tokens. -[2025-11-10 01:55:23,552][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 55.71%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:33 -[2025-11-10 01:55:24,217][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:55:24,221][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:55:24,223][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:55:25,068][__main__][INFO] - Iteration 496 took 57s (36.94% Gen, 61.59% Train). Generation: 21s, Training: 35s. Estimated remaining time: 40h 15m 52s. Estimated total time: 48h 12m 20s. Time estimates for 10 more iterations: 9m 38s, 100 more iterations: 1h 36m 24s, 500 more iterations: 8h 2m 3s. -[2025-11-10 01:55:25,070][__main__][INFO] - Starting iteration 496. -[2025-11-10 01:55:25,552][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 49 and human policies 1. -[2025-11-10 01:55:25,553][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:55:59,564][__main__][INFO] - Number of regex retries in iteration 496: 0 -[2025-11-10 01:55:59,565][__main__][INFO] - agents played in iteration 496 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:56:00,337][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:56:00,365][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:56:00,391][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:56:00,415][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:56:00,416][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:56:00,417][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:56:01,159][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:56:01,616][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:56:02,117][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:56:02,621][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:56:03,118][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:56:03,619][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:56:04,117][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:56:04,615][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:56:05,115][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:56:05,617][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:56:06,116][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:56:06,617][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:56:07,117][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:56:07,615][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:56:08,116][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:56:08,614][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:56:09,113][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:56:09,618][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:56:10,142][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:56:10,648][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:56:11,149][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:56:11,649][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:56:12,159][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:56:12,663][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:56:13,184][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:56:13,685][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:56:14,181][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:56:14,680][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:56:15,175][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:56:15,679][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:56:16,174][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:56:16,671][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:56:17,167][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:56:17,663][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:56:18,158][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:56:18,655][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:56:19,151][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:56:19,663][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:56:20,159][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:56:20,656][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:56:21,158][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:56:21,654][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:56:22,156][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:56:22,654][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:56:23,150][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:56:23,643][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:56:24,140][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:56:24,635][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:56:25,130][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:56:25,627][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:56:26,124][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:56:26,619][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:56:27,113][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:56:27,610][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:56:28,109][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:56:28,608][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:56:29,107][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:56:29,607][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:56:30,110][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:56:30,612][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:56:31,113][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:56:31,617][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:56:32,123][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:56:32,624][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:56:33,129][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10111 tokens. -[2025-11-10 01:56:33,973][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 55.69%, Block Peak % of device VRAM: 62.38%, ΔTime: 00:00:32 -[2025-11-10 01:56:34,742][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:56:34,744][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:56:34,746][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:56:35,665][__main__][INFO] - Iteration 497 took 1m 10s (48.51% Gen, 50.18% Train). Generation: 34s, Training: 35s. Estimated remaining time: 50h 28m 4s. Estimated total time: 58h 25m 42s. Time estimates for 10 more iterations: 11m 41s, 100 more iterations: 1h 56m 51s, 500 more iterations: 9h 44m 17s. -[2025-11-10 01:56:35,668][__main__][INFO] - Starting iteration 497. -[2025-11-10 01:56:36,142][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 49 and human policies 1. -[2025-11-10 01:56:36,143][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:57:02,257][__main__][INFO] - Number of regex retries in iteration 497: 0 -[2025-11-10 01:57:02,259][__main__][INFO] - agents played in iteration 497 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:57:03,245][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:57:03,301][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:57:03,342][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:57:03,376][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:57:03,377][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:57:03,378][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:57:04,149][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:57:04,604][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:57:05,109][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:57:05,608][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:57:06,112][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:57:06,614][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:57:07,116][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:57:07,615][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:57:08,114][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:57:08,618][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:57:09,123][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:57:09,625][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:57:10,123][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:57:10,627][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:57:11,126][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:57:11,626][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:57:12,127][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:57:12,623][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:57:13,119][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:57:13,618][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:57:14,116][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:57:14,616][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:57:15,119][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:57:15,618][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:57:16,123][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:57:16,621][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:57:17,116][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:57:17,615][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:57:18,113][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:57:18,613][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:57:19,107][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:57:19,606][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:57:20,109][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:57:20,606][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:57:21,103][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:57:21,601][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:57:22,100][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:57:22,601][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:57:23,099][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:57:23,598][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:57:24,095][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:57:24,592][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:57:25,092][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:57:25,587][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:57:26,082][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:57:26,579][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:57:27,077][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:57:27,574][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:57:28,072][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:57:28,571][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:57:29,069][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:57:29,566][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:57:30,060][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:57:30,566][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:57:31,064][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:57:31,562][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:57:32,075][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:57:32,574][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:57:33,073][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:57:33,567][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:57:34,069][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:57:34,584][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:57:35,083][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:57:35,585][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:57:36,085][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10176 tokens. -[2025-11-10 01:57:36,924][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.02%, Current % of VRAM taken: 55.52%, Block Peak % of device VRAM: 62.21%, ΔTime: 00:00:32 -[2025-11-10 01:57:37,592][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:57:37,594][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:57:37,596][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:57:38,574][__main__][INFO] - Iteration 498 took 1m 2s (41.83% Gen, 56.60% Train). Generation: 26s, Training: 35s. Estimated remaining time: 44h 2m 56s. Estimated total time: 52h 1m 37s. Time estimates for 10 more iterations: 10m 24s, 100 more iterations: 1h 44m 3s, 500 more iterations: 8h 40m 16s. -[2025-11-10 01:57:38,576][__main__][INFO] - Starting iteration 498. -[2025-11-10 01:57:39,065][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 49 and human policies 1. -[2025-11-10 01:57:39,066][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:58:09,758][__main__][INFO] - Number of regex retries in iteration 498: 0 -[2025-11-10 01:58:09,759][__main__][INFO] - agents played in iteration 498 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:58:10,534][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:58:10,561][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:58:10,588][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:58:10,612][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:58:10,612][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:58:10,614][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:58:11,340][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:58:11,796][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:58:12,296][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:58:12,799][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:58:13,295][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:58:13,793][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:58:14,304][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:58:14,799][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:58:15,296][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:58:15,794][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:58:16,292][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:58:16,795][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:58:17,298][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:58:17,795][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:58:18,296][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:58:18,792][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:58:19,288][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:58:19,786][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:58:20,291][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:58:20,802][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:58:21,303][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:58:21,799][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:58:22,297][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:58:22,797][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:58:23,300][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:58:23,809][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:58:24,308][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:58:24,810][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:58:25,309][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:58:25,809][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:58:26,307][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:58:26,811][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:58:27,313][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:58:27,807][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:58:28,303][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:58:28,796][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:58:29,295][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:58:29,796][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:58:30,294][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:58:30,792][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:58:31,304][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:58:31,801][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:58:32,300][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:58:32,798][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:58:33,295][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:58:33,793][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:58:34,293][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:58:34,791][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:58:35,304][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:58:35,799][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:58:36,296][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:58:36,795][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:58:37,291][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:58:37,808][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:58:38,305][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:58:38,801][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:58:39,298][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:58:39,798][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:58:40,298][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:58:40,800][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:58:41,304][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:58:41,815][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:58:42,314][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:58:42,817][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:58:43,322][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10245 tokens. -[2025-11-10 01:58:44,158][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 55.59%, Block Peak % of device VRAM: 62.54%, ΔTime: 00:00:32 -[2025-11-10 01:58:44,926][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:58:44,928][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:58:44,930][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:58:45,836][__main__][INFO] - Iteration 499 took 1m 6s (45.97% Gen, 52.67% Train). Generation: 30s, Training: 35s. Estimated remaining time: 47h 38m 44s. Estimated total time: 55h 38m 33s. Time estimates for 10 more iterations: 11m 7s, 100 more iterations: 1h 51m 17s, 500 more iterations: 9h 16m 25s. -[2025-11-10 01:58:45,838][__main__][INFO] - Starting iteration 499. -[2025-11-10 01:58:46,306][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 49 and human policies 1. -[2025-11-10 01:58:46,307][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 01:59:13,051][__main__][INFO] - Number of regex retries in iteration 499: 0 -[2025-11-10 01:59:13,053][__main__][INFO] - agents played in iteration 499 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 01:59:13,899][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:59:13,922][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:59:13,945][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:59:13,967][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 01:59:13,968][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 01:59:13,968][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 01:59:14,744][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 01:59:15,203][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 01:59:15,710][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 01:59:16,211][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 01:59:16,715][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 01:59:17,215][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 01:59:17,717][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 01:59:18,221][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 01:59:18,723][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 01:59:19,223][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 01:59:19,722][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 01:59:20,218][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 01:59:20,714][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 01:59:21,213][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 01:59:21,711][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 01:59:22,213][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 01:59:22,713][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 01:59:23,214][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 01:59:23,713][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 01:59:24,213][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 01:59:24,716][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 01:59:25,219][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 01:59:25,720][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 01:59:26,219][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 01:59:26,717][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 01:59:27,218][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 01:59:27,721][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 01:59:28,215][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 01:59:28,715][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 01:59:29,215][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 01:59:29,710][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 01:59:30,205][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 01:59:30,705][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 01:59:31,200][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 01:59:31,715][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 01:59:32,211][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 01:59:32,710][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 01:59:33,210][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 01:59:33,704][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 01:59:34,208][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 01:59:34,703][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 01:59:35,197][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 01:59:35,712][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 01:59:36,206][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 01:59:36,700][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 01:59:37,201][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 01:59:37,696][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 01:59:38,198][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 01:59:38,695][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 01:59:39,191][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 01:59:39,693][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 01:59:40,194][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 01:59:40,699][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 01:59:41,198][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 01:59:41,696][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 01:59:42,197][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 01:59:42,700][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 01:59:43,206][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 01:59:43,712][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 01:59:44,219][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 01:59:44,738][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 01:59:45,240][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 01:59:45,745][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 01:59:46,251][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 01:59:46,755][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10273 tokens. -[2025-11-10 01:59:47,603][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 55.71%, Block Peak % of device VRAM: 62.53%, ΔTime: 00:00:32 -[2025-11-10 01:59:48,243][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 01:59:48,245][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 01:59:48,247][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 01:59:49,069][__main__][INFO] - Iteration 500 took 1m 2s (42.61% Gen, 56.07% Train). Generation: 26s, Training: 35s. Estimated remaining time: 44h 17m 19s. Estimated total time: 52h 18m 11s. Time estimates for 10 more iterations: 10m 27s, 100 more iterations: 1h 44m 36s, 500 more iterations: 8h 43m 1s. -[2025-11-10 01:59:49,071][__main__][INFO] - Starting iteration 500. -[2025-11-10 01:59:49,546][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 49 and human policies 1. -[2025-11-10 01:59:49,547][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:00:19,709][__main__][INFO] - Number of regex retries in iteration 500: 0 -[2025-11-10 02:00:19,710][__main__][INFO] - agents played in iteration 500 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:00:20,478][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:00:20,506][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:00:20,533][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:00:20,557][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:00:20,557][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:00:20,558][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:00:21,291][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:00:21,752][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:00:22,254][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:00:22,752][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:00:23,253][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:00:23,752][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:00:24,252][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:00:24,756][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:00:25,253][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:00:25,752][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:00:26,252][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:00:26,753][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:00:27,253][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:00:27,750][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:00:28,251][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:00:28,750][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:00:29,247][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:00:29,749][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:00:30,249][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:00:30,750][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:00:31,246][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:00:31,746][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:00:32,254][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:00:32,755][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:00:33,262][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:00:33,766][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:00:34,260][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:00:34,757][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:00:35,253][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:00:35,748][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:00:36,252][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:00:36,751][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:00:37,247][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:00:37,741][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:00:38,237][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:00:38,732][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:00:39,229][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:00:39,725][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:00:40,236][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:00:40,731][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:00:41,226][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:00:41,725][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:00:42,220][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:00:42,716][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:00:43,213][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:00:43,708][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:00:44,213][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:00:44,713][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:00:45,211][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:00:45,737][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:00:46,234][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:00:46,731][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:00:47,234][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:00:47,734][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:00:48,253][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:00:48,753][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:00:49,258][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:00:49,752][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:00:50,249][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:00:50,755][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:00:51,253][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:00:51,751][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:00:52,253][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:00:52,751][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:00:53,254][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9981 tokens. -[2025-11-10 02:00:54,077][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 13.99%, Current % of VRAM taken: 55.49%, Block Peak % of device VRAM: 62.06%, ΔTime: 00:00:32 -[2025-11-10 02:00:54,849][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:00:54,851][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:00:54,852][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:00:56,638][__main__][INFO] - Iteration 501 took 1m 7s (44.96% Gen, 52.38% Train). Generation: 30s, Training: 35s. Estimated remaining time: 47h 52m 40s. Estimated total time: 55h 54m 39s. Time estimates for 10 more iterations: 11m 10s, 100 more iterations: 1h 51m 49s, 500 more iterations: 9h 19m 6s. -[2025-11-10 02:00:56,641][__main__][INFO] - Starting iteration 501. -[2025-11-10 02:00:57,129][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 50 and human policies 1. -[2025-11-10 02:00:57,130][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:01:22,130][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 02:01:25,410][__main__][INFO] - Number of regex retries in iteration 501: 1 -[2025-11-10 02:01:25,411][__main__][INFO] - agents played in iteration 501 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:01:26,193][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:01:26,221][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:01:26,248][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:01:26,272][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:01:26,272][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:01:26,273][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:01:27,011][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:01:27,463][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:01:27,968][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:01:28,463][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:01:28,967][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:01:29,468][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:01:29,967][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:01:30,467][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:01:30,966][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:01:31,474][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:01:31,979][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:01:32,480][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:01:32,980][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:01:33,481][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:01:34,010][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:01:34,520][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:01:35,016][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:01:35,527][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:01:36,023][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:01:36,519][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:01:37,026][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:01:37,522][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:01:38,026][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:01:38,521][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:01:39,017][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:01:39,528][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:01:40,027][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:01:40,523][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:01:41,031][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:01:41,529][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:01:42,038][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:01:42,536][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:01:43,034][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:01:43,536][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:01:44,036][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:01:44,532][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:01:45,032][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:01:45,530][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:01:46,044][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:01:46,547][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:01:47,052][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:01:47,550][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:01:48,049][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:01:48,558][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:01:49,054][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:01:49,555][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:01:50,058][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:01:51,612][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:01:52,118][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:01:52,623][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:01:53,129][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:01:53,628][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:01:54,133][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:01:54,636][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:01:55,159][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:01:55,664][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:01:56,169][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:01:56,667][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:01:57,168][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:01:57,671][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:01:58,170][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:01:58,675][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:01:59,185][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:01:59,686][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:02:00,185][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10197 tokens. -[2025-11-10 02:02:01,038][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 55.59%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:34 -[2025-11-10 02:02:01,699][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:02:01,700][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:02:01,702][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:02:02,530][__main__][INFO] - Iteration 502 took 1m 5s (43.24% Gen, 55.49% Train). Generation: 28s, Training: 36s. Estimated remaining time: 46h 26m 58s. Estimated total time: 54h 30m 4s. Time estimates for 10 more iterations: 10m 54s, 100 more iterations: 1h 49m 0s, 500 more iterations: 9h 5m 0s. -[2025-11-10 02:02:02,534][__main__][INFO] - Starting iteration 502. -[2025-11-10 02:02:03,068][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 50 and human policies 1. -[2025-11-10 02:02:03,069][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:02:25,716][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 02:02:38,178][__main__][INFO] - Number of regex retries in iteration 502: 1 -[2025-11-10 02:02:38,178][__main__][INFO] - agents played in iteration 502 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:02:38,963][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:02:38,990][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:02:39,017][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:02:39,040][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:02:39,041][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:02:39,042][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:02:39,770][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:02:40,223][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:02:40,729][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:02:41,225][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:02:41,720][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:02:42,222][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:02:42,718][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:02:43,213][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:02:43,713][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:02:44,209][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:02:44,709][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:02:45,206][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:02:45,701][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:02:46,198][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:02:46,695][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:02:47,195][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:02:47,691][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:02:48,186][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:02:48,684][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:02:49,185][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:02:49,681][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:02:50,173][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:02:50,668][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:02:51,179][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:02:51,673][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:02:52,166][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:02:52,661][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:02:53,158][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:02:53,667][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:02:54,161][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:02:54,659][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:02:55,166][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:02:55,663][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:02:56,166][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:02:56,667][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:02:57,187][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:02:57,703][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:02:58,210][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:02:58,709][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:02:59,216][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:02:59,717][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:03:00,219][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:03:00,724][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:03:01,224][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:03:01,728][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:03:02,229][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:03:02,732][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:03:03,236][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:03:03,737][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:03:04,239][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:03:04,739][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:03:05,243][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:03:05,754][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:03:06,255][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:03:06,759][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:03:07,261][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:03:07,762][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:03:08,267][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:03:08,772][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:03:09,273][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:03:09,775][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:03:10,279][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:03:10,787][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:03:11,287][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:03:11,792][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10114 tokens. -[2025-11-10 02:03:12,635][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 55.66%, Block Peak % of device VRAM: 62.37%, ΔTime: 00:00:32 -[2025-11-10 02:03:13,390][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:03:13,394][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:03:13,395][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:03:14,277][__main__][INFO] - Iteration 503 took 1m 11s (49.30% Gen, 49.46% Train). Generation: 35s, Training: 35s. Estimated remaining time: 51h 16m 11s. Estimated total time: 59h 20m 29s. Time estimates for 10 more iterations: 11m 52s, 100 more iterations: 1h 58m 40s, 500 more iterations: 9h 53m 24s. -[2025-11-10 02:03:14,279][__main__][INFO] - Starting iteration 503. -[2025-11-10 02:03:14,748][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 50 and human policies 1. -[2025-11-10 02:03:14,750][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:03:34,312][__main__][INFO] - Number of regex retries in iteration 503: 0 -[2025-11-10 02:03:34,313][__main__][INFO] - agents played in iteration 503 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:03:35,103][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:03:35,132][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:03:35,160][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:03:35,183][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:03:35,183][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:03:35,184][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:03:35,915][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:03:36,372][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:03:36,873][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:03:37,373][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:03:37,876][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:03:38,391][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:03:38,887][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:03:39,382][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:03:39,901][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:03:40,394][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:03:40,893][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:03:41,391][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:03:41,885][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:03:42,396][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:03:42,889][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:03:43,382][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:03:43,880][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:03:44,383][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:03:44,897][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:03:45,397][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:03:45,895][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:03:46,395][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:03:46,893][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:03:47,394][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:03:47,894][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:03:48,394][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:03:48,904][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:03:49,403][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:03:49,900][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:03:50,399][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:03:50,896][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:03:51,402][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:03:51,897][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:03:52,394][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:03:52,890][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:03:53,388][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:03:53,883][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:03:54,386][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:03:54,887][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:03:55,389][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:03:55,889][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:03:56,390][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:03:56,897][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:03:57,401][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:03:57,905][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:03:58,406][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:03:58,907][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:03:59,414][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:03:59,916][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:04:00,417][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:04:00,923][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:04:01,427][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:04:01,930][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:04:02,429][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:04:02,928][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:04:03,429][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:04:03,928][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:04:04,433][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:04:04,935][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:04:05,434][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:04:05,938][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:04:06,438][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:04:06,940][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:04:07,439][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:04:07,934][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10124 tokens. -[2025-11-10 02:04:08,706][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 55.63%, Block Peak % of device VRAM: 62.45%, ΔTime: 00:00:32 -[2025-11-10 02:04:09,432][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:04:09,434][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:04:09,435][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:04:10,364][__main__][INFO] - Iteration 504 took 55s (35.17% Gen, 63.15% Train). Generation: 19s, Training: 35s. Estimated remaining time: 38h 15m 36s. Estimated total time: 46h 20m 49s. Time estimates for 10 more iterations: 9m 16s, 100 more iterations: 1h 32m 41s, 500 more iterations: 7h 43m 28s. -[2025-11-10 02:04:10,366][__main__][INFO] - Starting iteration 504. -[2025-11-10 02:04:10,853][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 50 and human policies 1. -[2025-11-10 02:04:10,853][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:04:39,293][__main__][INFO] - Number of regex retries in iteration 504: 0 -[2025-11-10 02:04:39,295][__main__][INFO] - agents played in iteration 504 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:04:40,075][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:04:40,098][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:04:40,121][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:04:40,144][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:04:40,144][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:04:40,145][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:04:40,911][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:04:41,374][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:04:41,874][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:04:42,380][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:04:42,892][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:04:43,394][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:04:43,916][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:04:44,413][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:04:44,910][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:04:45,413][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:04:45,908][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:04:46,403][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:04:46,898][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:04:47,398][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:04:47,898][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:04:48,394][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:04:48,893][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:04:49,399][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:04:49,893][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:04:50,391][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:04:50,890][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:04:51,383][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:04:51,887][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:04:52,381][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:04:52,881][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:04:53,385][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:04:53,891][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:04:54,393][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:04:54,896][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:04:55,398][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:04:55,901][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:04:56,404][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:04:56,910][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:04:57,413][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:04:57,914][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:04:58,438][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:04:58,941][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:04:59,443][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:04:59,944][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:05:00,447][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:05:00,949][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:05:01,454][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:05:01,956][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:05:02,461][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:05:02,964][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:05:03,468][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:05:03,972][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:05:04,479][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:05:04,984][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:05:05,488][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:05:05,992][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:05:06,496][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:05:06,998][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:05:07,519][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:05:08,020][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:05:08,522][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:05:09,025][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:05:09,530][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:05:10,035][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:05:10,538][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:05:11,041][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:05:11,547][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:05:12,052][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:05:12,555][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:05:13,058][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10200 tokens. -[2025-11-10 02:05:13,916][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 55.56%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:33 -[2025-11-10 02:05:14,550][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:05:14,552][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:05:14,554][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:05:15,380][__main__][INFO] - Iteration 505 took 1m 4s (44.08% Gen, 54.64% Train). Generation: 28s, Training: 35s. Estimated remaining time: 45h 40m 6s. Estimated total time: 53h 46m 24s. Time estimates for 10 more iterations: 10m 45s, 100 more iterations: 1h 47m 32s, 500 more iterations: 8h 57m 44s. -[2025-11-10 02:05:15,382][__main__][INFO] - Starting iteration 505. -[2025-11-10 02:05:15,849][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 50 and human policies 1. -[2025-11-10 02:05:15,850][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:05:43,340][__main__][INFO] - Number of regex retries in iteration 505: 0 -[2025-11-10 02:05:43,341][__main__][INFO] - agents played in iteration 505 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:05:44,124][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:05:44,147][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:05:44,171][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:05:44,194][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:05:44,194][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:05:44,195][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:05:44,940][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:05:45,418][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:05:45,921][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:05:46,419][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:05:46,918][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:05:47,418][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:05:47,930][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:05:48,434][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:05:48,940][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:05:49,437][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:05:49,932][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:05:50,429][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:05:50,924][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:05:51,422][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:05:51,935][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:05:52,431][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:05:52,926][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:05:53,423][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:05:53,920][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:05:54,417][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:05:54,914][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:05:55,417][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:05:55,932][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:05:56,428][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:05:56,924][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:05:57,423][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:05:57,925][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:05:58,431][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:05:58,929][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:05:59,431][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:05:59,931][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:06:00,430][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:06:00,927][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:06:01,426][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:06:01,924][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:06:02,422][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:06:02,921][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:06:03,420][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:06:03,946][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:06:04,451][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:06:04,957][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:06:05,460][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:06:05,960][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:06:06,464][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:06:06,963][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:06:07,462][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:06:07,962][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:06:08,463][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:06:08,970][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:06:09,471][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:06:09,971][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:06:10,472][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:06:10,973][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:06:11,477][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:06:11,981][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:06:12,481][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:06:12,988][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:06:13,492][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:06:13,998][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:06:14,502][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:06:15,006][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:06:15,528][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:06:16,030][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:06:16,535][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:06:17,038][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10186 tokens. -[2025-11-10 02:06:17,883][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 55.68%, Block Peak % of device VRAM: 62.45%, ΔTime: 00:00:32 -[2025-11-10 02:06:18,609][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:06:18,610][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:06:18,612][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:06:19,523][__main__][INFO] - Iteration 506 took 1m 3s (43.17% Gen, 55.39% Train). Generation: 27s, Training: 35s. Estimated remaining time: 44h 56m 19s. Estimated total time: 53h 3m 41s. Time estimates for 10 more iterations: 10m 36s, 100 more iterations: 1h 46m 7s, 500 more iterations: 8h 50m 36s. -[2025-11-10 02:06:19,525][__main__][INFO] - Starting iteration 506. -[2025-11-10 02:06:19,984][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 50 and human policies 1. -[2025-11-10 02:06:19,984][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:06:48,689][__main__][INFO] - Number of regex retries in iteration 506: 0 -[2025-11-10 02:06:48,691][__main__][INFO] - agents played in iteration 506 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:06:49,465][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:06:49,490][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:06:49,515][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:06:49,538][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:06:49,539][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:06:49,539][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:06:50,279][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:06:50,743][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:06:51,246][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:06:51,744][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:06:52,245][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:06:52,742][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:06:53,244][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:06:53,738][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:06:54,236][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:06:54,740][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:06:55,237][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:06:55,737][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:06:56,242][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:06:56,738][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:06:57,241][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:06:57,741][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:06:58,239][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:06:58,744][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:06:59,241][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:06:59,747][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:07:00,250][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:07:00,747][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:07:01,246][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:07:01,739][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:07:02,237][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:07:02,744][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:07:03,238][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:07:03,735][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:07:04,235][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:07:04,730][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:07:05,231][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:07:05,727][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:07:06,232][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:07:06,744][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:07:07,245][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:07:07,745][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:07:08,241][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:07:08,738][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:07:09,247][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:07:09,745][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:07:10,242][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:07:10,744][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:07:11,249][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:07:11,754][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:07:12,256][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:07:12,757][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:07:13,257][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:07:13,757][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:07:14,257][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:07:14,761][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:07:15,261][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:07:15,766][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:07:16,270][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:07:16,771][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:07:17,274][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:07:17,774][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:07:18,278][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:07:18,783][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:07:19,285][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:07:19,788][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:07:20,295][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:07:20,800][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:07:21,310][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:07:21,811][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:07:22,313][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10095 tokens. -[2025-11-10 02:07:23,170][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 55.67%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 -[2025-11-10 02:07:23,827][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:07:23,830][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:07:23,831][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:07:24,640][__main__][INFO] - Iteration 507 took 1m 4s (44.40% Gen, 54.35% Train). Generation: 28s, Training: 35s. Estimated remaining time: 45h 44m 23s. Estimated total time: 53h 52m 51s. Time estimates for 10 more iterations: 10m 46s, 100 more iterations: 1h 47m 45s, 500 more iterations: 8h 58m 48s. -[2025-11-10 02:07:24,642][__main__][INFO] - Starting iteration 507. -[2025-11-10 02:07:25,117][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 50 and human policies 1. -[2025-11-10 02:07:25,118][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:07:48,592][__main__][INFO] - Number of regex retries in iteration 507: 0 -[2025-11-10 02:07:48,593][__main__][INFO] - agents played in iteration 507 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:07:49,373][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:07:49,399][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:07:49,424][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:07:49,447][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:07:49,447][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:07:49,449][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:07:50,207][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:07:50,661][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:07:51,163][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:07:51,662][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:07:52,160][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:07:52,661][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:07:53,157][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:07:53,658][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:07:54,159][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:07:54,658][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:07:55,155][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:07:55,654][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:07:56,153][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:07:56,651][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:07:57,148][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:07:57,645][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:07:58,152][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:07:58,649][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:07:59,150][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:07:59,654][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:08:00,152][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:08:00,655][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:08:01,153][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:08:01,653][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:08:02,159][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:08:02,657][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:08:03,155][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:08:03,649][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:08:04,143][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:08:04,645][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:08:05,142][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:08:05,638][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:08:06,148][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:08:06,642][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:08:07,139][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:08:07,638][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:08:08,135][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:08:08,637][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:08:09,133][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:08:09,631][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:08:10,128][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:08:10,622][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:08:11,127][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:08:11,621][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:08:12,118][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:08:12,618][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:08:13,119][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:08:13,622][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:08:14,125][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:08:14,625][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:08:15,129][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:08:15,632][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:08:16,137][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:08:16,657][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:08:17,160][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:08:17,662][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:08:18,161][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:08:18,661][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:08:19,162][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:08:19,661][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:08:20,165][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:08:20,666][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:08:21,165][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:08:21,671][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:08:22,174][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10119 tokens. -[2025-11-10 02:08:23,040][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.05%, Current % of VRAM taken: 55.54%, Block Peak % of device VRAM: 62.08%, ΔTime: 00:00:32 -[2025-11-10 02:08:23,757][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:08:23,758][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:08:23,760][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:08:24,686][__main__][INFO] - Iteration 508 took 59s (39.41% Gen, 59.04% Train). Generation: 23s, Training: 35s. Estimated remaining time: 41h 29m 1s. Estimated total time: 49h 38m 28s. Time estimates for 10 more iterations: 9m 55s, 100 more iterations: 1h 39m 16s, 500 more iterations: 8h 16m 24s. -[2025-11-10 02:08:24,688][__main__][INFO] - Starting iteration 508. -[2025-11-10 02:08:25,155][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 50 and human policies 1. -[2025-11-10 02:08:25,155][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:08:47,288][__main__][INFO] - Number of regex retries in iteration 508: 0 -[2025-11-10 02:08:47,290][__main__][INFO] - agents played in iteration 508 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:08:48,160][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:08:48,188][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:08:48,215][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:08:48,238][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:08:48,239][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:08:48,239][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:08:49,057][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:08:49,526][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:08:50,033][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:08:50,535][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:08:51,036][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:08:51,541][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:08:52,059][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:08:52,559][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:08:53,062][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:08:53,563][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:08:54,065][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:08:54,566][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:08:55,067][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:08:55,572][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:08:56,077][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:08:56,576][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:08:57,076][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:08:57,573][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:08:58,073][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:08:58,574][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:08:59,072][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:08:59,569][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:09:00,067][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:09:00,564][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:09:01,066][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:09:01,565][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:09:02,061][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:09:02,556][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:09:03,054][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:09:03,550][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:09:04,050][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:09:04,546][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:09:05,055][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:09:05,552][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:09:06,056][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:09:06,569][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:09:07,067][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:09:07,572][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:09:08,076][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:09:08,582][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:09:09,084][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:09:09,580][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:09:10,076][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:09:10,573][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:09:11,068][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:09:11,574][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:09:12,074][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:09:12,568][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:09:13,066][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:09:13,562][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:09:14,068][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:09:14,566][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:09:15,061][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:09:15,558][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:09:16,055][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:09:16,552][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:09:17,048][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:09:17,546][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:09:18,063][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:09:18,563][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:09:19,066][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:09:19,565][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:09:20,064][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:09:20,570][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:09:21,071][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10049 tokens. -[2025-11-10 02:09:21,922][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 55.65%, Block Peak % of device VRAM: 62.15%, ΔTime: 00:00:32 -[2025-11-10 02:09:22,567][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:09:22,569][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:09:22,571][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:09:23,402][__main__][INFO] - Iteration 509 took 58s (38.00% Gen, 60.57% Train). Generation: 22s, Training: 35s. Estimated remaining time: 40h 21m 57s. Estimated total time: 48h 32m 23s. Time estimates for 10 more iterations: 9m 42s, 100 more iterations: 1h 37m 4s, 500 more iterations: 8h 5m 23s. -[2025-11-10 02:09:23,404][__main__][INFO] - Starting iteration 509. -[2025-11-10 02:09:23,866][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 50 and human policies 1. -[2025-11-10 02:09:23,866][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:09:51,329][__main__][INFO] - Number of regex retries in iteration 509: 0 -[2025-11-10 02:09:51,330][__main__][INFO] - agents played in iteration 509 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:09:52,164][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:09:52,188][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:09:52,212][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:09:52,235][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:09:52,236][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:09:52,238][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:09:53,060][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:09:53,519][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:09:54,024][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:09:54,527][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:09:55,028][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:09:55,530][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:09:56,035][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:09:56,538][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:09:57,041][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:09:57,545][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:09:58,044][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:09:58,556][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:09:59,055][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:09:59,560][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:10:00,060][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:10:00,557][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:10:01,060][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:10:01,559][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:10:02,062][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:10:02,561][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:10:03,062][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:10:03,562][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:10:04,061][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:10:04,560][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:10:05,064][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:10:05,561][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:10:06,062][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:10:06,561][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:10:07,065][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:10:07,580][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:10:08,076][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:10:08,574][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:10:09,074][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:10:09,578][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:10:10,086][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:10:10,588][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:10:11,086][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:10:11,585][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:10:12,082][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:10:12,581][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:10:13,077][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:10:13,574][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:10:14,090][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:10:14,589][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:10:15,082][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:10:15,577][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:10:16,074][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:10:16,570][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:10:17,064][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:10:17,560][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:10:18,071][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:10:18,572][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:10:19,067][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:10:19,572][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:10:20,068][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:10:20,569][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:10:21,066][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:10:21,564][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:10:22,060][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:10:22,561][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:10:23,063][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:10:23,572][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:10:24,077][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:10:24,598][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:10:25,101][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10267 tokens. -[2025-11-10 02:10:25,944][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 55.70%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 -[2025-11-10 02:10:26,687][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:10:26,689][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:10:26,691][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:10:27,597][__main__][INFO] - Iteration 510 took 1m 3s (43.09% Gen, 55.49% Train). Generation: 27s, Training: 35s. Estimated remaining time: 44h 55m 4s. Estimated total time: 53h 6m 35s. Time estimates for 10 more iterations: 10m 37s, 100 more iterations: 1h 46m 13s, 500 more iterations: 8h 51m 5s. -[2025-11-10 02:10:27,599][__main__][INFO] - Starting iteration 510. -[2025-11-10 02:10:28,094][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 50 and human policies 1. -[2025-11-10 02:10:28,094][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:10:57,369][__main__][INFO] - Number of regex retries in iteration 510: 0 -[2025-11-10 02:10:57,371][__main__][INFO] - agents played in iteration 510 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:10:58,189][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:10:58,214][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:10:58,238][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:10:58,261][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:10:58,262][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:10:58,263][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:10:59,081][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:10:59,543][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:11:00,047][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:11:00,547][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:11:01,045][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:11:01,544][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:11:02,048][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:11:02,548][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:11:03,049][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:11:03,550][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:11:04,059][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:11:04,561][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:11:05,061][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:11:05,561][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:11:06,061][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:11:06,565][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:11:07,069][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:11:07,570][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:11:08,084][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:11:08,582][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:11:09,085][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:11:09,583][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:11:10,081][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:11:10,577][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:11:11,076][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:11:11,574][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:11:12,078][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:11:12,576][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:11:13,079][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:11:13,578][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:11:14,076][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:11:14,575][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:11:15,078][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:11:15,575][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:11:16,079][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:11:16,577][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:11:17,100][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:11:17,595][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:11:18,091][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:11:18,594][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:11:19,092][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:11:19,588][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:11:20,081][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:11:20,575][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:11:21,077][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:11:21,577][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:11:22,077][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:11:22,589][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:11:23,085][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:11:23,579][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:11:24,085][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:11:24,578][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:11:25,076][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:11:25,571][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:11:26,071][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:11:26,566][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:11:27,069][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:11:27,570][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:11:28,078][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:11:28,580][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:11:29,077][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:11:29,587][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:11:30,085][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:11:30,594][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:11:31,096][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10078 tokens. -[2025-11-10 02:11:31,936][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 55.65%, Block Peak % of device VRAM: 62.15%, ΔTime: 00:00:32 -[2025-11-10 02:11:32,584][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:11:32,586][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:11:32,588][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:11:34,306][__main__][INFO] - Iteration 511 took 1m 6s (44.22% Gen, 53.19% Train). Generation: 29s, Training: 35s. Estimated remaining time: 46h 58m 1s. Estimated total time: 55h 10m 39s. Time estimates for 10 more iterations: 11m 2s, 100 more iterations: 1h 50m 21s, 500 more iterations: 9h 11m 46s. -[2025-11-10 02:11:34,308][__main__][INFO] - Starting iteration 511. -[2025-11-10 02:11:34,784][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 51 and human policies 1. -[2025-11-10 02:11:34,785][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:11:53,916][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 02:12:05,141][__main__][INFO] - Number of regex retries in iteration 511: 1 -[2025-11-10 02:12:05,141][__main__][INFO] - agents played in iteration 511 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:12:05,948][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:12:05,971][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:12:05,996][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:12:06,019][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:12:06,020][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:12:06,021][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:12:06,838][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:12:07,297][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:12:07,803][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:12:08,304][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:12:08,814][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:12:09,312][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:12:09,810][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:12:10,307][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:12:10,806][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:12:11,319][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:12:11,818][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:12:12,324][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:12:12,823][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:12:13,321][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:12:13,822][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:12:14,323][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:12:14,820][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:12:15,317][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:12:15,815][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:12:16,316][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:12:16,814][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:12:17,313][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:12:17,820][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:12:18,320][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:12:18,819][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:12:19,317][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:12:19,823][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:12:20,326][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:12:20,824][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:12:21,320][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:12:21,819][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:12:22,314][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:12:22,816][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:12:23,316][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:12:23,813][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:12:24,319][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:12:24,815][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:12:25,315][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:12:25,814][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:12:26,313][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:12:26,809][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:12:27,306][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:12:27,802][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:12:28,314][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:12:28,809][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:12:29,305][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:12:29,805][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:12:30,300][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:12:30,807][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:12:31,305][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:12:31,823][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:12:32,323][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:12:32,822][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:12:33,324][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:12:33,824][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:12:34,324][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:12:34,826][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:12:35,326][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:12:35,825][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:12:36,326][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:12:36,827][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:12:37,331][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:12:37,830][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:12:38,330][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:12:38,831][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10125 tokens. -[2025-11-10 02:12:39,678][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 55.63%, Block Peak % of device VRAM: 62.40%, ΔTime: 00:00:32 -[2025-11-10 02:12:40,437][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:12:40,439][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:12:40,441][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:12:41,359][__main__][INFO] - Iteration 512 took 1m 6s (45.60% Gen, 53.02% Train). Generation: 30s, Training: 35s. Estimated remaining time: 47h 15m 3s. Estimated total time: 55h 28m 48s. Time estimates for 10 more iterations: 11m 5s, 100 more iterations: 1h 50m 57s, 500 more iterations: 9h 14m 48s. -[2025-11-10 02:12:41,362][__main__][INFO] - Starting iteration 512. -[2025-11-10 02:12:41,826][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 51 and human policies 1. -[2025-11-10 02:12:41,827][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:13:08,373][__main__][INFO] - Number of regex retries in iteration 512: 0 -[2025-11-10 02:13:08,375][__main__][INFO] - agents played in iteration 512 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:13:09,218][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:13:09,255][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:13:09,287][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:13:09,313][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:13:09,313][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:13:09,314][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:13:10,058][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:13:10,518][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:13:11,021][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:13:11,523][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:13:12,019][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:13:12,517][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:13:13,015][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:13:13,519][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:13:14,017][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:13:14,518][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:13:15,016][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:13:15,517][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:13:16,015][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:13:16,516][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:13:17,016][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:13:17,518][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:13:18,017][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:13:18,514][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:13:19,014][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:13:19,524][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:13:20,023][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:13:20,522][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:13:21,020][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:13:21,519][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:13:22,017][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:13:22,515][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:13:23,012][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:13:23,522][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:13:24,020][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:13:24,516][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:13:25,037][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:13:25,536][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:13:26,034][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:13:26,532][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:13:27,028][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:13:27,528][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:13:28,023][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:13:28,518][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:13:29,015][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:13:29,511][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:13:30,007][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:13:30,510][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:13:31,008][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:13:31,505][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:13:32,002][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:13:32,498][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:13:32,997][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:13:33,494][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:13:33,992][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:13:34,500][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:13:34,994][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:13:35,507][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:13:36,002][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:13:36,501][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:13:36,996][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:13:37,491][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:13:37,983][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:13:38,480][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:13:38,981][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:13:39,482][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:13:39,985][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:13:40,488][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:13:41,001][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:13:41,501][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:13:42,008][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10184 tokens. -[2025-11-10 02:13:42,836][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 55.71%, Block Peak % of device VRAM: 62.11%, ΔTime: 00:00:32 -[2025-11-10 02:13:43,504][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:13:43,506][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:13:43,507][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:13:44,334][__main__][INFO] - Iteration 513 took 1m 2s (42.47% Gen, 56.20% Train). Generation: 26s, Training: 35s. Estimated remaining time: 43h 50m 36s. Estimated total time: 52h 5m 24s. Time estimates for 10 more iterations: 10m 25s, 100 more iterations: 1h 44m 10s, 500 more iterations: 8h 40m 54s. -[2025-11-10 02:13:44,336][__main__][INFO] - Starting iteration 513. -[2025-11-10 02:13:44,806][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 51 and human policies 1. -[2025-11-10 02:13:44,807][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:14:11,602][__main__][INFO] - Number of regex retries in iteration 513: 0 -[2025-11-10 02:14:11,603][__main__][INFO] - agents played in iteration 513 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:14:12,370][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:14:12,393][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:14:12,416][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:14:12,438][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:14:12,439][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:14:12,440][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:14:13,248][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:14:13,717][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:14:14,222][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:14:14,725][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:14:15,224][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:14:15,721][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:14:16,234][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:14:16,732][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:14:17,230][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:14:17,733][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:14:18,233][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:14:18,736][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:14:19,237][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:14:19,737][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:14:20,250][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:14:20,750][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:14:21,249][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:14:21,748][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:14:22,247][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:14:22,750][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:14:23,252][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:14:23,747][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:14:24,245][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:14:24,739][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:14:25,240][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:14:25,735][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:14:26,237][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:14:26,739][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:14:27,240][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:14:27,736][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:14:28,245][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:14:28,741][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:14:29,240][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:14:29,734][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:14:30,229][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:14:30,734][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:14:31,230][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:14:31,725][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:14:32,221][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:14:32,719][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:14:33,215][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:14:33,713][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:14:34,209][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:14:34,721][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:14:35,218][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:14:35,714][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:14:36,216][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:14:36,714][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:14:37,232][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:14:37,728][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:14:38,223][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:14:38,736][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:14:39,231][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:14:39,725][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:14:40,221][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:14:40,716][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:14:41,216][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:14:41,712][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:14:42,206][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:14:42,718][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:14:43,212][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:14:43,707][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:14:44,202][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:14:44,696][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:14:45,193][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10067 tokens. -[2025-11-10 02:14:45,996][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.04%, Current % of VRAM taken: 55.53%, Block Peak % of device VRAM: 62.24%, ΔTime: 00:00:32 -[2025-11-10 02:14:46,749][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:14:46,751][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:14:46,753][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:14:47,678][__main__][INFO] - Iteration 514 took 1m 2s (42.62% Gen, 55.91% Train). Generation: 26s, Training: 35s. Estimated remaining time: 44h 7m 47s. Estimated total time: 52h 23m 37s. Time estimates for 10 more iterations: 10m 28s, 100 more iterations: 1h 44m 47s, 500 more iterations: 8h 43m 56s. -[2025-11-10 02:14:47,680][__main__][INFO] - Starting iteration 514. -[2025-11-10 02:14:48,154][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 51 and human policies 1. -[2025-11-10 02:14:48,155][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:15:11,997][__main__][INFO] - Number of regex retries in iteration 514: 0 -[2025-11-10 02:15:11,998][__main__][INFO] - agents played in iteration 514 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:15:12,801][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:15:12,824][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:15:12,847][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:15:12,869][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:15:12,870][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:15:12,871][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:15:13,691][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:15:14,150][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:15:14,657][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:15:15,160][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:15:15,663][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:15:16,164][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:15:16,670][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:15:17,171][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:15:17,673][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:15:18,171][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:15:18,671][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:15:19,175][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:15:19,673][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:15:20,176][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:15:20,680][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:15:21,180][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:15:21,679][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:15:22,179][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:15:22,677][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:15:23,182][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:15:23,682][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:15:24,182][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:15:24,681][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:15:25,176][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:15:25,673][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:15:26,170][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:15:26,668][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:15:27,169][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:15:27,667][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:15:28,163][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:15:28,661][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:15:29,158][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:15:29,655][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:15:30,152][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:15:30,649][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:15:31,147][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:15:31,643][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:15:32,139][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:15:32,653][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:15:33,151][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:15:33,648][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:15:34,150][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:15:34,650][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:15:35,161][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:15:35,659][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:15:36,156][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:15:36,655][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:15:37,153][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:15:37,660][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:15:38,156][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:15:38,655][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:15:39,168][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:15:39,665][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:15:40,160][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:15:40,657][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:15:41,154][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:15:41,653][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:15:42,149][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:15:42,644][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:15:43,141][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:15:43,638][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:15:44,135][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:15:44,633][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:15:45,132][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:15:45,658][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10150 tokens. -[2025-11-10 02:15:46,477][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 55.62%, Block Peak % of device VRAM: 62.47%, ΔTime: 00:00:32 -[2025-11-10 02:15:47,246][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:15:47,248][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:15:47,250][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:15:48,145][__main__][INFO] - Iteration 515 took 59s (39.74% Gen, 58.76% Train). Generation: 23s, Training: 35s. Estimated remaining time: 41h 42m 44s. Estimated total time: 49h 59m 35s. Time estimates for 10 more iterations: 9m 59s, 100 more iterations: 1h 39m 59s, 500 more iterations: 8h 19m 55s. -[2025-11-10 02:15:48,147][__main__][INFO] - Starting iteration 515. -[2025-11-10 02:15:48,610][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 51 and human policies 1. -[2025-11-10 02:15:48,611][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:16:16,374][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 02:16:17,256][__main__][INFO] - Number of regex retries in iteration 515: 1 -[2025-11-10 02:16:17,257][__main__][INFO] - agents played in iteration 515 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:16:18,025][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:16:18,053][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:16:18,079][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:16:18,102][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:16:18,103][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:16:18,103][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:16:18,851][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:16:19,308][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:16:19,811][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:16:20,309][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:16:20,812][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:16:21,315][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:16:21,814][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:16:22,312][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:16:22,813][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:16:23,316][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:16:23,815][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:16:24,312][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:16:24,812][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:16:25,315][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:16:25,817][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:16:26,314][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:16:26,812][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:16:27,313][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:16:27,816][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:16:28,312][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:16:28,813][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:16:29,311][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:16:29,812][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:16:30,311][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:16:30,809][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:16:31,307][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:16:31,806][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:16:32,302][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:16:32,801][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:16:33,296][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:16:33,791][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:16:34,292][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:16:34,787][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:16:35,286][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:16:35,789][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:16:36,286][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:16:36,786][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:16:37,282][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:16:37,777][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:16:38,274][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:16:38,770][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:16:39,270][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:16:39,766][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:16:40,264][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:16:40,760][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:16:41,258][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:16:41,754][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:16:42,252][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:16:42,751][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:16:43,247][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:16:43,748][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:16:44,244][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:16:44,738][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:16:45,237][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:16:45,734][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:16:46,235][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:16:46,730][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:16:47,229][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:16:47,733][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:16:48,234][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:16:48,747][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:16:49,247][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:16:49,750][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:16:50,252][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:16:50,755][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10209 tokens. -[2025-11-10 02:16:51,634][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.22%, ΔTime: 00:00:32 -[2025-11-10 02:16:52,309][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:16:52,310][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:16:52,312][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:16:53,146][__main__][INFO] - Iteration 516 took 1m 4s (44.39% Gen, 54.32% Train). Generation: 28s, Training: 35s. Estimated remaining time: 45h 28m 51s. Estimated total time: 53h 46m 48s. Time estimates for 10 more iterations: 10m 45s, 100 more iterations: 1h 47m 33s, 500 more iterations: 8h 57m 48s. -[2025-11-10 02:16:53,148][__main__][INFO] - Starting iteration 516. -[2025-11-10 02:16:53,614][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 51 and human policies 1. -[2025-11-10 02:16:53,614][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:17:08,972][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 02:17:20,822][__main__][INFO] - Number of regex retries in iteration 516: 1 -[2025-11-10 02:17:20,823][__main__][INFO] - agents played in iteration 516 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:17:21,608][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:17:21,635][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:17:21,662][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:17:21,685][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:17:21,686][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:17:21,686][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:17:22,440][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:17:22,896][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:17:23,403][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:17:23,902][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:17:24,399][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:17:24,897][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:17:25,398][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:17:25,895][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:17:26,393][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:17:26,900][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:17:27,399][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:17:27,899][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:17:28,425][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:17:28,926][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:17:29,431][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:17:29,933][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:17:30,435][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:17:30,935][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:17:31,434][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:17:31,931][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:17:32,439][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:17:32,940][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:17:33,456][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:17:33,956][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:17:34,455][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:17:34,956][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:17:35,454][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:17:35,965][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:17:36,459][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:17:36,954][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:17:37,450][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:17:37,948][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:17:38,442][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:17:38,936][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:17:39,432][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:17:39,932][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:17:40,426][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:17:40,922][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:17:41,419][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:17:41,916][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:17:42,409][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:17:42,904][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:17:43,399][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:17:43,896][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:17:44,395][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:17:44,892][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:17:45,386][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:17:45,886][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:17:46,384][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:17:46,878][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:17:47,373][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:17:47,872][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:17:48,367][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:17:48,861][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:17:49,359][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:17:49,858][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:17:50,372][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:17:50,878][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:17:51,380][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:17:51,880][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:17:52,380][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:17:52,882][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:17:53,383][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:17:53,883][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:17:54,390][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10102 tokens. -[2025-11-10 02:17:55,219][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 55.67%, Block Peak % of device VRAM: 62.27%, ΔTime: 00:00:32 -[2025-11-10 02:17:55,988][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:17:55,990][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:17:55,992][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:17:56,912][__main__][INFO] - Iteration 517 took 1m 3s (42.98% Gen, 55.56% Train). Generation: 27s, Training: 35s. Estimated remaining time: 44h 25m 57s. Estimated total time: 52h 44m 57s. Time estimates for 10 more iterations: 10m 32s, 100 more iterations: 1h 45m 29s, 500 more iterations: 8h 47m 29s. -[2025-11-10 02:17:56,914][__main__][INFO] - Starting iteration 517. -[2025-11-10 02:17:57,375][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 51 and human policies 1. -[2025-11-10 02:17:57,376][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:18:22,533][__main__][INFO] - Number of regex retries in iteration 517: 0 -[2025-11-10 02:18:22,535][__main__][INFO] - agents played in iteration 517 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:18:23,410][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:18:23,438][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:18:23,465][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:18:23,488][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:18:23,488][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:18:23,489][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:18:24,318][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:18:24,777][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:18:25,282][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:18:25,800][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:18:26,302][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:18:26,807][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:18:27,305][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:18:27,807][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:18:28,305][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:18:28,805][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:18:29,300][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:18:29,803][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:18:30,299][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:18:30,798][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:18:31,294][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:18:31,793][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:18:32,295][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:18:32,793][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:18:33,290][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:18:33,795][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:18:34,294][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:18:34,794][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:18:35,289][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:18:35,786][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:18:36,284][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:18:36,780][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:18:37,278][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:18:37,794][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:18:38,289][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:18:38,790][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:18:39,288][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:18:39,786][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:18:40,285][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:18:40,781][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:18:41,276][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:18:41,776][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:18:42,275][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:18:42,771][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:18:43,272][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:18:43,767][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:18:44,285][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:18:44,780][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:18:45,275][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:18:45,772][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:18:46,270][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:18:46,775][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:18:47,272][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:18:47,766][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:18:48,262][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:18:48,759][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:18:49,254][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:18:49,750][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:18:50,247][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:18:50,746][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:18:51,248][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:18:51,750][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:18:52,252][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:18:52,756][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:18:53,256][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:18:53,757][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:18:54,259][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:18:54,775][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:18:55,280][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:18:55,782][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:18:56,285][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10118 tokens. -[2025-11-10 02:18:57,105][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 55.60%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:32 -[2025-11-10 02:18:57,759][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:18:57,761][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:18:57,762][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:18:58,600][__main__][INFO] - Iteration 518 took 1m 1s (41.09% Gen, 57.54% Train). Generation: 25s, Training: 35s. Estimated remaining time: 42h 41m 16s. Estimated total time: 51h 1m 17s. Time estimates for 10 more iterations: 10m 12s, 100 more iterations: 1h 42m 2s, 500 more iterations: 8h 30m 12s. -[2025-11-10 02:18:58,603][__main__][INFO] - Starting iteration 518. -[2025-11-10 02:18:59,068][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 51 and human policies 1. -[2025-11-10 02:18:59,068][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:19:15,983][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 02:19:19,352][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 02:19:27,469][__main__][INFO] - Number of regex retries in iteration 518: 2 -[2025-11-10 02:19:27,470][__main__][INFO] - agents played in iteration 518 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:19:28,272][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:19:28,296][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:19:28,319][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:19:28,342][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:19:28,342][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:19:28,343][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:19:29,195][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:19:29,872][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:19:30,385][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:19:30,882][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:19:31,398][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:19:31,895][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:19:32,393][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:19:32,904][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:19:33,403][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:19:33,902][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:19:34,398][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:19:34,894][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:19:35,399][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:19:35,896][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:19:36,393][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:19:36,896][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:19:37,394][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:19:37,895][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:19:38,392][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:19:38,893][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:19:39,404][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:19:39,901][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:19:40,403][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:19:40,901][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:19:41,398][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:19:41,899][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:19:42,398][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:19:42,902][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:19:43,416][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:19:43,914][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:19:44,413][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:19:44,915][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:19:45,416][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:19:45,922][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:19:46,420][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:19:46,919][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:19:47,418][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:19:47,915][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:19:48,412][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:19:48,910][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:19:49,405][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:19:49,914][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:19:50,411][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:19:50,913][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:19:51,408][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:19:51,905][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:19:52,409][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:19:52,904][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:19:53,400][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:19:53,913][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:19:54,412][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:19:54,912][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:19:55,407][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:19:55,904][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:19:56,406][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:19:56,917][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:19:57,419][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:19:57,924][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:19:58,423][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:19:58,926][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:19:59,426][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:19:59,927][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:20:00,432][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:20:00,935][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:20:01,437][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10047 tokens. -[2025-11-10 02:20:02,276][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 55.56%, Block Peak % of device VRAM: 62.17%, ΔTime: 00:00:33 -[2025-11-10 02:20:03,027][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:20:03,029][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:20:03,031][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:20:03,905][__main__][INFO] - Iteration 519 took 1m 4s (43.80% Gen, 54.85% Train). Generation: 28s, Training: 35s. Estimated remaining time: 45h 40m 48s. Estimated total time: 54h 1m 55s. Time estimates for 10 more iterations: 10m 48s, 100 more iterations: 1h 48m 3s, 500 more iterations: 9h 0m 19s. -[2025-11-10 02:20:03,908][__main__][INFO] - Starting iteration 519. -[2025-11-10 02:20:04,387][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 51 and human policies 1. -[2025-11-10 02:20:04,387][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:20:36,220][__main__][INFO] - Number of regex retries in iteration 519: 0 -[2025-11-10 02:20:36,222][__main__][INFO] - agents played in iteration 519 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:20:37,053][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:20:37,083][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:20:37,110][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:20:37,147][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:20:37,148][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:20:37,148][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:20:37,911][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:20:38,366][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:20:38,876][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:20:39,379][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:20:39,875][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:20:40,373][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:20:40,874][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:20:41,372][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:20:41,874][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:20:42,371][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:20:42,870][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:20:43,366][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:20:43,864][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:20:44,363][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:20:44,861][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:20:45,362][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:20:45,862][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:20:46,367][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:20:46,867][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:20:47,368][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:20:47,869][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:20:48,369][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:20:48,873][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:20:49,371][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:20:49,877][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:20:50,379][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:20:50,876][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:20:51,374][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:20:51,871][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:20:52,369][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:20:52,864][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:20:53,361][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:20:53,855][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:20:54,355][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:20:54,856][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:20:55,356][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:20:55,859][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:20:56,380][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:20:56,882][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:20:57,383][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:20:57,881][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:20:58,387][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:20:58,883][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:20:59,378][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:20:59,885][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:21:00,380][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:21:00,877][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:21:01,373][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:21:01,868][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:21:02,373][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:21:02,926][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:21:03,435][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:21:03,939][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:21:04,443][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:21:04,947][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:21:05,445][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:21:05,949][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:21:06,453][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:21:06,955][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:21:07,461][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:21:07,962][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:21:08,463][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:21:08,969][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:21:09,469][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:21:09,970][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10124 tokens. -[2025-11-10 02:21:10,818][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 55.65%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:32 -[2025-11-10 02:21:11,485][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:21:11,487][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:21:11,489][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:21:12,300][__main__][INFO] - Iteration 520 took 1m 7s (46.88% Gen, 51.93% Train). Generation: 31s, Training: 35s. Estimated remaining time: 48h 13m 26s. Estimated total time: 56h 35m 42s. Time estimates for 10 more iterations: 11m 19s, 100 more iterations: 1h 53m 11s, 500 more iterations: 9h 25m 57s. -[2025-11-10 02:21:12,302][__main__][INFO] - Starting iteration 520. -[2025-11-10 02:21:12,783][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 51 and human policies 1. -[2025-11-10 02:21:12,784][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:21:39,750][__main__][INFO] - Number of regex retries in iteration 520: 0 -[2025-11-10 02:21:39,751][__main__][INFO] - agents played in iteration 520 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:21:40,526][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:21:40,554][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:21:40,580][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:21:40,603][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:21:40,604][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:21:40,605][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:21:41,358][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:21:41,818][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:21:42,320][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:21:42,820][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:21:43,318][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:21:43,822][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:21:44,331][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:21:44,829][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:21:45,327][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:21:45,825][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:21:46,323][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:21:46,830][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:21:47,331][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:21:47,827][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:21:48,327][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:21:48,826][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:21:49,332][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:21:49,838][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:21:50,337][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:21:50,840][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:21:51,337][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:21:51,837][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:21:52,338][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:21:52,843][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:21:53,344][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:21:53,838][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:21:54,335][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:21:54,831][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:21:55,325][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:21:55,822][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:21:56,320][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:21:56,820][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:21:57,327][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:21:57,824][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:21:58,320][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:21:58,817][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:21:59,315][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:21:59,820][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:22:00,316][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:22:00,813][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:22:01,325][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:22:01,821][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:22:02,318][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:22:02,814][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:22:03,309][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:22:03,815][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:22:04,312][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:22:04,809][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:22:05,324][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:22:05,823][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:22:06,320][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:22:06,821][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:22:07,327][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:22:07,836][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:22:08,338][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:22:08,839][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:22:09,336][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:22:09,836][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:22:10,342][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:22:10,844][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:22:11,346][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:22:11,847][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:22:12,348][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:22:12,854][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:22:13,357][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10075 tokens. -[2025-11-10 02:22:14,198][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 55.72%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:32 -[2025-11-10 02:22:14,939][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:22:14,940][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:22:14,942][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:22:16,710][__main__][INFO] - Iteration 521 took 1m 3s (42.18% Gen, 55.05% Train). Generation: 26s, Training: 35s. Estimated remaining time: 44h 53m 2s. Estimated total time: 53h 16m 22s. Time estimates for 10 more iterations: 10m 39s, 100 more iterations: 1h 46m 32s, 500 more iterations: 8h 52m 43s. -[2025-11-10 02:22:16,713][__main__][INFO] - Starting iteration 521. -[2025-11-10 02:22:17,192][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 52 and human policies 1. -[2025-11-10 02:22:17,193][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:22:35,580][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 02:22:37,805][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 02:22:46,931][__main__][INFO] - Number of regex retries in iteration 521: 2 -[2025-11-10 02:22:46,932][__main__][INFO] - agents played in iteration 521 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:22:47,744][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:22:47,772][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:22:47,799][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:22:47,822][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:22:47,823][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:22:47,824][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:22:48,576][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:22:49,029][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:22:49,529][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:22:50,037][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:22:50,534][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:22:51,030][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:22:51,530][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:22:52,031][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:22:52,534][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:22:53,030][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:22:53,527][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:22:54,036][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:22:54,533][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:22:55,028][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:22:55,528][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:22:56,024][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:22:56,529][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:22:57,029][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:22:57,529][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:22:58,046][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:22:58,542][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:22:59,044][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:22:59,544][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:23:00,045][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:23:00,548][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:23:01,041][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:23:01,537][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:23:02,041][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:23:02,558][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:23:03,059][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:23:03,556][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:23:04,053][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:23:04,556][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:23:05,049][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:23:05,543][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:23:06,057][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:23:06,556][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:23:07,054][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:23:07,560][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:23:08,054][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:23:08,558][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:23:09,059][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:23:09,563][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:23:10,068][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:23:10,569][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:23:11,070][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:23:11,572][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:23:12,074][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:23:12,579][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:23:13,077][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:23:13,579][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:23:14,080][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:23:14,579][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:23:15,078][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:23:15,576][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:23:16,077][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:23:16,579][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:23:17,083][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:23:17,585][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:23:18,087][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:23:18,592][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:23:19,094][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:23:19,597][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:23:20,102][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:23:20,602][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10079 tokens. -[2025-11-10 02:23:21,437][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 55.67%, Block Peak % of device VRAM: 62.18%, ΔTime: 00:00:32 -[2025-11-10 02:23:22,098][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:23:22,100][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:23:22,102][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:23:22,926][__main__][INFO] - Iteration 522 took 1m 5s (45.24% Gen, 53.50% Train). Generation: 29s, Training: 35s. Estimated remaining time: 46h 22m 18s. Estimated total time: 54h 46m 44s. Time estimates for 10 more iterations: 10m 57s, 100 more iterations: 1h 49m 33s, 500 more iterations: 9h 7m 47s. -[2025-11-10 02:23:22,928][__main__][INFO] - Starting iteration 522. -[2025-11-10 02:23:23,400][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 52 and human policies 1. -[2025-11-10 02:23:23,400][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:23:47,720][__main__][INFO] - Number of regex retries in iteration 522: 0 -[2025-11-10 02:23:47,721][__main__][INFO] - agents played in iteration 522 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:23:48,517][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:23:48,544][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:23:48,569][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:23:48,593][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:23:48,593][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:23:48,594][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:23:49,438][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:23:49,900][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:23:50,417][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:23:50,918][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:23:51,421][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:23:51,924][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:23:52,430][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:23:52,929][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:23:53,447][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:23:53,947][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:23:54,446][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:23:54,950][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:23:55,459][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:23:55,958][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:23:56,460][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:23:56,962][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:23:57,464][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:23:57,960][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:23:58,455][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:23:58,959][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:23:59,455][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:23:59,950][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:24:00,450][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:24:00,951][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:24:01,458][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:24:01,957][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:24:02,458][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:24:02,964][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:24:03,459][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:24:03,957][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:24:04,452][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:24:04,947][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:24:05,449][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:24:05,946][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:24:06,449][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:24:06,953][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:24:07,453][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:24:07,951][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:24:08,449][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:24:08,946][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:24:09,456][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:24:09,952][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:24:10,455][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:24:10,959][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:24:11,467][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:24:11,972][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:24:12,474][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:24:12,976][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:24:13,478][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:24:13,980][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:24:14,483][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:24:14,982][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:24:15,487][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:24:15,991][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:24:16,497][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:24:17,000][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:24:17,501][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:24:18,003][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:24:18,505][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:24:19,006][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:24:19,510][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:24:20,014][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:24:20,519][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:24:21,024][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:24:21,525][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10094 tokens. -[2025-11-10 02:24:22,385][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 55.65%, Block Peak % of device VRAM: 62.20%, ΔTime: 00:00:32 -[2025-11-10 02:24:23,115][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:24:23,118][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:24:23,121][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:24:24,066][__main__][INFO] - Iteration 523 took 1m 0s (40.09% Gen, 58.35% Train). Generation: 24s, Training: 35s. Estimated remaining time: 42h 7m 52s. Estimated total time: 50h 33m 19s. Time estimates for 10 more iterations: 10m 6s, 100 more iterations: 1h 41m 6s, 500 more iterations: 8h 25m 33s. -[2025-11-10 02:24:24,069][__main__][INFO] - Starting iteration 523. -[2025-11-10 02:24:24,534][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 52 and human policies 1. -[2025-11-10 02:24:24,535][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:24:47,928][__main__][INFO] - Number of regex retries in iteration 523: 0 -[2025-11-10 02:24:47,930][__main__][INFO] - agents played in iteration 523 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:24:48,721][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:24:48,747][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:24:48,773][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:24:48,796][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:24:48,797][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:24:48,798][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:24:49,621][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:24:50,080][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:24:50,587][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:24:51,095][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:24:51,598][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:24:52,099][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:24:52,602][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:24:53,106][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:24:53,611][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:24:54,118][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:24:54,619][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:24:55,119][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:24:55,651][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:24:56,156][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:24:56,655][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:24:57,161][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:24:57,663][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:24:58,167][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:24:58,664][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:24:59,168][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:24:59,674][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:25:00,173][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:25:00,676][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:25:01,175][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:25:01,672][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:25:02,177][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:25:02,678][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:25:03,179][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:25:03,679][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:25:04,178][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:25:04,680][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:25:05,180][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:25:05,686][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:25:06,185][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:25:06,681][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:25:07,178][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:25:07,676][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:25:08,175][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:25:08,674][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:25:09,169][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:25:09,666][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:25:10,167][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:25:10,674][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:25:11,175][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:25:11,676][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:25:12,177][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:25:12,682][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:25:13,186][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:25:13,690][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:25:14,195][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:25:14,699][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:25:15,201][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:25:15,705][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:25:16,207][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:25:16,713][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:25:17,213][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:25:17,717][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:25:18,228][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:25:18,729][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:25:19,233][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:25:19,742][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:25:20,245][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:25:20,760][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:25:21,265][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:25:21,767][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10187 tokens. -[2025-11-10 02:25:22,686][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 55.59%, Block Peak % of device VRAM: 62.41%, ΔTime: 00:00:33 -[2025-11-10 02:25:23,348][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:25:23,351][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:25:23,352][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:25:24,172][__main__][INFO] - Iteration 524 took 59s (39.23% Gen, 59.39% Train). Generation: 23s, Training: 35s. Estimated remaining time: 41h 15m 26s. Estimated total time: 49h 41m 53s. Time estimates for 10 more iterations: 9m 56s, 100 more iterations: 1h 39m 23s, 500 more iterations: 8h 16m 58s. -[2025-11-10 02:25:24,174][__main__][INFO] - Starting iteration 524. -[2025-11-10 02:25:24,642][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 52 and human policies 1. -[2025-11-10 02:25:24,643][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:25:55,255][__main__][INFO] - Number of regex retries in iteration 524: 0 -[2025-11-10 02:25:55,256][__main__][INFO] - agents played in iteration 524 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:25:56,048][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:25:56,083][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:25:56,112][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:25:56,136][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:25:56,137][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:25:56,138][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:25:56,965][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:25:57,427][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:25:57,936][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:25:58,435][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:25:58,934][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:25:59,438][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:25:59,941][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:26:00,440][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:26:00,936][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:26:01,457][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:26:01,959][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:26:02,455][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:26:02,957][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:26:03,453][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:26:03,960][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:26:04,460][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:26:04,959][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:26:05,485][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:26:05,985][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:26:06,483][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:26:06,981][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:26:07,479][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:26:07,985][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:26:08,483][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:26:08,981][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:26:09,480][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:26:09,981][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:26:10,484][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:26:10,984][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:26:11,486][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:26:11,983][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:26:12,482][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:26:12,982][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:26:13,490][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:26:13,995][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:26:14,497][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:26:14,997][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:26:15,500][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:26:15,996][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:26:16,497][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:26:16,993][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:26:17,488][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:26:17,982][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:26:18,478][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:26:18,975][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:26:19,471][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:26:19,965][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:26:20,467][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:26:20,966][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:26:21,471][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:26:21,976][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:26:22,479][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:26:22,981][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:26:23,481][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:26:23,988][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:26:24,492][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:26:24,995][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:26:25,496][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:26:25,996][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:26:26,497][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:26:27,000][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:26:27,509][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:26:28,013][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:26:28,515][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:26:29,019][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10138 tokens. -[2025-11-10 02:26:29,892][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 55.56%, Block Peak % of device VRAM: 62.20%, ΔTime: 00:00:32 -[2025-11-10 02:26:30,631][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:26:30,633][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:26:30,635][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:26:31,591][__main__][INFO] - Iteration 525 took 1m 6s (45.72% Gen, 52.84% Train). Generation: 30s, Training: 35s. Estimated remaining time: 47h 19m 55s. Estimated total time: 55h 47m 30s. Time estimates for 10 more iterations: 11m 9s, 100 more iterations: 1h 51m 35s, 500 more iterations: 9h 17m 55s. -[2025-11-10 02:26:31,594][__main__][INFO] - Starting iteration 525. -[2025-11-10 02:26:32,061][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 52 and human policies 1. -[2025-11-10 02:26:32,061][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:26:58,462][__main__][INFO] - Number of regex retries in iteration 525: 0 -[2025-11-10 02:26:58,464][__main__][INFO] - agents played in iteration 525 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:26:59,272][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:26:59,296][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:26:59,320][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:26:59,344][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:26:59,344][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:26:59,346][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:27:00,168][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:27:00,630][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:27:01,141][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:27:01,641][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:27:02,144][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:27:02,643][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:27:03,145][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:27:03,646][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:27:04,144][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:27:04,652][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:27:05,153][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:27:05,654][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:27:06,159][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:27:06,659][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:27:07,181][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:27:07,676][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:27:08,175][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:27:08,684][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:27:09,182][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:27:09,689][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:27:10,204][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:27:10,702][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:27:11,203][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:27:11,701][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:27:12,204][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:27:12,701][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:27:13,199][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:27:13,699][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:27:14,198][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:27:14,695][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:27:15,198][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:27:15,695][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:27:16,196][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:27:16,698][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:27:17,195][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:27:17,697][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:27:18,196][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:27:18,696][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:27:19,198][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:27:19,693][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:27:20,189][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:27:20,686][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:27:21,185][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:27:21,684][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:27:22,190][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:27:22,690][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:27:23,195][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:27:23,694][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:27:24,193][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:27:24,695][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:27:25,197][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:27:25,698][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:27:26,199][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:27:26,699][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:27:27,205][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:27:27,704][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:27:28,204][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:27:28,708][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:27:29,208][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:27:29,708][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:27:30,211][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:27:30,712][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:27:31,219][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:27:31,725][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:27:32,228][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10049 tokens. -[2025-11-10 02:27:33,083][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 55.64%, Block Peak % of device VRAM: 62.36%, ΔTime: 00:00:32 -[2025-11-10 02:27:33,722][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:27:33,724][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:27:33,726][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:27:34,541][__main__][INFO] - Iteration 526 took 1m 2s (42.26% Gen, 56.44% Train). Generation: 26s, Training: 35s. Estimated remaining time: 43h 35m 24s. Estimated total time: 52h 4m 2s. Time estimates for 10 more iterations: 10m 24s, 100 more iterations: 1h 44m 8s, 500 more iterations: 8h 40m 40s. -[2025-11-10 02:27:34,543][__main__][INFO] - Starting iteration 526. -[2025-11-10 02:27:35,018][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 52 and human policies 1. -[2025-11-10 02:27:35,018][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:28:05,212][__main__][INFO] - Number of regex retries in iteration 526: 0 -[2025-11-10 02:28:05,214][__main__][INFO] - agents played in iteration 526 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:28:06,003][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:28:06,031][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:28:06,057][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:28:06,081][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:28:06,082][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:28:06,083][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:28:06,948][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:28:07,406][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:28:07,915][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:28:08,423][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:28:08,923][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:28:09,424][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:28:09,924][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:28:10,420][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:28:10,922][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:28:11,420][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:28:11,921][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:28:12,423][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:28:12,922][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:28:13,420][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:28:13,921][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:28:14,427][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:28:14,925][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:28:15,420][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:28:15,918][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:28:16,416][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:28:16,916][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:28:17,413][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:28:17,912][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:28:18,411][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:28:18,910][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:28:19,408][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:28:19,909][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:28:20,413][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:28:20,916][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:28:21,415][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:28:21,914][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:28:22,414][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:28:22,913][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:28:23,410][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:28:23,905][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:28:24,401][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:28:24,906][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:28:25,407][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:28:25,909][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:28:26,410][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:28:26,908][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:28:27,413][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:28:27,917][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:28:28,418][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:28:28,920][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:28:29,421][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:28:29,922][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:28:30,423][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:28:30,923][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:28:31,431][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:28:31,937][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:28:32,438][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:28:32,945][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:28:33,447][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:28:33,946][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:28:34,447][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:28:34,948][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:28:35,458][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:28:35,960][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:28:36,466][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:28:36,987][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:28:37,491][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:28:38,003][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:28:38,508][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:28:39,012][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10122 tokens. -[2025-11-10 02:28:39,896][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 55.59%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 -[2025-11-10 02:28:40,629][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:28:40,630][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:28:40,633][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:28:41,630][__main__][INFO] - Iteration 527 took 1m 6s (45.33% Gen, 53.17% Train). Generation: 30s, Training: 35s. Estimated remaining time: 47h 0m 54s. Estimated total time: 55h 30m 39s. Time estimates for 10 more iterations: 11m 6s, 100 more iterations: 1h 51m 1s, 500 more iterations: 9h 15m 6s. -[2025-11-10 02:28:41,633][__main__][INFO] - Starting iteration 527. -[2025-11-10 02:28:42,126][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 52 and human policies 1. -[2025-11-10 02:28:42,126][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:28:54,452][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 02:29:05,309][__main__][INFO] - Number of regex retries in iteration 527: 1 -[2025-11-10 02:29:05,311][__main__][INFO] - agents played in iteration 527 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:29:06,228][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:29:06,265][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:29:06,294][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:29:06,320][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:29:06,321][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:29:06,321][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:29:07,189][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:29:07,654][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:29:08,157][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:29:08,672][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:29:09,173][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:29:09,694][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:29:10,195][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:29:10,696][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:29:11,208][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:29:11,716][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:29:12,220][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:29:12,724][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:29:13,226][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:29:13,730][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:29:14,230][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:29:14,727][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:29:15,228][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:29:15,725][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:29:16,223][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:29:16,717][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:29:17,214][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:29:17,714][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:29:18,208][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:29:18,703][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:29:19,199][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:29:19,696][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:29:20,193][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:29:20,693][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:29:21,188][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:29:21,699][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:29:22,195][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:29:22,692][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:29:23,190][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:29:23,687][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:29:24,196][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:29:24,700][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:29:25,197][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:29:25,697][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:29:26,195][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:29:26,694][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:29:27,195][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:29:27,703][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:29:28,216][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:29:28,711][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:29:29,212][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:29:29,724][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:29:30,222][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:29:30,724][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:29:31,229][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:29:31,726][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:29:32,232][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:29:32,737][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:29:33,237][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:29:33,743][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:29:34,244][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:29:34,750][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:29:35,257][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:29:35,767][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:29:36,269][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:29:36,769][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:29:37,272][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:29:37,789][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:29:38,294][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:29:38,798][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:29:39,302][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10031 tokens. -[2025-11-10 02:29:40,157][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 55.61%, Block Peak % of device VRAM: 62.15%, ΔTime: 00:00:32 -[2025-11-10 02:29:40,814][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:29:40,816][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:29:40,818][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:29:41,703][__main__][INFO] - Iteration 528 took 59s (38.91% Gen, 59.60% Train). Generation: 23s, Training: 35s. Estimated remaining time: 41h 8m 8s. Estimated total time: 49h 38m 53s. Time estimates for 10 more iterations: 9m 55s, 100 more iterations: 1h 39m 17s, 500 more iterations: 8h 16m 28s. -[2025-11-10 02:29:41,705][__main__][INFO] - Starting iteration 528. -[2025-11-10 02:29:42,169][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 52 and human policies 1. -[2025-11-10 02:29:42,170][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:30:14,269][__main__][INFO] - Number of regex retries in iteration 528: 0 -[2025-11-10 02:30:14,270][__main__][INFO] - agents played in iteration 528 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:30:15,057][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:30:15,086][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:30:15,113][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:30:15,136][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:30:15,137][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:30:15,138][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:30:15,909][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:30:16,368][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:30:16,868][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:30:17,366][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:30:17,862][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:30:18,359][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:30:18,857][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:30:19,358][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:30:19,855][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:30:20,356][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:30:20,860][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:30:21,361][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:30:21,863][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:30:22,367][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:30:22,866][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:30:23,368][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:30:23,869][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:30:24,368][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:30:24,867][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:30:25,366][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:30:25,869][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:30:26,368][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:30:26,869][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:30:27,364][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:30:27,861][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:30:28,356][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:30:28,859][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:30:29,356][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:30:29,851][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:30:30,348][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:30:30,842][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:30:31,336][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:30:31,832][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:30:32,329][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:30:32,828][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:30:33,326][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:30:33,820][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:30:34,318][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:30:34,817][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:30:35,315][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:30:35,816][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:30:36,322][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:30:36,832][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:30:37,333][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:30:37,833][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:30:38,334][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:30:38,839][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:30:39,340][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:30:39,843][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:30:40,349][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:30:40,852][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:30:41,353][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:30:41,854][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:30:42,355][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:30:42,858][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:30:43,361][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:30:43,863][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:30:44,368][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:30:44,874][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:30:45,374][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:30:45,874][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:30:46,382][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:30:46,883][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:30:47,387][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:30:47,901][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10118 tokens. -[2025-11-10 02:30:48,789][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 55.61%, Block Peak % of device VRAM: 62.28%, ΔTime: 00:00:32 -[2025-11-10 02:30:49,532][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:30:49,534][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:30:49,536][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:30:50,482][__main__][INFO] - Iteration 529 took 1m 8s (46.99% Gen, 51.62% Train). Generation: 32s, Training: 35s. Estimated remaining time: 48h 23m 45s. Estimated total time: 56h 55m 39s. Time estimates for 10 more iterations: 11m 23s, 100 more iterations: 1h 53m 51s, 500 more iterations: 9h 29m 16s. -[2025-11-10 02:30:50,484][__main__][INFO] - Starting iteration 529. -[2025-11-10 02:30:50,970][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 52 and human policies 1. -[2025-11-10 02:30:50,971][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:31:04,866][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 02:31:07,014][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 02:31:17,743][__main__][INFO] - Number of regex retries in iteration 529: 2 -[2025-11-10 02:31:17,744][__main__][INFO] - agents played in iteration 529 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:31:18,570][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:31:18,596][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:31:18,622][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:31:18,646][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:31:18,646][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:31:18,647][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:31:19,503][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:31:19,960][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:31:20,466][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:31:20,972][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:31:21,472][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:31:21,972][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:31:22,472][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:31:22,976][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:31:23,473][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:31:23,969][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:31:24,465][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:31:24,960][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:31:25,456][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:31:25,954][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:31:26,453][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:31:26,949][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:31:27,446][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:31:27,944][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:31:28,444][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:31:28,944][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:31:29,451][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:31:29,951][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:31:30,449][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:31:30,979][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:31:31,478][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:31:31,975][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:31:32,473][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:31:32,971][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:31:33,466][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:31:33,961][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:31:34,455][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:31:34,950][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:31:35,448][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:31:35,951][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:31:36,451][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:31:36,950][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:31:37,449][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:31:37,944][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:31:38,439][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:31:38,940][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:31:39,437][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:31:39,938][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:31:40,435][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:31:40,932][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:31:41,436][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:31:41,937][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:31:42,439][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:31:42,939][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:31:43,440][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:31:43,944][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:31:44,447][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:31:44,952][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:31:45,459][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:31:45,961][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:31:46,464][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:31:46,968][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:31:47,478][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:31:47,980][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:31:48,480][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:31:48,981][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:31:49,491][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:31:49,997][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:31:50,500][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:31:51,003][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:31:51,505][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10020 tokens. -[2025-11-10 02:31:52,356][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.04%, Current % of VRAM taken: 55.53%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:32 -[2025-11-10 02:31:53,005][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:31:53,007][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:31:53,009][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:31:53,918][__main__][INFO] - Iteration 530 took 1m 2s (42.53% Gen, 56.02% Train). Generation: 26s, Training: 35s. Estimated remaining time: 43h 54m 27s. Estimated total time: 52h 27m 24s. Time estimates for 10 more iterations: 10m 29s, 100 more iterations: 1h 44m 54s, 500 more iterations: 8h 44m 34s. -[2025-11-10 02:31:53,920][__main__][INFO] - Starting iteration 530. -[2025-11-10 02:31:54,420][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 52 and human policies 1. -[2025-11-10 02:31:54,421][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:32:11,608][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 02:32:22,346][__main__][INFO] - Number of regex retries in iteration 530: 1 -[2025-11-10 02:32:22,346][__main__][INFO] - agents played in iteration 530 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:32:23,221][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:32:23,257][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:32:23,282][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:32:23,305][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:32:23,305][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:32:23,306][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:32:24,129][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:32:24,591][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:32:25,100][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:32:25,600][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:32:26,124][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:32:26,624][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:32:27,121][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:32:27,621][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:32:28,123][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:32:28,618][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:32:29,116][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:32:29,615][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:32:30,132][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:32:30,630][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:32:31,125][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:32:31,625][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:32:32,121][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:32:32,631][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:32:33,127][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:32:33,627][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:32:34,125][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:32:34,620][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:32:35,117][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:32:35,614][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:32:36,110][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:32:36,629][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:32:37,127][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:32:37,624][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:32:38,120][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:32:38,618][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:32:39,131][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:32:39,639][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:32:40,136][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:32:40,632][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:32:41,130][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:32:41,626][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:32:42,127][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:32:42,621][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:32:43,118][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:32:43,616][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:32:44,111][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:32:44,621][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:32:45,114][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:32:45,613][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:32:46,115][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:32:46,614][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:32:47,111][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:32:47,610][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:32:48,109][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:32:48,609][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:32:49,109][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:32:49,609][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:32:50,108][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:32:50,605][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:32:51,120][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:32:51,620][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:32:52,128][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:32:52,628][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:32:53,127][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:32:53,637][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:32:54,140][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:32:54,640][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:32:55,149][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:32:55,651][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:32:56,151][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10125 tokens. -[2025-11-10 02:32:56,972][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 55.61%, Block Peak % of device VRAM: 62.16%, ΔTime: 00:00:32 -[2025-11-10 02:32:57,718][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:32:57,720][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:32:57,722][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:32:59,547][__main__][INFO] - Iteration 531 took 1m 5s (42.88% Gen, 54.32% Train). Generation: 27s, Training: 35s. Estimated remaining time: 45h 42m 18s. Estimated total time: 54h 16m 21s. Time estimates for 10 more iterations: 10m 51s, 100 more iterations: 1h 48m 32s, 500 more iterations: 9h 2m 43s. -[2025-11-10 02:32:59,550][__main__][INFO] - Starting iteration 531. -[2025-11-10 02:33:00,019][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 53 and human policies 1. -[2025-11-10 02:33:00,020][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:33:11,108][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 02:33:21,107][__main__][INFO] - Number of regex retries in iteration 531: 1 -[2025-11-10 02:33:21,108][__main__][INFO] - agents played in iteration 531 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:33:21,979][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:33:22,002][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:33:22,025][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:33:22,048][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:33:22,049][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:33:22,049][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:33:22,885][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:33:23,345][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:33:23,853][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:33:24,351][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:33:24,855][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:33:25,363][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:33:25,861][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:33:26,357][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:33:26,852][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:33:27,351][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:33:27,846][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:33:28,346][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:33:28,867][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:33:29,369][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:33:29,872][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:33:30,368][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:33:30,865][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:33:31,360][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:33:31,859][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:33:32,355][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:33:32,854][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:33:33,352][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:33:33,846][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:33:34,352][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:33:34,853][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:33:35,348][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:33:35,857][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:33:36,355][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:33:36,856][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:33:37,356][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:33:37,856][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:33:38,364][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:33:38,861][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:33:39,358][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:33:39,862][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:33:40,363][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:33:40,861][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:33:41,359][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:33:41,855][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:33:42,362][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:33:42,857][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:33:43,358][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:33:43,856][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:33:44,355][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:33:44,859][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:33:45,354][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:33:45,850][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:33:46,362][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:33:46,858][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:33:47,355][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:33:47,850][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:33:48,352][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:33:48,865][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:33:49,366][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:33:49,868][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:33:50,367][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:33:50,865][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:33:51,371][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:33:51,872][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:33:52,374][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:33:52,878][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:33:53,379][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:33:53,888][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:33:54,389][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:33:54,892][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10003 tokens. -[2025-11-10 02:33:55,742][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 55.61%, Block Peak % of device VRAM: 62.24%, ΔTime: 00:00:32 -[2025-11-10 02:33:56,451][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:33:56,452][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:33:56,454][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:33:57,386][__main__][INFO] - Iteration 532 took 57s (36.76% Gen, 61.61% Train). Generation: 21s, Training: 35s. Estimated remaining time: 39h 13m 21s. Estimated total time: 47h 48m 21s. Time estimates for 10 more iterations: 9m 33s, 100 more iterations: 1h 35m 36s, 500 more iterations: 7h 58m 3s. -[2025-11-10 02:33:57,397][__main__][INFO] - Starting iteration 532. -[2025-11-10 02:33:58,090][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 53 and human policies 1. -[2025-11-10 02:33:58,091][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:34:19,913][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 02:34:27,809][__main__][INFO] - Number of regex retries in iteration 532: 1 -[2025-11-10 02:34:27,810][__main__][INFO] - agents played in iteration 532 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:34:28,596][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:34:28,619][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:34:28,645][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:34:28,668][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:34:28,669][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:34:28,670][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:34:29,506][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:34:29,965][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:34:30,471][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:34:30,977][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:34:31,473][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:34:31,969][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:34:32,468][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:34:32,968][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:34:33,467][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:34:33,966][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:34:34,464][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:34:34,965][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:34:35,462][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:34:35,960][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:34:36,458][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:34:36,958][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:34:37,456][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:34:37,955][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:34:38,454][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:34:38,951][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:34:39,448][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:34:39,949][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:34:40,445][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:34:40,942][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:34:41,438][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:34:41,936][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:34:42,435][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:34:42,938][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:34:43,435][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:34:43,930][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:34:44,424][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:34:44,919][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:34:45,416][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:34:45,916][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:34:46,416][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:34:46,915][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:34:47,413][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:34:47,915][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:34:48,411][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:34:48,910][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:34:49,413][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:34:49,913][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:34:50,415][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:34:50,916][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:34:51,421][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:34:51,921][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:34:52,423][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:34:52,922][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:34:53,424][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:34:53,926][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:34:54,427][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:34:54,928][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:34:55,430][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:34:55,936][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:34:56,438][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:34:56,938][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:34:57,440][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:34:57,941][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:34:58,453][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:34:58,954][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:34:59,461][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:34:59,961][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:35:00,464][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:35:00,968][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:35:01,471][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9990 tokens. -[2025-11-10 02:35:02,330][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:32 -[2025-11-10 02:35:03,066][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:35:03,068][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:35:03,070][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:35:04,045][__main__][INFO] - Iteration 533 took 1m 5s (45.05% Gen, 53.46% Train). Generation: 29s, Training: 35s. Estimated remaining time: 46h 21m 40s. Estimated total time: 54h 57m 47s. Time estimates for 10 more iterations: 10m 59s, 100 more iterations: 1h 49m 55s, 500 more iterations: 9h 9m 37s. -[2025-11-10 02:35:04,048][__main__][INFO] - Starting iteration 533. -[2025-11-10 02:35:04,515][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 53 and human policies 1. -[2025-11-10 02:35:04,516][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:35:30,270][__main__][INFO] - Number of regex retries in iteration 533: 0 -[2025-11-10 02:35:30,271][__main__][INFO] - agents played in iteration 533 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:35:31,285][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:35:31,311][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:35:31,337][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:35:31,360][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:35:31,360][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:35:31,361][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:35:32,167][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:35:32,625][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:35:33,135][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:35:33,640][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:35:34,145][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:35:34,641][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:35:35,139][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:35:35,648][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:35:36,145][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:35:36,646][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:35:37,142][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:35:37,642][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:35:38,141][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:35:38,638][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:35:39,135][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:35:39,637][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:35:40,134][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:35:40,632][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:35:41,136][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:35:41,637][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:35:42,138][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:35:42,640][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:35:43,139][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:35:43,653][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:35:44,151][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:35:44,664][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:35:45,161][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:35:45,659][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:35:46,169][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:35:46,674][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:35:47,182][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:35:47,681][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:35:48,177][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:35:48,692][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:35:49,195][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:35:49,694][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:35:50,193][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:35:50,694][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:35:51,201][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:35:51,701][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:35:52,201][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:35:52,701][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:35:53,198][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:35:53,697][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:35:54,196][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:35:54,693][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:35:55,763][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:35:57,026][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:35:57,532][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:35:58,033][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:35:58,558][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:35:59,064][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:35:59,565][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:36:00,070][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:36:00,571][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:36:01,076][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:36:01,582][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:36:02,089][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:36:02,593][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:36:03,092][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:36:03,601][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:36:04,104][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:36:04,605][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:36:05,110][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:36:05,609][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10139 tokens. -[2025-11-10 02:36:06,572][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 55.61%, Block Peak % of device VRAM: 62.20%, ΔTime: 00:00:34 -[2025-11-10 02:36:07,243][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:36:07,245][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:36:07,247][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:36:08,094][__main__][INFO] - Iteration 534 took 1m 3s (40.51% Gen, 58.16% Train). Generation: 25s, Training: 36s. Estimated remaining time: 44h 21m 49s. Estimated total time: 52h 59m 0s. Time estimates for 10 more iterations: 10m 35s, 100 more iterations: 1h 45m 58s, 500 more iterations: 8h 49m 50s. -[2025-11-10 02:36:08,101][__main__][INFO] - Starting iteration 534. -[2025-11-10 02:36:08,569][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 53 and human policies 1. -[2025-11-10 02:36:08,570][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:36:30,179][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 02:36:41,232][__main__][INFO] - Number of regex retries in iteration 534: 1 -[2025-11-10 02:36:41,233][__main__][INFO] - agents played in iteration 534 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:36:42,039][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:36:42,062][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:36:42,087][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:36:42,110][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:36:42,111][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:36:42,111][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:36:42,956][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:36:43,419][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:36:43,925][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:36:44,433][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:36:44,932][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:36:45,429][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:36:45,930][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:36:46,426][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:36:46,924][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:36:47,425][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:36:47,924][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:36:48,421][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:36:48,919][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:36:49,416][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:36:49,916][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:36:50,418][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:36:50,917][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:36:51,418][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:36:51,919][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:36:52,423][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:36:52,921][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:36:53,421][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:36:53,922][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:36:54,423][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:36:54,920][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:36:55,418][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:36:55,919][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:36:56,429][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:36:56,933][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:36:57,435][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:36:57,947][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:36:58,446][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:36:58,949][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:36:59,459][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:36:59,963][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:37:00,501][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:37:01,007][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:37:01,511][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:37:02,022][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:37:02,528][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:37:03,037][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:37:03,539][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:37:04,043][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:37:04,551][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:37:05,056][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:37:05,563][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:37:06,066][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:37:06,570][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:37:07,073][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:37:07,573][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:37:08,073][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:37:08,582][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:37:09,080][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:37:09,588][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:37:10,089][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:37:10,592][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:37:11,090][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:37:11,589][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:37:12,086][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:37:12,585][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:37:13,084][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:37:13,601][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:37:14,101][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:37:14,606][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:37:15,111][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10184 tokens. -[2025-11-10 02:37:15,888][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.29%, Current % of VRAM taken: 55.79%, Block Peak % of device VRAM: 62.49%, ΔTime: 00:00:32 -[2025-11-10 02:37:16,631][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:37:16,633][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:37:16,635][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:37:17,639][__main__][INFO] - Iteration 535 took 1m 9s (47.29% Gen, 51.25% Train). Generation: 32s, Training: 35s. Estimated remaining time: 48h 55m 11s. Estimated total time: 57h 33m 32s. Time estimates for 10 more iterations: 11m 30s, 100 more iterations: 1h 55m 7s, 500 more iterations: 9h 35m 35s. -[2025-11-10 02:37:17,641][__main__][INFO] - Starting iteration 535. -[2025-11-10 02:37:18,106][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 53 and human policies 1. -[2025-11-10 02:37:18,106][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:37:46,195][__main__][INFO] - Number of regex retries in iteration 535: 0 -[2025-11-10 02:37:46,196][__main__][INFO] - agents played in iteration 535 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:37:46,978][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.67%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:37:47,007][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.67%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:37:47,033][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.67%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:37:47,056][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.67%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:37:47,056][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:37:47,057][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:37:47,812][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:37:48,269][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:37:48,770][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:37:49,273][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:37:49,769][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:37:50,271][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:37:50,770][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:37:51,265][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:37:51,762][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:37:52,261][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:37:52,758][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:37:53,258][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:37:53,763][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:37:54,272][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:37:55,023][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:37:56,231][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:37:56,739][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:37:57,240][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:37:57,743][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:37:58,250][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:37:58,756][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:37:59,260][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:37:59,761][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:38:00,265][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:38:00,783][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:38:01,286][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:38:01,787][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:38:02,295][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:38:02,795][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:38:03,298][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:38:03,802][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:38:04,303][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:38:04,806][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:38:05,306][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:38:05,814][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:38:06,316][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:38:06,819][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:38:07,320][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:38:07,821][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:38:08,326][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:38:08,826][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:38:09,329][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:38:09,835][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:38:10,335][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:38:10,837][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:38:11,340][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:38:11,840][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:38:12,341][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:38:12,842][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:38:13,343][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:38:13,846][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:38:14,348][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:38:14,847][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:38:15,354][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:38:15,854][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:38:16,355][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:38:16,852][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:38:17,379][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:38:17,880][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:38:18,378][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:38:18,878][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:38:19,382][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:38:19,880][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:38:20,400][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:38:20,900][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10159 tokens. -[2025-11-10 02:38:21,722][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 55.66%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:33 -[2025-11-10 02:38:22,355][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:38:22,357][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:38:22,358][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:38:23,193][__main__][INFO] - Iteration 536 took 1m 5s (43.16% Gen, 55.56% Train). Generation: 28s, Training: 36s. Estimated remaining time: 45h 34m 57s. Estimated total time: 54h 14m 23s. Time estimates for 10 more iterations: 10m 50s, 100 more iterations: 1h 48m 28s, 500 more iterations: 9h 2m 23s. -[2025-11-10 02:38:23,195][__main__][INFO] - Starting iteration 536. -[2025-11-10 02:38:23,659][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 53 and human policies 1. -[2025-11-10 02:38:23,660][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:38:55,180][__main__][INFO] - Number of regex retries in iteration 536: 0 -[2025-11-10 02:38:55,181][__main__][INFO] - agents played in iteration 536 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:38:55,976][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:38:56,000][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:38:56,022][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:38:56,044][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:38:56,045][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:38:56,046][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:38:56,875][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:38:57,334][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:38:57,838][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:38:58,338][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:38:58,837][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:38:59,335][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:38:59,832][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:39:00,327][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:39:00,828][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:39:01,324][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:39:01,829][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:39:02,331][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:39:02,832][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:39:03,334][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:39:03,838][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:39:04,346][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:39:04,853][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:39:05,354][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:39:05,855][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:39:06,358][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:39:06,859][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:39:07,360][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:39:07,863][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:39:08,365][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:39:08,872][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:39:09,375][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:39:09,881][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:39:10,384][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:39:10,887][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:39:11,389][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:39:11,900][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:39:12,406][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:39:12,934][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:39:13,438][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:39:13,941][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:39:14,443][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:39:14,946][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:39:15,454][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:39:15,955][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:39:16,459][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:39:16,964][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:39:17,499][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:39:18,012][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:39:18,518][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:39:19,023][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:39:19,522][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:39:20,020][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:39:20,521][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:39:21,020][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:39:21,518][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:39:22,016][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:39:22,515][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:39:23,013][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:39:23,519][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:39:24,018][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:39:24,526][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:39:25,026][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:39:25,526][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:39:26,034][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:39:26,532][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:39:27,029][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:39:27,528][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:39:28,026][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:39:28,531][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:39:29,038][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10111 tokens. -[2025-11-10 02:39:30,014][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 55.72%, Block Peak % of device VRAM: 62.16%, ΔTime: 00:00:33 -[2025-11-10 02:39:30,750][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:39:30,751][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:39:30,753][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:39:31,753][__main__][INFO] - Iteration 537 took 1m 8s (46.29% Gen, 52.24% Train). Generation: 31s, Training: 35s. Estimated remaining time: 48h 4m 8s. Estimated total time: 56h 44m 43s. Time estimates for 10 more iterations: 11m 20s, 100 more iterations: 1h 53m 29s, 500 more iterations: 9h 27m 27s. -[2025-11-10 02:39:31,755][__main__][INFO] - Starting iteration 537. -[2025-11-10 02:39:32,213][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 53 and human policies 1. -[2025-11-10 02:39:32,213][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:39:46,724][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 02:39:57,689][__main__][INFO] - Number of regex retries in iteration 537: 1 -[2025-11-10 02:39:57,690][__main__][INFO] - agents played in iteration 537 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:39:59,825][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:39:59,870][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:39:59,893][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:39:59,915][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:39:59,916][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:39:59,916][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:40:00,694][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:40:01,153][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:40:01,653][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:40:02,151][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:40:02,648][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:40:03,149][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:40:03,645][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:40:04,150][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:40:04,646][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:40:05,144][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:40:05,648][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:40:06,151][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:40:06,654][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:40:07,156][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:40:07,658][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:40:08,159][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:40:08,663][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:40:09,166][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:40:09,667][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:40:10,170][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:40:10,672][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:40:11,175][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:40:11,677][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:40:12,180][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:40:12,684][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:40:13,194][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:40:13,695][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:40:14,200][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:40:14,709][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:40:15,214][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:40:15,725][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:40:16,226][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:40:16,727][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:40:17,239][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:40:17,741][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:40:18,243][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:40:18,747][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:40:19,249][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:40:19,754][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:40:20,255][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:40:20,758][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:40:21,262][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:40:21,761][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:40:22,261][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:40:22,761][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:40:23,260][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:40:23,759][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:40:24,260][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:40:24,759][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:40:25,260][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:40:25,763][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:40:26,264][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:40:26,761][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:40:27,257][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:40:27,756][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:40:28,254][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:40:28,752][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:40:29,249][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:40:29,746][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:40:30,248][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:40:30,743][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:40:31,241][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:40:31,740][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:40:32,236][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:40:32,737][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10109 tokens. -[2025-11-10 02:40:33,514][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 55.66%, Block Peak % of device VRAM: 62.28%, ΔTime: 00:00:32 -[2025-11-10 02:40:34,158][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:40:34,159][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:40:34,161][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:40:34,997][__main__][INFO] - Iteration 538 took 1m 2s (40.58% Gen, 58.09% Train). Generation: 25s, Training: 36s. Estimated remaining time: 43h 37m 35s. Estimated total time: 52h 19m 13s. Time estimates for 10 more iterations: 10m 27s, 100 more iterations: 1h 44m 38s, 500 more iterations: 8h 43m 12s. -[2025-11-10 02:40:34,999][__main__][INFO] - Starting iteration 538. -[2025-11-10 02:40:35,465][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 53 and human policies 1. -[2025-11-10 02:40:35,465][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:41:06,476][__main__][INFO] - Number of regex retries in iteration 538: 0 -[2025-11-10 02:41:06,477][__main__][INFO] - agents played in iteration 538 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:41:07,296][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:41:07,319][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:41:07,342][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:41:07,365][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:41:07,366][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:41:07,366][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:41:08,180][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:41:08,633][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:41:09,133][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:41:09,630][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:41:10,128][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:41:10,625][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:41:11,121][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:41:11,621][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:41:12,123][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:41:12,622][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:41:13,122][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:41:13,625][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:41:14,121][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:41:14,632][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:41:15,131][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:41:15,632][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:41:16,148][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:41:16,646][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:41:17,146][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:41:17,653][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:41:18,153][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:41:18,656][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:41:19,159][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:41:19,659][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:41:20,162][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:41:20,669][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:41:21,168][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:41:21,673][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:41:22,171][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:41:22,677][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:41:23,177][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:41:23,678][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:41:24,179][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:41:24,685][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:41:25,189][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:41:25,692][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:41:26,196][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:41:26,726][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:41:27,229][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:41:27,731][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:41:28,234][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:41:28,735][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:41:29,233][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:41:29,730][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:41:30,226][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:41:30,737][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:41:31,234][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:41:31,729][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:41:32,225][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:41:32,721][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:41:33,234][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:41:33,737][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:41:34,234][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:41:34,738][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:41:35,235][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:41:35,739][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:41:36,237][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:41:36,735][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:41:37,235][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:41:37,736][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:41:38,234][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:41:38,730][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:41:39,226][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:41:39,747][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:41:40,246][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10042 tokens. -[2025-11-10 02:41:41,077][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.09%, Current % of VRAM taken: 55.59%, Block Peak % of device VRAM: 62.22%, ΔTime: 00:00:32 -[2025-11-10 02:41:41,791][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:41:41,793][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:41:41,794][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:41:42,749][__main__][INFO] - Iteration 539 took 1m 7s (46.09% Gen, 52.49% Train). Generation: 31s, Training: 35s. Estimated remaining time: 47h 21m 28s. Estimated total time: 56h 4m 14s. Time estimates for 10 more iterations: 11m 12s, 100 more iterations: 1h 52m 8s, 500 more iterations: 9h 20m 42s. -[2025-11-10 02:41:42,751][__main__][INFO] - Starting iteration 539. -[2025-11-10 02:41:43,216][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 53 and human policies 1. -[2025-11-10 02:41:43,217][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:42:05,181][__main__][INFO] - Number of regex retries in iteration 539: 0 -[2025-11-10 02:42:05,182][__main__][INFO] - agents played in iteration 539 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:42:06,059][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:42:06,085][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:42:06,110][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:42:06,132][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:42:06,133][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:42:06,134][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:42:06,967][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:42:07,439][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:42:07,945][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:42:08,452][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:42:08,951][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:42:09,450][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:42:09,957][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:42:10,454][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:42:10,952][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:42:11,452][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:42:11,949][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:42:12,449][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:42:12,951][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:42:13,453][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:42:13,961][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:42:14,462][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:42:14,964][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:42:15,466][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:42:15,970][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:42:16,470][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:42:16,978][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:42:17,482][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:42:17,984][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:42:18,485][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:42:18,990][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:42:19,957][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:42:21,254][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:42:21,759][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:42:22,264][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:42:22,765][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:42:23,271][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:42:23,770][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:42:24,273][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:42:24,774][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:42:25,274][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:42:25,774][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:42:26,274][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:42:26,773][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:42:27,277][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:42:27,780][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:42:28,283][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:42:28,780][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:42:29,277][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:42:29,776][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:42:30,274][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:42:30,774][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:42:31,269][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:42:31,765][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:42:32,265][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:42:32,760][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:42:33,261][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:42:33,754][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:42:34,250][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:42:34,752][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:42:35,248][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:42:35,744][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:42:36,240][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:42:36,739][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:42:37,234][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:42:37,729][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:42:38,227][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:42:38,743][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:42:39,239][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:42:39,739][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:42:40,241][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10117 tokens. -[2025-11-10 02:42:41,129][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.05%, Current % of VRAM taken: 55.55%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:34 -[2025-11-10 02:42:41,784][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:42:41,786][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:42:41,788][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:42:42,691][__main__][INFO] - Iteration 540 took 59s (36.93% Gen, 61.55% Train). Generation: 21s, Training: 36s. Estimated remaining time: 40h 49m 59s. Estimated total time: 49h 33m 45s. Time estimates for 10 more iterations: 9m 54s, 100 more iterations: 1h 39m 7s, 500 more iterations: 8h 15m 37s. -[2025-11-10 02:42:42,693][__main__][INFO] - Starting iteration 540. -[2025-11-10 02:42:43,158][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 53 and human policies 1. -[2025-11-10 02:42:43,158][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:43:02,758][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 02:43:13,960][__main__][INFO] - Number of regex retries in iteration 540: 1 -[2025-11-10 02:43:13,961][__main__][INFO] - agents played in iteration 540 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:43:14,773][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:43:14,797][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:43:14,820][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:43:14,843][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:43:14,844][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:43:14,844][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:43:15,641][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:43:16,096][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:43:16,602][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:43:17,107][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:43:17,610][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:43:18,116][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:43:18,625][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:43:19,127][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:43:19,627][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:43:20,136][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:43:20,637][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:43:21,159][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:43:21,661][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:43:22,165][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:43:22,666][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:43:23,174][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:43:23,679][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:43:24,178][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:43:24,677][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:43:25,184][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:43:25,684][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:43:26,186][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:43:26,690][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:43:27,195][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:43:27,692][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:43:28,188][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:43:28,685][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:43:29,187][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:43:29,684][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:43:30,187][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:43:30,689][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:43:31,187][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:43:31,684][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:43:32,183][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:43:32,682][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:43:33,180][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:43:33,676][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:43:34,174][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:43:34,670][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:43:35,166][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:43:35,670][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:43:36,166][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:43:36,663][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:43:37,170][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:43:37,669][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:43:38,167][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:43:38,661][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:43:39,158][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:43:39,663][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:43:40,165][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:43:40,662][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:43:41,158][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:43:41,652][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:43:42,146][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:43:42,645][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:43:43,140][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:43:43,648][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:43:44,144][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:43:44,646][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:43:45,146][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:43:45,640][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:43:46,149][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:43:46,645][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:43:47,141][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:43:47,656][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10017 tokens. -[2025-11-10 02:43:48,413][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 55.66%, Block Peak % of device VRAM: 62.15%, ΔTime: 00:00:32 -[2025-11-10 02:43:49,180][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:43:49,182][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:43:49,184][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:43:51,046][__main__][INFO] - Iteration 541 took 1m 7s (45.37% Gen, 51.88% Train). Generation: 30s, Training: 35s. Estimated remaining time: 47h 49m 34s. Estimated total time: 56h 34m 28s. Time estimates for 10 more iterations: 11m 18s, 100 more iterations: 1h 53m 8s, 500 more iterations: 9h 25m 44s. -[2025-11-10 02:43:51,049][__main__][INFO] - Starting iteration 541. -[2025-11-10 02:43:51,535][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 54 and human policies 1. -[2025-11-10 02:43:51,535][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:44:02,587][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 02:44:12,410][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 02:44:13,251][__main__][INFO] - Number of regex retries in iteration 541: 2 -[2025-11-10 02:44:13,251][__main__][INFO] - agents played in iteration 541 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:44:14,054][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:44:14,076][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:44:14,099][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:44:14,121][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:44:14,122][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:44:14,123][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:44:14,990][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:44:15,453][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:44:15,960][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:44:16,470][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:44:16,970][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:44:17,471][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:44:17,975][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:44:18,477][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:44:18,984][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:44:19,483][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:44:19,984][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:44:20,490][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:44:20,990][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:44:21,491][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:44:21,993][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:44:22,495][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:44:22,998][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:44:23,498][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:44:24,006][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:44:24,509][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:44:25,016][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:44:25,517][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:44:26,019][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:44:26,521][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:44:27,030][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:44:27,529][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:44:28,029][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:44:28,540][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:44:29,040][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:44:29,547][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:44:30,046][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:44:30,544][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:44:31,060][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:44:31,559][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:44:32,058][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:44:32,560][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:44:33,061][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:44:33,581][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:44:34,079][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:44:34,582][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:44:35,077][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:44:35,576][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:44:36,079][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:44:36,579][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:44:37,077][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:44:37,589][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:44:38,086][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:44:38,584][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:44:39,080][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:44:39,579][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:44:40,090][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:44:40,587][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:44:41,088][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:44:41,585][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:44:42,087][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:44:42,595][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:44:43,098][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:44:43,601][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:44:44,099][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:44:44,598][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:44:45,095][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:44:45,590][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:44:46,087][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:44:46,589][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:44:47,085][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10095 tokens. -[2025-11-10 02:44:47,871][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 55.56%, Block Peak % of device VRAM: 62.21%, ΔTime: 00:00:32 -[2025-11-10 02:44:48,592][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:44:48,594][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:44:48,595][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:44:49,532][__main__][INFO] - Iteration 542 took 57s (37.44% Gen, 60.94% Train). Generation: 21s, Training: 35s. Estimated remaining time: 39h 34m 1s. Estimated total time: 48h 19m 53s. Time estimates for 10 more iterations: 9m 39s, 100 more iterations: 1h 36m 39s, 500 more iterations: 8h 3m 18s. -[2025-11-10 02:44:49,534][__main__][INFO] - Starting iteration 542. -[2025-11-10 02:44:49,997][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 54 and human policies 1. -[2025-11-10 02:44:49,998][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:45:12,724][mllm.models.large_language_model_local][WARNING] - Response Proposal: x hats, y books, 10 balls +This proposal is based on the per-item values given. Since Alice values hats much more than I do, and I value books much more than Alice does, I aim to maximize my points by taking all the hats and leaving the books and balls to Alice. This way, I ensure I get the full value from the hats, which is 1 per item, while Alice gets the full value from the books and balls, which is 10 and 1 per item respectively. did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 00:56:00,724][__main__][INFO] - Number of regex retries in iteration 261: 1 +[2026-03-26 00:56:00,725][__main__][INFO] - agents played in iteration 261 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:56:01,500][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:56:01,523][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:56:01,546][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:56:01,569][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:56:01,570][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:56:01,570][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:56:02,229][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:56:02,683][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:56:03,189][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:56:03,692][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:56:04,193][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:56:04,694][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:56:05,194][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:56:05,695][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:56:06,194][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:56:06,695][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:56:07,195][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:56:07,696][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:56:08,195][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:56:08,694][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:56:09,195][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:56:09,695][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:56:10,195][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:56:10,692][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:56:11,191][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:56:11,691][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:56:12,190][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:56:12,689][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:56:13,186][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:56:13,687][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:56:14,186][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:56:14,694][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:56:15,195][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:56:15,698][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:56:16,197][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:56:16,715][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:56:17,222][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:56:17,722][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:56:18,221][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:56:18,720][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:56:19,220][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:56:19,719][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:56:20,218][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:56:20,718][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:56:21,217][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:56:21,716][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:56:22,213][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:56:22,713][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:56:23,212][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:56:23,710][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:56:24,210][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:56:24,711][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:56:25,211][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:56:25,716][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:56:26,215][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:56:26,711][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:56:27,208][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:56:27,708][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:56:28,203][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:56:28,699][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:56:29,199][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:56:29,699][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:56:30,195][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:56:30,692][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:56:31,189][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:56:31,684][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:56:32,183][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:56:32,678][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:56:33,174][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:56:33,672][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:56:34,169][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10431 tokens. +[2026-03-26 00:56:34,856][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 60.65%, Block Peak % of device VRAM: 62.56%, ΔTime: 00:00:32 +[2026-03-26 00:56:35,678][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:56:35,680][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:56:35,682][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:56:36,435][__main__][INFO] - Iteration 262 took 54s (34.34% Gen, 64.27% Train). Generation: 18s, Training: 34s. Estimated remaining time: 41h 28m 35s. Estimated total time: 45h 19m 31s. Time estimates for 10 more iterations: 9m 3s, 100 more iterations: 1h 30m 39s, 500 more iterations: 7h 33m 15s. +[2026-03-26 00:56:36,438][__main__][INFO] - Starting iteration 262. +[2026-03-26 00:56:36,837][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 26 and human policies 1. +[2026-03-26 00:56:36,837][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:56:41,929][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 00:56:49,499][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 00:56:56,037][__main__][INFO] - Number of regex retries in iteration 262: 2 +[2026-03-26 00:56:56,038][__main__][INFO] - agents played in iteration 262 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:56:56,813][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.62%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:56:56,836][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.62%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:56:56,859][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.62%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:56:56,882][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.62%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:56:56,883][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:56:56,883][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:56:57,523][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:56:57,983][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:56:58,488][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:56:58,987][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:56:59,485][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:56:59,986][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:57:00,484][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:57:00,985][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:57:01,481][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:57:01,983][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:57:02,479][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:57:02,976][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:57:03,474][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:57:03,971][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:57:04,471][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:57:04,970][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:57:05,466][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:57:05,962][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:57:06,460][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:57:06,957][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:57:07,456][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:57:07,954][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:57:08,451][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:57:08,948][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:57:09,446][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:57:09,948][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:57:10,447][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:57:10,944][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:57:11,441][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:57:11,938][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:57:12,435][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:57:12,935][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:57:13,434][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:57:13,932][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:57:14,429][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:57:14,925][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:57:15,424][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:57:15,919][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:57:16,421][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:57:16,916][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:57:17,413][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:57:17,913][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:57:18,412][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:57:18,910][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:57:19,423][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:57:19,926][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:57:20,426][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:57:20,927][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:57:21,427][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:57:21,922][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:57:22,419][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:57:22,915][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:57:23,411][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:57:23,906][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:57:24,402][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:57:24,901][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:57:25,398][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:57:25,895][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:57:26,392][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:57:26,893][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:57:27,389][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:57:27,889][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:57:28,385][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:57:28,883][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:57:29,387][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10355 tokens. +[2026-03-26 00:57:30,076][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.17%, ΔTime: 00:00:32 +[2026-03-26 00:57:30,832][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:57:30,835][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:57:30,836][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:57:31,627][__main__][INFO] - Iteration 263 took 54s (35.04% Gen, 63.51% Train). Generation: 19s, Training: 34s. Estimated remaining time: 41h 47m 42s. Estimated total time: 45h 39m 33s. Time estimates for 10 more iterations: 9m 7s, 100 more iterations: 1h 31m 19s, 500 more iterations: 7h 36m 35s. +[2026-03-26 00:57:31,630][__main__][INFO] - Starting iteration 263. +[2026-03-26 00:57:32,028][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 26 and human policies 1. +[2026-03-26 00:57:32,029][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:57:51,291][__main__][INFO] - Number of regex retries in iteration 263: 0 +[2026-03-26 00:57:51,292][__main__][INFO] - agents played in iteration 263 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:57:52,055][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:57:52,079][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:57:52,102][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:57:52,125][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:57:52,126][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:57:52,126][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:57:52,793][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:57:53,250][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:57:53,749][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:57:54,251][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:57:54,752][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:57:55,252][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:57:55,748][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:57:56,244][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:57:56,742][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:57:57,240][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:57:57,744][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:57:58,243][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:57:58,742][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:57:59,242][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:57:59,740][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:58:00,238][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:58:00,736][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:58:01,234][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:58:01,735][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:58:02,232][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:58:02,731][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:58:03,230][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:58:03,729][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:58:04,226][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:58:04,725][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:58:05,220][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:58:05,716][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:58:06,214][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:58:06,711][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:58:07,207][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:58:07,701][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:58:08,197][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:58:08,694][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:58:09,194][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:58:09,698][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:58:10,198][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:58:10,697][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:58:11,196][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:58:11,693][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:58:12,196][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:58:12,694][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:58:13,190][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:58:13,686][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:58:14,182][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:58:14,677][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:58:15,175][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:58:15,671][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:58:16,166][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:58:16,661][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:58:17,157][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:58:17,656][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:58:18,152][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:58:18,647][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:58:19,146][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:58:19,642][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:58:20,139][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:58:20,637][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:58:21,136][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:58:21,635][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:58:22,134][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:58:22,634][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:58:23,132][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:58:23,633][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:58:24,131][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:58:24,631][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10416 tokens. +[2026-03-26 00:58:25,348][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 60.66%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 +[2026-03-26 00:58:26,095][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:58:26,097][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:58:26,099][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:58:26,828][__main__][INFO] - Iteration 264 took 54s (35.15% Gen, 63.52% Train). Generation: 19s, Training: 34s. Estimated remaining time: 41h 47m 13s. Estimated total time: 45h 40m 0s. Time estimates for 10 more iterations: 9m 8s, 100 more iterations: 1h 31m 20s, 500 more iterations: 7h 36m 40s. +[2026-03-26 00:58:26,830][__main__][INFO] - Starting iteration 264. +[2026-03-26 00:58:27,232][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 26 and human policies 1. +[2026-03-26 00:58:27,232][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:58:37,067][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 00:58:37,331][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls -Given the values, I prioritize the balls heavily since they provide the highest value for me. Bob seems to value books the most and hats and balls equally, so I will try to claim as many balls as possible while leaving a small portion to Bob to avoid any proportional distribution issues. I will keep 10 hats and 0 books, as they have low value for me, and propose to take all 10 balls. did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 02:45:15,677][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 02:45:19,383][__main__][INFO] - Number of regex retries in iteration 542: 2 -[2025-11-10 02:45:19,384][__main__][INFO] - agents played in iteration 542 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:45:20,215][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:45:20,253][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:45:20,288][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:45:20,315][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:45:20,315][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:45:20,316][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:45:21,191][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:45:21,655][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:45:22,180][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:45:22,684][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:45:23,187][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:45:23,697][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:45:24,203][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:45:24,713][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:45:25,222][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:45:25,726][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:45:26,234][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:45:26,735][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:45:27,238][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:45:27,742][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:45:28,251][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:45:28,758][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:45:29,261][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:45:29,786][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:45:30,287][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:45:30,794][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:45:31,309][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:45:31,817][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:45:32,320][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:45:32,827][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:45:33,329][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:45:33,831][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:45:34,333][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:45:34,839][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:45:35,347][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:45:35,851][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:45:36,353][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:45:36,855][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:45:37,360][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:45:37,878][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:45:38,384][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:45:38,886][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:45:39,384][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:45:39,883][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:45:40,391][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:45:40,889][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:45:41,389][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:45:41,899][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:45:42,403][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:45:42,905][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:45:43,402][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:45:43,904][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:45:44,414][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:45:44,918][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:45:45,419][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:45:45,917][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:45:46,416][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:45:46,914][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:45:47,414][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:45:47,910][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:45:48,410][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:45:48,908][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:45:49,408][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:45:49,907][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:45:50,402][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:45:50,900][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:45:51,399][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:45:51,901][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:45:52,410][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:45:52,908][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:45:53,404][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10050 tokens. -[2025-11-10 02:45:54,158][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 55.59%, Block Peak % of device VRAM: 62.42%, ΔTime: 00:00:32 -[2025-11-10 02:45:54,809][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:45:54,812][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:45:54,814][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:45:55,762][__main__][INFO] - Iteration 543 took 1m 5s (44.68% Gen, 53.87% Train). Generation: 29s, Training: 35s. Estimated remaining time: 46h 1m 17s. Estimated total time: 54h 48m 16s. Time estimates for 10 more iterations: 10m 57s, 100 more iterations: 1h 49m 36s, 500 more iterations: 9h 8m 2s. -[2025-11-10 02:45:55,764][__main__][INFO] - Starting iteration 543. -[2025-11-10 02:45:56,236][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 54 and human policies 1. -[2025-11-10 02:45:56,236][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:46:17,869][__main__][INFO] - Number of regex retries in iteration 543: 0 -[2025-11-10 02:46:17,870][__main__][INFO] - agents played in iteration 543 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:46:18,762][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:46:18,787][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:46:18,811][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:46:18,834][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:46:18,835][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:46:18,835][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:46:19,624][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:46:20,109][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:46:20,617][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:46:21,121][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:46:21,631][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:46:22,136][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:46:22,638][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:46:23,141][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:46:23,644][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:46:24,149][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:46:24,653][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:46:25,160][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:46:25,662][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:46:26,167][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:46:26,685][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:46:27,191][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:46:27,704][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:46:28,208][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:46:28,711][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:46:29,215][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:46:29,719][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:46:30,221][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:46:30,723][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:46:31,224][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:46:31,737][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:46:32,240][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:46:32,741][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:46:33,242][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:46:33,745][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:46:34,248][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:46:34,746][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:46:35,248][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:46:35,751][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:46:36,252][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:46:36,753][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:46:37,255][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:46:37,757][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:46:38,259][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:46:38,762][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:46:39,261][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:46:39,768][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:46:40,271][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:46:40,784][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:46:41,286][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:46:41,783][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:46:42,283][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:46:42,785][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:46:43,300][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:46:43,804][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:46:44,301][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:46:44,801][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:46:45,299][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:46:45,799][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:46:46,297][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:46:46,794][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:46:47,296][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:46:47,792][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:46:48,289][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:46:48,799][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:46:49,304][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:46:49,804][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:46:50,302][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:46:50,802][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:46:51,305][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:46:51,804][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10104 tokens. -[2025-11-10 02:46:52,609][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.04%, Current % of VRAM taken: 55.53%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:32 -[2025-11-10 02:46:53,343][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:46:53,345][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:46:53,347][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:46:54,325][__main__][INFO] - Iteration 544 took 58s (37.24% Gen, 61.07% Train). Generation: 21s, Training: 35s. Estimated remaining time: 39h 36m 31s. Estimated total time: 48h 24m 29s. Time estimates for 10 more iterations: 9m 40s, 100 more iterations: 1h 36m 48s, 500 more iterations: 8h 4m 4s. -[2025-11-10 02:46:54,327][__main__][INFO] - Starting iteration 544. -[2025-11-10 02:46:54,806][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 54 and human policies 1. -[2025-11-10 02:46:54,807][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:47:17,141][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 02:47:22,968][__main__][INFO] - Number of regex retries in iteration 544: 1 -[2025-11-10 02:47:22,969][__main__][INFO] - agents played in iteration 544 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:47:23,804][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:47:23,840][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:47:23,868][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:47:23,894][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:47:23,895][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:47:23,896][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:47:24,769][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:47:25,239][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:47:25,755][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:47:26,259][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:47:26,772][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:47:27,274][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:47:27,788][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:47:28,296][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:47:28,802][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:47:29,322][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:47:29,827][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:47:30,330][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:47:30,841][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:47:31,348][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:47:31,864][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:47:32,368][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:47:32,877][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:47:33,386][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:47:33,889][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:47:34,399][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:47:34,906][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:47:35,413][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:47:35,936][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:47:36,439][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:47:36,946][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:47:37,451][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:47:37,957][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:47:38,464][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:47:38,967][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:47:39,472][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:47:39,980][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:47:40,488][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:47:40,992][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:47:41,492][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:47:41,996][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:47:42,520][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:47:43,023][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:47:43,525][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:47:44,029][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:47:44,533][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:47:45,038][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:47:45,541][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:47:46,045][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:47:46,545][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:47:47,045][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:47:47,547][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:47:48,048][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:47:48,549][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:47:49,053][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:47:49,551][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:47:50,057][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:47:50,560][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:47:51,058][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:47:51,559][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:47:52,062][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:47:52,564][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:47:53,064][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:47:53,562][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:47:54,059][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:47:54,558][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:47:55,060][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:47:55,563][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:47:56,066][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:47:56,566][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:47:57,073][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10136 tokens. -[2025-11-10 02:47:57,865][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 55.66%, Block Peak % of device VRAM: 62.20%, ΔTime: 00:00:33 -[2025-11-10 02:47:58,508][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:47:58,510][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:47:58,512][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:47:59,416][__main__][INFO] - Iteration 545 took 1m 4s (43.58% Gen, 55.01% Train). Generation: 28s, Training: 35s. Estimated remaining time: 45h 1m 28s. Estimated total time: 53h 50m 30s. Time estimates for 10 more iterations: 10m 46s, 100 more iterations: 1h 47m 41s, 500 more iterations: 8h 58m 25s. -[2025-11-10 02:47:59,418][__main__][INFO] - Starting iteration 545. -[2025-11-10 02:47:59,883][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 54 and human policies 1. -[2025-11-10 02:47:59,883][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:48:22,720][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls +Given the per-item values, I aim to maximize my value by focusing on the items where my value is significantly higher than Bob's. The hats have a value of 10 for me and only 1 for Bob, making them the most valuable for me. Since I have a much lower valuation for books and balls compared to Bob, it would be disadvantageous to propose any significant amount for those items. did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 00:58:45,755][__main__][INFO] - Number of regex retries in iteration 264: 2 +[2026-03-26 00:58:45,757][__main__][INFO] - agents played in iteration 264 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:58:46,516][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:58:46,540][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:58:46,563][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:58:46,587][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:58:46,588][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:58:46,588][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:58:47,245][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:58:47,700][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:58:48,204][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:58:48,703][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:58:49,200][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:58:49,697][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:58:50,194][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:58:50,691][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:58:51,186][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:58:51,688][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:58:52,191][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:58:52,691][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:58:53,194][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:58:53,694][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:58:54,193][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:58:54,693][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:58:55,193][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:58:55,691][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:58:56,194][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:58:56,694][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:58:57,195][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:58:57,692][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:58:58,191][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:58:58,686][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:58:59,183][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:58:59,679][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:59:00,179][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:59:00,676][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:59:01,179][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:59:01,679][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:59:02,180][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:59:02,680][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:59:03,179][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:59:03,675][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:59:04,174][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:59:04,669][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:59:05,167][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:59:05,664][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 00:59:06,160][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 00:59:06,655][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 00:59:07,152][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 00:59:07,651][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 00:59:08,153][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 00:59:08,650][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 00:59:09,148][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 00:59:09,645][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 00:59:10,144][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 00:59:10,640][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 00:59:11,138][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 00:59:11,637][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 00:59:12,152][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 00:59:12,656][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 00:59:13,154][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 00:59:13,654][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 00:59:14,154][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 00:59:14,655][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 00:59:15,156][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 00:59:15,653][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 00:59:16,150][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 00:59:16,648][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 00:59:17,144][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 00:59:17,640][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 00:59:18,137][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 00:59:18,633][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 00:59:19,130][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10371 tokens. +[2026-03-26 00:59:19,819][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.58%, Block Peak % of device VRAM: 62.37%, ΔTime: 00:00:32 +[2026-03-26 00:59:20,570][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 00:59:20,572][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 00:59:20,573][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 00:59:21,310][__main__][INFO] - Iteration 265 took 54s (34.25% Gen, 64.38% Train). Generation: 18s, Training: 34s. Estimated remaining time: 41h 10m 16s. Estimated total time: 45h 3m 57s. Time estimates for 10 more iterations: 9m 0s, 100 more iterations: 1h 30m 7s, 500 more iterations: 7h 30m 39s. +[2026-03-26 00:59:21,312][__main__][INFO] - Starting iteration 265. +[2026-03-26 00:59:21,711][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 26 and human policies 1. +[2026-03-26 00:59:21,712][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 00:59:39,626][__main__][INFO] - Number of regex retries in iteration 265: 0 +[2026-03-26 00:59:39,627][__main__][INFO] - agents played in iteration 265 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 00:59:40,391][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:59:40,415][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:59:40,438][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:59:40,461][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 00:59:40,462][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 00:59:40,462][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 00:59:41,111][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 00:59:41,570][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 00:59:42,073][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 00:59:42,570][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 00:59:43,073][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 00:59:43,575][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 00:59:44,077][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 00:59:44,578][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 00:59:45,077][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 00:59:45,576][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 00:59:46,076][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 00:59:46,575][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 00:59:47,075][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 00:59:47,575][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 00:59:48,074][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 00:59:48,571][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 00:59:49,071][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 00:59:49,571][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 00:59:50,072][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 00:59:50,571][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 00:59:51,071][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 00:59:51,569][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 00:59:52,066][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 00:59:52,568][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 00:59:53,071][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 00:59:53,570][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 00:59:54,069][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 00:59:54,567][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 00:59:55,066][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 00:59:55,562][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 00:59:56,060][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 00:59:56,561][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 00:59:57,058][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 00:59:57,562][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 00:59:58,059][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 00:59:58,560][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 00:59:59,059][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 00:59:59,560][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:00:00,059][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:00:00,557][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:00:01,053][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:00:01,553][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:00:02,054][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:00:02,553][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:00:03,066][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:00:03,570][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:00:04,071][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:00:04,571][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:00:05,068][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:00:05,574][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:00:06,071][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:00:06,574][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:00:07,073][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:00:07,572][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:00:08,072][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:00:08,573][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:00:09,073][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:00:09,569][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:00:10,069][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:00:10,568][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:00:11,069][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:00:11,569][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:00:12,068][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:00:12,567][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:00:13,066][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10528 tokens. +[2026-03-26 01:00:13,744][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 60.60%, Block Peak % of device VRAM: 62.27%, ΔTime: 00:00:32 +[2026-03-26 01:00:14,496][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:00:14,498][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:00:14,499][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:00:15,226][__main__][INFO] - Iteration 266 took 53s (33.48% Gen, 65.16% Train). Generation: 17s, Training: 34s. Estimated remaining time: 40h 41m 10s. Estimated total time: 44h 35m 45s. Time estimates for 10 more iterations: 8m 55s, 100 more iterations: 1h 29m 11s, 500 more iterations: 7h 25m 57s. +[2026-03-26 01:00:15,228][__main__][INFO] - Starting iteration 266. +[2026-03-26 01:00:15,626][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 26 and human policies 1. +[2026-03-26 01:00:15,626][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:00:26,373][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 01:00:34,075][__main__][INFO] - Number of regex retries in iteration 266: 1 +[2026-03-26 01:00:34,076][__main__][INFO] - agents played in iteration 266 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:00:34,842][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:00:34,866][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:00:34,889][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:00:34,912][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:00:34,913][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:00:34,913][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:00:35,564][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:00:36,015][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:00:36,523][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:00:37,024][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:00:37,521][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:00:38,017][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:00:38,514][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:00:39,015][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:00:39,516][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:00:40,019][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:00:40,517][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:00:41,014][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:00:41,510][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:00:42,006][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:00:42,502][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:00:43,000][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:00:43,500][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:00:44,001][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:00:44,500][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:00:44,999][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:00:45,496][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:00:45,996][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:00:46,491][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:00:46,988][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:00:47,490][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:00:47,989][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:00:48,485][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:00:48,983][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:00:49,484][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:00:49,982][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:00:50,482][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:00:50,977][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:00:51,478][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:00:51,976][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:00:52,477][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:00:52,978][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:00:53,477][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:00:53,981][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:00:54,479][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:00:55,010][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:00:55,510][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:00:56,005][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:00:56,500][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:00:56,997][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:00:57,492][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:00:57,990][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:00:58,487][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:00:58,983][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:00:59,482][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:00:59,979][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:01:00,478][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:01:00,975][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:01:01,473][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:01:01,970][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:01:02,469][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:01:02,964][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:01:03,459][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:01:03,961][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:01:04,459][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:01:04,957][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:01:05,456][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:01:05,955][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:01:06,455][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:01:06,956][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:01:07,454][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10377 tokens. +[2026-03-26 01:01:08,130][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 60.62%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:32 +[2026-03-26 01:01:08,906][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:01:08,908][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:01:08,910][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:01:09,699][__main__][INFO] - Iteration 267 took 54s (34.12% Gen, 64.42% Train). Generation: 18s, Training: 34s. Estimated remaining time: 41h 8m 12s. Estimated total time: 45h 3m 42s. Time estimates for 10 more iterations: 9m 0s, 100 more iterations: 1h 30m 7s, 500 more iterations: 7h 30m 37s. +[2026-03-26 01:01:09,702][__main__][INFO] - Starting iteration 267. +[2026-03-26 01:01:10,100][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 26 and human policies 1. +[2026-03-26 01:01:10,100][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:01:14,930][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 01:01:15,014][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 01:01:19,927][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 01:01:20,970][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls + +Given the values, I propose taking all the hats because my value for hats is significantly higher (10) compared to my value for books and balls (1). Bob values books more highly than I do, so I avoid proposing any books. Balls have equal value for both of us, but since my overall value for books is higher than his for balls, I choose to take all the hats to maximize my score in this round. did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 01:01:29,108][__main__][INFO] - Number of regex retries in iteration 267: 4 +[2026-03-26 01:01:29,109][__main__][INFO] - agents played in iteration 267 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:01:29,867][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:01:29,891][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:01:29,914][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:01:29,937][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:01:29,938][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:01:29,938][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:01:30,583][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:01:31,039][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:01:31,542][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:01:32,043][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:01:32,541][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:01:33,039][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:01:33,539][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:01:34,040][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:01:34,539][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:01:35,042][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:01:35,540][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:01:36,035][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:01:36,536][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:01:37,036][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:01:37,535][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:01:38,034][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:01:38,530][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:01:39,031][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:01:39,528][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:01:40,023][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:01:40,524][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:01:41,021][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:01:41,522][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:01:42,021][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:01:42,520][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:01:43,019][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:01:43,518][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:01:44,014][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:01:44,514][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:01:45,013][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:01:45,512][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:01:46,009][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:01:46,509][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:01:47,010][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:01:47,506][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:01:48,005][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:01:48,504][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:01:49,003][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:01:49,501][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:01:49,999][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:01:50,499][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:01:50,998][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:01:51,497][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:01:51,997][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:01:52,497][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:01:52,995][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:01:53,495][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:01:53,996][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:01:54,497][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:01:54,997][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:01:55,496][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:01:55,994][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:01:56,492][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:01:56,990][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:01:57,488][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:01:57,984][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:01:58,484][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:01:58,983][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:01:59,481][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:01:59,980][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:02:00,479][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:02:00,978][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:02:01,477][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:02:01,979][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:02:02,479][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10536 tokens. +[2026-03-26 01:02:03,156][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 60.64%, Block Peak % of device VRAM: 62.28%, ΔTime: 00:00:32 +[2026-03-26 01:02:09,440][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:02:09,442][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:02:09,443][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:02:10,237][__main__][INFO] - Iteration 268 took 1m 0s (31.61% Gen, 67.07% Train). Generation: 19s, Training: 40s. Estimated remaining time: 46h 10m 22s. Estimated total time: 50h 6m 52s. Time estimates for 10 more iterations: 10m 1s, 100 more iterations: 1h 40m 13s, 500 more iterations: 8h 21m 8s. +[2026-03-26 01:02:10,239][__main__][INFO] - Starting iteration 268. +[2026-03-26 01:02:10,637][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 26 and human policies 1. +[2026-03-26 01:02:10,638][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:02:50,628][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 01:02:57,290][__main__][INFO] - Number of regex retries in iteration 268: 1 +[2026-03-26 01:02:57,290][__main__][INFO] - agents played in iteration 268 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:02:58,052][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:02:58,075][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:02:58,098][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:02:58,122][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:02:58,122][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:02:58,123][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:02:58,778][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:02:59,235][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:02:59,739][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:03:00,239][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:03:00,738][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:03:01,237][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:03:01,736][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:03:02,237][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:03:02,737][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:03:03,238][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:03:03,736][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:03:04,235][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:03:04,734][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:03:05,229][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:03:05,726][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:03:06,231][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:03:06,732][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:03:07,228][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:03:07,724][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:03:08,219][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:03:08,719][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:03:09,215][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:03:09,713][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:03:10,208][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:03:10,705][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:03:11,207][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:03:11,707][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:03:12,208][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:03:12,705][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:03:13,202][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:03:13,703][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:03:14,200][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:03:14,703][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:03:15,199][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:03:15,696][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:03:16,197][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:03:16,694][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:03:17,191][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:03:17,688][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:03:18,184][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:03:18,681][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:03:19,180][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:03:19,677][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:03:20,177][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:03:20,677][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:03:21,174][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:03:21,672][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:03:22,169][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:03:22,669][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:03:23,169][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:03:23,670][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:03:24,171][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:03:24,667][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:03:25,168][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:03:25,665][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:03:26,162][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:03:26,659][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:03:27,157][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:03:27,654][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:03:28,154][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:03:28,652][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:03:29,151][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:03:29,648][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:03:30,143][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:03:30,642][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10326 tokens. +[2026-03-26 01:03:31,327][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 60.64%, Block Peak % of device VRAM: 62.28%, ΔTime: 00:00:32 +[2026-03-26 01:03:32,092][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:03:32,095][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:03:32,097][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:03:32,879][__main__][INFO] - Iteration 269 took 1m 22s (56.73% Gen, 42.32% Train). Generation: 46s, Training: 34s. Estimated remaining time: 64h 34m 12s. Estimated total time: 68h 32m 5s. Time estimates for 10 more iterations: 13m 42s, 100 more iterations: 2h 17m 4s, 500 more iterations: 11h 25m 20s. +[2026-03-26 01:03:32,881][__main__][INFO] - Starting iteration 269. +[2026-03-26 01:03:33,279][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 26 and human policies 1. +[2026-03-26 01:03:33,280][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:03:56,333][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 01:03:57,120][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 01:04:09,294][__main__][INFO] - Number of regex retries in iteration 269: 2 +[2026-03-26 01:04:09,295][__main__][INFO] - agents played in iteration 269 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:04:10,068][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:04:10,091][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:04:10,114][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:04:10,137][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:04:10,138][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:04:10,138][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:04:10,797][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:04:11,251][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:04:11,753][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:04:12,250][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:04:12,751][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:04:13,246][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:04:13,741][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:04:14,242][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:04:14,745][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:04:15,240][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:04:15,736][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:04:16,236][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:04:16,733][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:04:17,232][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:04:17,733][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:04:18,236][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:04:18,732][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:04:19,229][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:04:19,730][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:04:20,228][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:04:20,724][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:04:21,223][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:04:21,722][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:04:22,220][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:04:22,718][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:04:23,214][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:04:23,713][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:04:24,212][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:04:24,711][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:04:25,210][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:04:25,710][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:04:26,208][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:04:26,703][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:04:27,200][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:04:27,699][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:04:28,196][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:04:28,724][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:04:29,223][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:04:29,718][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:04:30,215][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:04:30,714][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:04:31,211][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:04:31,713][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:04:32,213][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:04:32,708][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:04:33,205][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:04:33,703][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:04:34,201][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:04:34,697][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:04:35,194][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:04:35,695][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:04:36,192][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:04:36,692][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:04:37,192][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:04:37,687][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:04:38,183][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:04:38,679][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:04:39,176][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:04:39,675][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:04:40,171][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:04:40,668][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:04:41,168][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:04:41,664][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:04:42,159][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:04:42,655][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10363 tokens. +[2026-03-26 01:04:43,333][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.09%, Current % of VRAM taken: 60.52%, Block Peak % of device VRAM: 62.11%, ΔTime: 00:00:32 +[2026-03-26 01:04:44,082][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:04:44,084][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:04:44,085][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:04:44,819][__main__][INFO] - Iteration 270 took 1m 11s (50.34% Gen, 48.63% Train). Generation: 36s, Training: 34s. Estimated remaining time: 55h 37m 55s. Estimated total time: 59h 37m 0s. Time estimates for 10 more iterations: 11m 55s, 100 more iterations: 1h 59m 14s, 500 more iterations: 9h 56m 10s. +[2026-03-26 01:04:44,821][__main__][INFO] - Starting iteration 270. +[2026-03-26 01:04:45,219][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 26 and human policies 1. +[2026-03-26 01:04:45,219][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:05:05,329][__main__][INFO] - Number of regex retries in iteration 270: 0 +[2026-03-26 01:05:05,330][__main__][INFO] - agents played in iteration 270 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:05:06,105][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:05:06,128][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:05:06,151][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:05:06,174][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:05:06,175][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:05:06,175][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:05:06,826][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:05:07,285][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:05:07,791][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:05:08,290][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:05:08,785][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:05:09,285][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:05:09,786][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:05:10,289][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:05:10,788][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:05:11,284][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:05:11,781][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:05:12,278][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:05:12,775][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:05:13,275][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:05:13,775][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:05:14,275][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:05:14,773][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:05:15,269][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:05:15,765][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:05:16,261][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:05:16,761][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:05:17,261][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:05:17,757][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:05:18,254][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:05:18,755][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:05:19,252][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:05:19,750][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:05:20,245][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:05:20,745][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:05:21,241][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:05:21,735][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:05:22,235][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:05:22,735][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:05:23,231][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:05:23,728][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:05:24,223][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:05:24,719][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:05:25,215][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:05:25,711][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:05:26,207][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:05:26,706][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:05:27,200][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:05:27,698][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:05:28,197][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:05:28,692][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:05:29,190][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:05:29,684][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:05:30,182][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:05:30,677][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:05:31,173][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:05:31,671][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:05:32,170][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:05:32,666][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:05:33,168][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:05:33,668][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:05:34,169][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:05:34,665][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:05:35,162][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:05:35,661][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:05:36,160][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:05:36,658][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:05:37,159][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:05:37,655][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:05:38,156][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:05:38,655][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10356 tokens. +[2026-03-26 01:05:39,334][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 60.64%, Block Peak % of device VRAM: 62.20%, ΔTime: 00:00:32 +[2026-03-26 01:05:40,086][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:05:40,088][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:05:40,090][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:05:41,394][__main__][INFO] - Iteration 271 took 56s (35.80% Gen, 61.88% Train). Generation: 20s, Training: 34s. Estimated remaining time: 42h 48m 46s. Estimated total time: 46h 48m 47s. Time estimates for 10 more iterations: 9m 21s, 100 more iterations: 1h 33m 37s, 500 more iterations: 7h 48m 7s. +[2026-03-26 01:05:41,396][__main__][INFO] - Starting iteration 271. +[2026-03-26 01:05:41,794][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 27 and human policies 1. +[2026-03-26 01:05:41,795][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:05:58,628][__main__][INFO] - Number of regex retries in iteration 271: 0 +[2026-03-26 01:05:58,629][__main__][INFO] - agents played in iteration 271 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:05:59,397][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:05:59,420][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:05:59,444][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:05:59,466][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:05:59,467][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:05:59,468][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:06:00,118][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:06:00,574][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:06:01,074][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:06:01,572][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:06:02,072][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:06:02,570][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:06:03,071][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:06:03,567][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:06:04,068][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:06:04,568][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:06:05,065][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:06:05,562][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:06:06,062][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:06:06,568][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:06:07,067][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:06:07,563][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:06:08,062][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:06:08,561][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:06:09,061][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:06:09,561][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:06:10,059][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:06:10,559][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:06:11,059][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:06:11,559][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:06:12,062][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:06:12,561][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:06:13,061][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:06:13,558][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:06:14,060][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:06:14,558][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:06:15,057][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:06:15,556][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:06:16,057][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:06:16,553][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:06:17,054][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:06:17,553][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:06:18,054][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:06:18,553][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:06:19,050][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:06:19,551][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:06:20,050][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:06:20,550][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:06:21,046][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:06:21,545][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:06:22,044][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:06:22,541][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:06:23,038][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:06:23,535][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:06:24,034][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:06:24,533][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:06:25,033][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:06:25,532][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:06:26,031][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:06:26,531][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:06:27,030][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:06:27,530][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:06:28,034][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:06:28,532][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:06:29,032][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:06:29,536][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:06:30,036][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:06:30,534][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:06:31,031][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:06:31,531][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:06:32,033][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10435 tokens. +[2026-03-26 01:06:32,720][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.58%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 +[2026-03-26 01:06:33,468][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:06:33,471][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:06:33,472][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:06:34,205][__main__][INFO] - Iteration 272 took 52s (32.12% Gen, 66.48% Train). Generation: 16s, Training: 34s. Estimated remaining time: 39h 39m 42s. Estimated total time: 43h 40m 35s. Time estimates for 10 more iterations: 8m 44s, 100 more iterations: 1h 27m 21s, 500 more iterations: 7h 16m 45s. +[2026-03-26 01:06:34,207][__main__][INFO] - Starting iteration 272. +[2026-03-26 01:06:34,608][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 27 and human policies 1. +[2026-03-26 01:06:34,608][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:06:52,654][__main__][INFO] - Number of regex retries in iteration 272: 0 +[2026-03-26 01:06:52,655][__main__][INFO] - agents played in iteration 272 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:06:53,428][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:06:53,451][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:06:53,474][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:06:53,497][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:06:53,498][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:06:53,498][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:06:54,152][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:06:54,605][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:06:55,107][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:06:55,607][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:06:56,111][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:06:56,610][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:06:57,115][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:06:57,612][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:06:58,109][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:06:58,612][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:06:59,111][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:06:59,611][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:07:00,107][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:07:00,608][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:07:01,108][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:07:01,607][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:07:02,108][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:07:02,606][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:07:03,127][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:07:03,629][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:07:04,129][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:07:04,627][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:07:05,125][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:07:05,623][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:07:06,124][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:07:06,619][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:07:07,114][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:07:07,611][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:07:08,108][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:07:08,609][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:07:09,105][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:07:09,600][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:07:10,096][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:07:10,593][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:07:11,092][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:07:11,594][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:07:12,092][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:07:12,589][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:07:13,085][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:07:13,581][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:07:14,080][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:07:14,577][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:07:15,076][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:07:15,582][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:07:16,082][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:07:16,582][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:07:17,080][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:07:17,575][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:07:18,072][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:07:18,570][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:07:19,067][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:07:19,563][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:07:20,061][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:07:20,557][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:07:21,056][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:07:21,551][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:07:22,046][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:07:22,541][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:07:23,036][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:07:23,531][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:07:24,025][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:07:24,520][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:07:25,024][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:07:25,519][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:07:26,018][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10335 tokens. +[2026-03-26 01:07:26,716][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 60.60%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:32 +[2026-03-26 01:07:27,472][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:07:27,475][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:07:27,476][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:07:28,326][__main__][INFO] - Iteration 273 took 53s (33.59% Gen, 64.82% Train). Generation: 18s, Training: 34s. Estimated remaining time: 40h 44m 7s. Estimated total time: 44h 45m 55s. Time estimates for 10 more iterations: 8m 57s, 100 more iterations: 1h 29m 31s, 500 more iterations: 7h 27m 39s. +[2026-03-26 01:07:28,328][__main__][INFO] - Starting iteration 273. +[2026-03-26 01:07:28,726][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 27 and human policies 1. +[2026-03-26 01:07:28,726][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:07:47,453][__main__][INFO] - Number of regex retries in iteration 273: 0 +[2026-03-26 01:07:47,454][__main__][INFO] - agents played in iteration 273 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:07:48,225][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:07:48,248][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:07:48,271][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:07:48,294][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:07:48,294][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:07:48,295][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:07:48,957][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:07:49,415][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:07:49,925][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:07:50,428][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:07:50,929][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:07:51,431][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:07:51,933][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:07:52,433][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:07:52,933][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:07:53,429][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:07:53,928][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:07:54,434][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:07:54,933][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:07:55,431][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:07:55,928][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:07:56,426][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:07:56,924][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:07:57,428][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:07:57,925][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:07:58,425][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:07:58,921][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:07:59,420][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:07:59,916][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:08:00,415][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:08:00,920][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:08:01,416][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:08:01,917][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:08:02,414][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:08:02,913][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:08:03,408][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:08:03,905][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:08:04,401][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:08:04,897][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:08:05,397][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:08:05,901][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:08:06,401][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:08:06,902][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:08:07,402][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:08:07,901][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:08:08,399][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:08:08,897][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:08:09,393][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:08:09,893][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:08:10,391][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:08:10,890][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:08:11,390][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:08:11,890][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:08:12,386][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:08:12,887][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:08:13,387][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:08:13,887][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:08:14,384][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:08:14,883][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:08:15,383][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:08:15,881][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:08:16,380][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:08:16,879][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:08:17,379][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:08:17,878][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:08:18,378][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:08:18,877][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:08:19,375][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:08:19,874][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:08:20,375][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:08:20,874][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10407 tokens. +[2026-03-26 01:08:21,566][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 60.62%, Block Peak % of device VRAM: 62.28%, ΔTime: 00:00:32 +[2026-03-26 01:08:22,321][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:08:22,323][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:08:22,325][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:08:23,119][__main__][INFO] - Iteration 274 took 54s (34.43% Gen, 64.11% Train). Generation: 18s, Training: 34s. Estimated remaining time: 41h 16m 59s. Estimated total time: 45h 19m 42s. Time estimates for 10 more iterations: 9m 3s, 100 more iterations: 1h 30m 39s, 500 more iterations: 7h 33m 17s. +[2026-03-26 01:08:23,122][__main__][INFO] - Starting iteration 274. +[2026-03-26 01:08:23,520][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 27 and human policies 1. +[2026-03-26 01:08:23,520][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:08:29,981][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 01:08:42,656][__main__][INFO] - Number of regex retries in iteration 274: 1 +[2026-03-26 01:08:42,656][__main__][INFO] - agents played in iteration 274 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:08:43,415][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:08:43,438][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:08:43,461][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:08:43,484][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:08:43,485][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:08:43,485][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:08:44,137][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:08:44,589][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:08:45,093][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:08:45,590][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:08:46,092][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:08:46,589][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:08:47,086][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:08:47,586][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:08:48,085][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:08:48,586][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:08:49,089][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:08:49,591][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:08:50,087][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:08:50,588][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:08:51,089][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:08:51,587][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:08:52,085][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:08:52,584][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:08:53,082][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:08:53,580][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:08:54,077][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:08:54,575][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:08:55,073][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:08:55,572][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:08:56,071][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:08:56,568][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:08:57,064][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:08:57,580][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:08:58,083][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:08:58,582][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:08:59,082][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:08:59,581][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:09:00,080][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:09:00,584][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:09:01,081][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:09:01,581][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:09:02,078][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:09:02,576][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:09:03,074][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:09:03,573][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:09:04,072][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:09:04,568][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:09:05,067][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:09:05,563][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:09:06,059][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:09:06,558][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:09:07,056][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:09:07,554][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:09:08,052][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:09:08,555][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:09:09,054][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:09:09,553][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:09:10,051][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:09:10,550][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:09:11,052][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:09:11,548][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:09:12,050][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:09:12,549][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:09:13,048][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:09:13,547][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:09:14,044][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:09:14,542][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:09:15,039][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:09:15,536][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:09:16,034][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10443 tokens. +[2026-03-26 01:09:16,715][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 60.64%, Block Peak % of device VRAM: 62.33%, ΔTime: 00:00:32 +[2026-03-26 01:09:17,466][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:09:17,468][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:09:17,470][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:09:18,359][__main__][INFO] - Iteration 275 took 54s (34.90% Gen, 63.48% Train). Generation: 19s, Training: 34s. Estimated remaining time: 41h 38m 20s. Estimated total time: 45h 41m 58s. Time estimates for 10 more iterations: 9m 8s, 100 more iterations: 1h 31m 23s, 500 more iterations: 7h 36m 59s. +[2026-03-26 01:09:18,361][__main__][INFO] - Starting iteration 275. +[2026-03-26 01:09:18,759][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 27 and human policies 1. +[2026-03-26 01:09:18,760][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:09:27,163][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls + +Given Alice's per-item values, she values hats much less than I do, so I will try to grab all the hats. Books and balls have the same value for both of us, so there's no strategic advantage in splitting those items. By taking all the hats, I maximize my potential points from this round. did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 01:09:28,819][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 01:09:36,595][__main__][INFO] - Number of regex retries in iteration 275: 2 +[2026-03-26 01:09:36,596][__main__][INFO] - agents played in iteration 275 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:09:37,354][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:09:37,378][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:09:37,401][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:09:37,423][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:09:37,424][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:09:37,424][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:09:38,085][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:09:38,543][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:09:39,053][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:09:39,555][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:09:40,053][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:09:40,556][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:09:41,061][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:09:41,560][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:09:42,058][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:09:42,557][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:09:43,056][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:09:43,557][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:09:44,064][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:09:44,564][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:09:45,064][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:09:45,563][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:09:46,061][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:09:46,560][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:09:47,059][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:09:47,565][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:09:48,061][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:09:48,558][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:09:49,054][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:09:49,554][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:09:50,054][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:09:50,554][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:09:51,053][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:09:51,549][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:09:52,047][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:09:52,550][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:09:53,051][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:09:53,551][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:09:54,049][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:09:54,545][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:09:55,044][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:09:55,540][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:09:56,037][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:09:56,533][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:09:57,029][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:09:57,527][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:09:58,024][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:09:58,525][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:09:59,020][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:09:59,518][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:10:00,013][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:10:00,509][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:10:01,005][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:10:01,500][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:10:01,998][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:10:02,497][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:10:02,994][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:10:03,493][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:10:03,991][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:10:04,490][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:10:04,986][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:10:05,485][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:10:05,984][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:10:06,483][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:10:06,984][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:10:07,481][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:10:07,978][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:10:08,476][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:10:08,976][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:10:09,476][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:10:09,972][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10357 tokens. +[2026-03-26 01:10:10,653][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 +[2026-03-26 01:10:11,398][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:10:11,402][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:10:11,404][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:10:12,278][__main__][INFO] - Iteration 276 took 53s (33.33% Gen, 65.04% Train). Generation: 17s, Training: 34s. Estimated remaining time: 40h 31m 25s. Estimated total time: 44h 35m 57s. Time estimates for 10 more iterations: 8m 55s, 100 more iterations: 1h 29m 11s, 500 more iterations: 7h 25m 59s. +[2026-03-26 01:10:12,281][__main__][INFO] - Starting iteration 276. +[2026-03-26 01:10:12,678][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 27 and human policies 1. +[2026-03-26 01:10:12,679][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:10:18,510][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 01:10:29,190][__main__][INFO] - Number of regex retries in iteration 276: 1 +[2026-03-26 01:10:29,191][__main__][INFO] - agents played in iteration 276 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:10:29,949][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:10:29,973][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:10:29,996][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:10:30,020][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:10:30,020][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:10:30,021][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:10:30,683][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:10:31,138][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:10:31,641][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:10:32,141][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:10:32,638][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:10:33,134][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:10:33,630][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:10:34,128][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:10:34,625][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:10:35,124][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:10:35,625][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:10:36,122][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:10:36,628][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:10:37,128][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:10:37,624][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:10:38,122][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:10:38,617][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:10:39,113][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:10:39,612][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:10:40,108][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:10:40,609][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:10:41,105][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:10:41,601][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:10:42,097][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:10:42,592][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:10:43,091][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:10:43,590][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:10:44,089][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:10:44,586][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:10:45,087][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:10:45,583][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:10:46,081][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:10:46,577][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:10:47,073][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:10:47,572][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:10:48,069][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:10:48,566][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:10:49,062][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:10:49,559][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:10:50,059][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:10:50,559][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:10:51,054][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:10:51,552][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:10:52,046][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:10:52,547][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:10:53,046][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:10:53,547][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:10:54,041][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:10:54,540][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:10:55,036][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:10:55,532][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:10:56,031][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:10:56,526][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:10:57,022][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:10:57,518][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:10:58,016][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:10:58,512][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:10:59,010][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:10:59,507][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:11:00,007][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:11:00,502][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:11:00,999][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:11:01,497][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:11:01,997][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:11:02,500][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10199 tokens. +[2026-03-26 01:11:03,188][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 60.61%, Block Peak % of device VRAM: 62.13%, ΔTime: 00:00:32 +[2026-03-26 01:11:03,949][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:11:03,951][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:11:03,953][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:11:04,838][__main__][INFO] - Iteration 277 took 52s (31.66% Gen, 66.64% Train). Generation: 16s, Training: 34s. Estimated remaining time: 39h 22m 35s. Estimated total time: 43h 28m 0s. Time estimates for 10 more iterations: 8m 41s, 100 more iterations: 1h 26m 56s, 500 more iterations: 7h 14m 40s. +[2026-03-26 01:11:04,842][__main__][INFO] - Starting iteration 277. +[2026-03-26 01:11:05,243][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 27 and human policies 1. +[2026-03-26 01:11:05,243][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:11:20,193][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 01:11:25,818][__main__][INFO] - Number of regex retries in iteration 277: 1 +[2026-03-26 01:11:25,819][__main__][INFO] - agents played in iteration 277 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:11:26,583][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:11:26,607][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:11:26,630][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:11:26,653][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:11:26,653][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:11:26,654][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:11:27,309][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:11:27,766][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:11:28,278][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:11:28,779][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:11:29,275][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:11:29,774][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:11:30,278][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:11:30,778][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:11:31,278][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:11:31,778][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:11:32,279][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:11:32,779][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:11:33,283][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:11:33,783][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:11:34,283][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:11:34,780][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:11:35,280][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:11:35,774][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:11:36,270][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:11:36,771][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:11:37,266][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:11:37,761][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:11:38,257][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:11:38,753][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:11:39,249][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:11:39,743][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:11:40,238][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:11:40,735][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:11:41,237][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:11:41,734][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:11:42,231][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:11:42,727][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:11:43,227][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:11:43,722][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:11:44,221][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:11:44,719][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:11:45,215][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:11:45,715][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:11:46,211][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:11:46,711][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:11:47,212][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:11:47,708][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:11:48,206][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:11:48,705][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:11:49,204][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:11:49,702][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:11:50,201][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:11:50,700][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:11:51,195][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:11:51,692][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:11:52,189][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:11:52,687][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:11:53,185][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:11:53,685][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:11:54,183][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:11:54,681][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:11:55,179][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:11:55,677][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:11:56,175][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:11:56,671][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:11:57,172][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:11:57,672][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:11:58,172][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:11:58,671][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:11:59,170][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10466 tokens. +[2026-03-26 01:11:59,852][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 60.65%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:32 +[2026-03-26 01:12:00,599][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:12:00,602][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:12:00,603][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:12:01,480][__main__][INFO] - Iteration 278 took 56s (36.59% Gen, 61.85% Train). Generation: 20s, Training: 34s. Estimated remaining time: 42h 45m 32s. Estimated total time: 46h 51m 53s. Time estimates for 10 more iterations: 9m 22s, 100 more iterations: 1h 33m 43s, 500 more iterations: 7h 48m 38s. +[2026-03-26 01:12:01,483][__main__][INFO] - Starting iteration 278. +[2026-03-26 01:12:01,884][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 27 and human policies 1. +[2026-03-26 01:12:01,884][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:12:20,493][__main__][INFO] - Number of regex retries in iteration 278: 0 +[2026-03-26 01:12:20,494][__main__][INFO] - agents played in iteration 278 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:12:21,257][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:12:21,280][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:12:21,304][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:12:21,326][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:12:21,327][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:12:21,327][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:12:21,989][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:12:22,444][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:12:22,948][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:12:23,449][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:12:23,949][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:12:24,447][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:12:24,947][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:12:25,451][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:12:25,951][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:12:26,451][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:12:26,955][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:12:27,453][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:12:27,950][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:12:28,446][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:12:28,944][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:12:29,441][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:12:29,939][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:12:30,435][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:12:30,931][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:12:31,427][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:12:31,924][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:12:32,419][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:12:32,913][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:12:33,407][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:12:33,903][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:12:34,401][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:12:34,900][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:12:35,399][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:12:35,899][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:12:36,399][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:12:36,896][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:12:37,395][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:12:37,891][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:12:38,388][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:12:38,883][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:12:39,382][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:12:39,878][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:12:40,379][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:12:40,879][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:12:41,377][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:12:41,879][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:12:42,377][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:12:42,877][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:12:43,397][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:12:43,894][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:12:44,396][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:12:44,893][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:12:45,392][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:12:45,891][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:12:46,390][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:12:46,887][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:12:47,385][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:12:47,882][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:12:48,379][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:12:48,875][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:12:49,373][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:12:49,872][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:12:50,368][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:12:50,863][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:12:51,359][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:12:51,857][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:12:52,357][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:12:52,853][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:12:53,351][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:12:53,849][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10399 tokens. +[2026-03-26 01:12:54,537][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.58%, Block Peak % of device VRAM: 62.24%, ΔTime: 00:00:32 +[2026-03-26 01:12:55,301][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:12:55,304][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:12:55,305][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:12:56,081][__main__][INFO] - Iteration 279 took 54s (34.34% Gen, 64.23% Train). Generation: 18s, Training: 34s. Estimated remaining time: 41h 2m 38s. Estimated total time: 45h 9m 53s. Time estimates for 10 more iterations: 9m 1s, 100 more iterations: 1h 30m 19s, 500 more iterations: 7h 31m 38s. +[2026-03-26 01:12:56,083][__main__][INFO] - Starting iteration 279. +[2026-03-26 01:12:56,481][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 27 and human policies 1. +[2026-03-26 01:12:56,481][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:13:12,881][__main__][INFO] - Number of regex retries in iteration 279: 0 +[2026-03-26 01:13:12,882][__main__][INFO] - agents played in iteration 279 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:13:13,642][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:13:13,665][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:13:13,688][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:13:13,711][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:13:13,712][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:13:13,712][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:13:14,379][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:13:14,835][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:13:15,340][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:13:15,840][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:13:16,339][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:13:16,836][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:13:17,334][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:13:17,832][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:13:18,331][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:13:18,827][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:13:19,324][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:13:19,824][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:13:20,322][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:13:20,819][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:13:21,319][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:13:21,815][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:13:22,315][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:13:22,814][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:13:23,313][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:13:23,809][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:13:24,307][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:13:24,803][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:13:25,303][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:13:25,807][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:13:26,311][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:13:26,807][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:13:27,305][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:13:27,803][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:13:28,299][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:13:28,799][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:13:29,295][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:13:29,794][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:13:30,292][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:13:30,792][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:13:31,292][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:13:31,789][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:13:32,289][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:13:32,789][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:13:33,292][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:13:33,791][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:13:34,291][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:13:34,788][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:13:35,284][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:13:35,780][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:13:36,280][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:13:36,781][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:13:37,281][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:13:37,779][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:13:38,276][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:13:38,777][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:13:39,274][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:13:39,769][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:13:40,266][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:13:40,765][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:13:41,261][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:13:41,760][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:13:42,259][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:13:42,755][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:13:43,255][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:13:43,750][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:13:44,245][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:13:44,741][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:13:45,236][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:13:45,733][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:13:46,228][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10279 tokens. +[2026-03-26 01:13:46,921][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 60.60%, Block Peak % of device VRAM: 62.27%, ΔTime: 00:00:32 +[2026-03-26 01:13:47,668][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:13:47,670][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:13:47,671][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:13:48,443][__main__][INFO] - Iteration 280 took 51s (31.56% Gen, 66.95% Train). Generation: 16s, Training: 34s. Estimated remaining time: 39h 10m 1s. Estimated total time: 43h 18m 9s. Time estimates for 10 more iterations: 8m 39s, 100 more iterations: 1h 26m 36s, 500 more iterations: 7h 13m 1s. +[2026-03-26 01:13:48,446][__main__][INFO] - Starting iteration 280. +[2026-03-26 01:13:48,842][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 27 and human policies 1. +[2026-03-26 01:13:48,843][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:14:06,890][__main__][INFO] - Number of regex retries in iteration 280: 0 +[2026-03-26 01:14:06,891][__main__][INFO] - agents played in iteration 280 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:14:07,653][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:14:07,676][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:14:07,700][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:14:07,723][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:14:07,724][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:14:07,724][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:14:08,397][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:14:08,851][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:14:09,355][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:14:09,853][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:14:10,361][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:14:10,861][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:14:11,358][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:14:11,859][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:14:12,356][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:14:12,863][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:14:13,363][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:14:13,863][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:14:14,359][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:14:14,856][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:14:15,355][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:14:15,860][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:14:16,361][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:14:16,859][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:14:17,356][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:14:17,852][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:14:18,349][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:14:18,845][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:14:19,341][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:14:19,840][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:14:20,335][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:14:20,834][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:14:21,332][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:14:21,831][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:14:22,326][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:14:22,823][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:14:23,321][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:14:23,818][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:14:24,317][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:14:24,819][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:14:25,319][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:14:25,817][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:14:26,315][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:14:26,814][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:14:27,316][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:14:27,815][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:14:28,314][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:14:28,811][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:14:29,309][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:14:29,805][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:14:30,303][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:14:30,801][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:14:31,295][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:14:31,792][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:14:32,292][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:14:32,788][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:14:33,283][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:14:33,779][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:14:34,281][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:14:34,777][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:14:35,273][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:14:35,768][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:14:36,265][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:14:36,770][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:14:37,266][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:14:37,764][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:14:38,260][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:14:38,755][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:14:39,254][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:14:39,752][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:14:40,249][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10392 tokens. +[2026-03-26 01:14:40,952][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.24%, ΔTime: 00:00:32 +[2026-03-26 01:14:41,706][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:14:41,708][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:14:41,710][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:14:43,014][__main__][INFO] - Iteration 281 took 54s (33.32% Gen, 64.27% Train). Generation: 18s, Training: 34s. Estimated remaining time: 40h 59m 35s. Estimated total time: 45h 8m 38s. Time estimates for 10 more iterations: 9m 1s, 100 more iterations: 1h 30m 17s, 500 more iterations: 7h 31m 26s. +[2026-03-26 01:14:43,019][__main__][INFO] - Starting iteration 281. +[2026-03-26 01:14:43,419][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 28 and human policies 1. +[2026-03-26 01:14:43,420][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:14:59,162][__main__][INFO] - Number of regex retries in iteration 281: 0 +[2026-03-26 01:14:59,163][__main__][INFO] - agents played in iteration 281 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:14:59,923][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:14:59,946][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:14:59,970][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:14:59,993][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:14:59,993][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:14:59,994][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:15:00,663][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:15:01,122][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:15:01,627][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:15:02,125][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:15:02,625][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:15:03,125][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:15:03,629][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:15:04,129][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:15:04,626][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:15:05,130][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:15:05,631][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:15:06,128][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:15:06,630][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:15:07,132][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:15:07,630][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:15:08,127][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:15:08,626][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:15:09,127][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:15:09,625][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:15:10,122][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:15:10,616][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:15:11,116][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:15:11,613][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:15:12,113][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:15:12,609][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:15:13,112][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:15:13,615][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:15:14,117][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:15:14,618][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:15:15,118][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:15:15,620][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:15:16,130][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:15:16,631][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:15:17,133][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:15:17,632][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:15:18,133][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:15:18,632][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:15:19,134][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:15:19,634][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:15:20,135][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:15:20,636][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:15:21,138][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:15:21,633][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:15:22,128][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:15:22,630][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:15:23,131][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:15:23,631][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:15:24,128][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:15:24,623][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:15:25,122][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:15:25,619][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:15:26,119][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:15:26,617][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:15:27,115][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:15:27,615][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:15:28,115][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:15:28,613][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:15:29,113][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:15:29,609][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:15:30,108][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:15:30,607][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:15:31,108][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:15:31,611][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:15:32,118][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:15:32,617][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10439 tokens. +[2026-03-26 01:15:33,332][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 60.64%, Block Peak % of device VRAM: 62.39%, ΔTime: 00:00:32 +[2026-03-26 01:15:34,077][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:15:34,080][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:15:34,082][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:15:34,927][__main__][INFO] - Iteration 282 took 51s (30.56% Gen, 67.79% Train). Generation: 15s, Training: 34s. Estimated remaining time: 38h 45m 31s. Estimated total time: 42h 55m 26s. Time estimates for 10 more iterations: 8m 35s, 100 more iterations: 1h 25m 50s, 500 more iterations: 7h 9m 14s. +[2026-03-26 01:15:34,930][__main__][INFO] - Starting iteration 282. +[2026-03-26 01:15:35,330][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 28 and human policies 1. +[2026-03-26 01:15:35,331][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:15:40,366][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 01:15:52,220][__main__][INFO] - Number of regex retries in iteration 282: 1 +[2026-03-26 01:15:52,221][__main__][INFO] - agents played in iteration 282 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:15:53,043][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:15:53,067][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:15:53,090][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:15:53,113][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:15:53,113][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:15:53,114][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:15:53,801][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:15:54,254][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:15:54,764][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:15:55,263][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:15:55,762][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:15:56,262][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:15:56,764][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:15:57,269][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:15:57,773][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:15:58,275][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:15:58,777][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:15:59,278][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:15:59,775][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:16:00,280][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:16:00,778][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:16:01,276][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:16:01,777][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:16:02,279][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:16:02,801][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:16:03,305][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:16:03,805][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:16:04,302][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:16:04,801][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:16:05,299][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:16:05,794][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:16:06,293][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:16:06,791][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:16:07,292][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:16:07,792][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:16:08,297][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:16:08,797][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:16:09,294][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:16:09,790][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:16:10,287][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:16:10,783][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:16:11,283][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:16:11,780][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:16:12,277][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:16:12,783][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:16:13,284][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:16:13,780][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:16:14,280][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:16:14,780][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:16:15,278][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:16:15,775][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:16:16,272][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:16:16,772][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:16:17,271][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:16:17,767][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:16:18,271][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:16:18,769][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:16:19,266][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:16:19,765][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:16:20,262][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:16:20,759][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:16:21,258][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:16:21,754][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:16:22,251][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:16:22,752][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:16:23,249][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:16:23,751][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:16:24,251][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:16:24,749][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:16:25,250][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:16:25,749][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10411 tokens. +[2026-03-26 01:16:26,456][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.21%, ΔTime: 00:00:32 +[2026-03-26 01:16:27,199][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:16:27,201][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:16:27,203][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:16:27,929][__main__][INFO] - Iteration 283 took 52s (32.11% Gen, 66.51% Train). Generation: 16s, Training: 34s. Estimated remaining time: 39h 39m 12s. Estimated total time: 43h 49m 59s. Time estimates for 10 more iterations: 8m 45s, 100 more iterations: 1h 27m 39s, 500 more iterations: 7h 18m 19s. +[2026-03-26 01:16:27,932][__main__][INFO] - Starting iteration 283. +[2026-03-26 01:16:28,332][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 28 and human policies 1. +[2026-03-26 01:16:28,333][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:16:33,204][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 01:16:45,875][__main__][INFO] - Number of regex retries in iteration 283: 1 +[2026-03-26 01:16:45,877][__main__][INFO] - agents played in iteration 283 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:16:46,639][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:16:46,663][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:16:46,686][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:16:46,709][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:16:46,710][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:16:46,711][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:16:47,382][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:16:47,838][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:16:48,342][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:16:48,844][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:16:49,341][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:16:49,837][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:16:50,339][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:16:50,839][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:16:51,342][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:16:51,838][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:16:52,337][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:16:52,836][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:16:53,334][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:16:53,833][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:16:54,330][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:16:54,828][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:16:55,332][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:16:55,827][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:16:56,321][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:16:56,816][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:16:57,314][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:16:57,809][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:16:58,305][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:16:58,803][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:16:59,300][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:16:59,806][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:17:00,310][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:17:00,810][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:17:01,310][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:17:01,809][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:17:02,311][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:17:02,811][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:17:03,311][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:17:03,810][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:17:04,310][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:17:04,808][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:17:05,308][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:17:05,812][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:17:06,312][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:17:06,812][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:17:07,310][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:17:07,808][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:17:08,303][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:17:08,801][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:17:09,299][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:17:09,794][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:17:10,299][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:17:10,799][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:17:11,296][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:17:11,796][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:17:12,292][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:17:12,788][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:17:13,289][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:17:13,784][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:17:14,282][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:17:14,781][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:17:15,278][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:17:15,775][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:17:16,273][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:17:16,768][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:17:17,264][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:17:17,762][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:17:18,259][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:17:18,758][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:17:19,256][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10377 tokens. +[2026-03-26 01:17:19,965][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.58%, Block Peak % of device VRAM: 62.37%, ΔTime: 00:00:32 +[2026-03-26 01:17:20,712][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:17:20,714][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:17:20,716][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:17:21,443][__main__][INFO] - Iteration 284 took 53s (33.03% Gen, 65.60% Train). Generation: 17s, Training: 34s. Estimated remaining time: 40h 3m 53s. Estimated total time: 44h 15m 34s. Time estimates for 10 more iterations: 8m 51s, 100 more iterations: 1h 28m 31s, 500 more iterations: 7h 22m 35s. +[2026-03-26 01:17:21,445][__main__][INFO] - Starting iteration 284. +[2026-03-26 01:17:21,847][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 28 and human policies 1. +[2026-03-26 01:17:21,848][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:17:40,086][__main__][INFO] - Number of regex retries in iteration 284: 0 +[2026-03-26 01:17:40,087][__main__][INFO] - agents played in iteration 284 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:17:40,853][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:17:40,876][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:17:40,899][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:17:40,922][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:17:40,923][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:17:40,923][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:17:41,601][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:17:42,054][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:17:42,556][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:17:43,054][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:17:43,553][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:17:44,050][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:17:44,553][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:17:45,052][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:17:45,552][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:17:46,049][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:17:46,560][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:17:47,062][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:17:47,562][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:17:48,064][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:17:48,564][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:17:49,063][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:17:49,564][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:17:50,068][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:17:50,568][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:17:51,067][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:17:51,564][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:17:52,061][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:17:52,566][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:17:53,068][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:17:53,569][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:17:54,070][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:17:54,569][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:17:55,068][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:17:55,567][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:17:56,066][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:17:56,573][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:17:57,074][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:17:57,573][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:17:58,068][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:17:58,566][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:17:59,061][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:17:59,560][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:18:00,057][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:18:00,556][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:18:01,054][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:18:01,555][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:18:02,055][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:18:02,552][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:18:03,049][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:18:03,545][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:18:04,041][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:18:04,538][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:18:05,036][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:18:05,535][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:18:06,033][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:18:06,531][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:18:07,028][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:18:07,530][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:18:08,026][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:18:08,523][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:18:09,020][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:18:09,518][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:18:10,013][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:18:10,510][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:18:11,010][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:18:11,509][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:18:12,005][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:18:12,508][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:18:13,006][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:18:13,502][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10285 tokens. +[2026-03-26 01:18:14,232][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 60.49%, Block Peak % of device VRAM: 62.40%, ΔTime: 00:00:32 +[2026-03-26 01:18:14,977][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:18:14,979][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:18:14,981][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:18:15,853][__main__][INFO] - Iteration 285 took 54s (33.77% Gen, 64.61% Train). Generation: 18s, Training: 34s. Estimated remaining time: 40h 47m 44s. Estimated total time: 45h 0m 19s. Time estimates for 10 more iterations: 9m 0s, 100 more iterations: 1h 30m 0s, 500 more iterations: 7h 30m 3s. +[2026-03-26 01:18:15,856][__main__][INFO] - Starting iteration 285. +[2026-03-26 01:18:16,257][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 28 and human policies 1. +[2026-03-26 01:18:16,258][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:18:30,528][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 01:18:34,127][__main__][INFO] - Number of regex retries in iteration 285: 1 +[2026-03-26 01:18:34,128][__main__][INFO] - agents played in iteration 285 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:18:34,894][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:18:34,919][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:18:34,943][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:18:34,967][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:18:34,969][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:18:34,969][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:18:35,662][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:18:36,120][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:18:36,625][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:18:37,127][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:18:37,630][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:18:38,134][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:18:38,650][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:18:39,156][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:18:39,654][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:18:40,155][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:18:40,657][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:18:41,155][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:18:41,654][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:18:42,152][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:18:42,655][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:18:43,155][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:18:43,657][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:18:44,163][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:18:44,662][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:18:45,162][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:18:45,658][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:18:46,159][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:18:46,659][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:18:47,158][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:18:47,661][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:18:48,157][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:18:48,658][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:18:49,159][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:18:49,655][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:18:50,151][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:18:50,647][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:18:51,141][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:18:51,638][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:18:52,134][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:18:52,631][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:18:53,127][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:18:53,629][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:18:54,129][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:18:54,625][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:18:55,126][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:18:55,621][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:18:56,118][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:18:56,616][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:18:57,115][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:18:57,614][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:18:58,113][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:18:58,610][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:18:59,110][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:18:59,608][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:19:00,106][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:19:00,602][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:19:01,099][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:19:01,606][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:19:02,107][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:19:02,607][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:19:03,106][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:19:03,603][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:19:04,102][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:19:04,601][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:19:05,098][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:19:05,596][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:19:06,094][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:19:06,591][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:19:07,088][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:19:07,584][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10334 tokens. +[2026-03-26 01:19:08,311][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.09%, Current % of VRAM taken: 60.52%, Block Peak % of device VRAM: 62.33%, ΔTime: 00:00:32 +[2026-03-26 01:19:09,045][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:19:09,047][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:19:09,049][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:19:09,931][__main__][INFO] - Iteration 286 took 53s (33.29% Gen, 65.06% Train). Generation: 17s, Training: 34s. Estimated remaining time: 40h 30m 14s. Estimated total time: 44h 43m 44s. Time estimates for 10 more iterations: 8m 56s, 100 more iterations: 1h 29m 27s, 500 more iterations: 7h 27m 17s. +[2026-03-26 01:19:09,934][__main__][INFO] - Starting iteration 286. +[2026-03-26 01:19:10,338][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 28 and human policies 1. +[2026-03-26 01:19:10,339][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:19:17,038][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 01:19:20,146][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 1 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 01:19:25,975][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 01:19:28,020][__main__][INFO] - Number of regex retries in iteration 286: 3 +[2026-03-26 01:19:28,021][__main__][INFO] - agents played in iteration 286 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:19:28,802][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:19:28,825][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:19:28,849][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:19:28,872][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:19:28,873][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:19:28,873][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:19:29,544][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:19:29,998][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:19:30,499][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:19:30,999][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:19:31,500][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:19:32,003][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:19:32,500][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:19:33,001][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:19:33,497][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:19:33,995][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:19:34,497][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:19:34,996][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:19:35,495][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:19:35,991][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:19:36,493][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:19:36,994][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:19:37,494][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:19:37,991][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:19:38,487][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:19:38,984][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:19:39,484][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:19:39,980][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:19:40,477][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:19:40,975][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:19:41,475][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:19:41,976][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:19:42,474][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:19:42,973][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:19:43,476][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:19:43,976][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:19:44,473][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:19:44,971][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:19:45,476][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:19:45,975][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:19:46,473][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:19:46,973][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:19:47,470][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:19:47,966][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:19:48,462][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:19:48,959][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:19:49,456][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:19:49,955][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:19:50,454][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:19:50,954][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:19:51,454][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:19:51,956][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:19:52,457][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:19:52,978][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:19:53,479][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:19:53,978][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:19:54,476][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:19:54,972][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:19:55,473][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:19:55,973][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:19:56,473][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:19:56,978][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:19:57,478][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:19:57,974][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:19:58,473][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:19:58,970][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:19:59,467][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:19:59,967][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:20:00,473][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:20:00,974][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:20:01,476][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10375 tokens. +[2026-03-26 01:20:02,189][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 60.56%, Block Peak % of device VRAM: 62.35%, ΔTime: 00:00:32 +[2026-03-26 01:20:02,955][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:20:02,957][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:20:02,959][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:20:03,704][__main__][INFO] - Iteration 287 took 53s (33.13% Gen, 65.47% Train). Generation: 17s, Training: 34s. Estimated remaining time: 40h 13m 56s. Estimated total time: 44h 28m 19s. Time estimates for 10 more iterations: 8m 53s, 100 more iterations: 1h 28m 56s, 500 more iterations: 7h 24m 43s. +[2026-03-26 01:20:03,707][__main__][INFO] - Starting iteration 287. +[2026-03-26 01:20:04,105][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 28 and human policies 1. +[2026-03-26 01:20:04,106][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:20:21,329][__main__][INFO] - Number of regex retries in iteration 287: 0 +[2026-03-26 01:20:21,329][__main__][INFO] - agents played in iteration 287 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:20:22,097][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:20:22,120][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:20:22,144][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:20:22,167][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:20:22,167][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:20:22,168][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:20:22,889][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:20:23,345][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:20:23,845][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:20:24,347][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:20:24,846][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:20:25,343][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:20:25,846][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:20:26,343][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:20:26,844][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:20:27,340][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:20:27,840][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:20:28,337][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:20:28,840][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:20:29,337][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:20:29,833][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:20:30,329][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:20:30,825][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:20:31,323][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:20:31,835][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:20:32,337][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:20:32,841][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:20:33,340][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:20:33,839][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:20:34,340][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:20:34,839][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:20:35,337][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:20:35,835][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:20:36,337][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:20:36,834][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:20:37,338][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:20:37,838][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:20:38,341][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:20:38,843][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:20:39,341][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:20:39,839][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:20:40,338][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:20:40,836][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:20:41,333][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:20:41,831][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:20:42,328][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:20:42,825][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:20:43,323][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:20:43,823][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:20:44,324][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:20:44,822][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:20:45,320][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:20:45,828][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:20:46,326][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:20:46,824][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:20:47,323][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:20:47,823][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:20:48,319][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:20:48,818][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:20:49,316][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:20:49,817][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:20:50,316][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:20:50,815][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:20:51,311][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:20:51,808][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:20:52,311][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:20:52,813][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:20:53,309][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:20:53,806][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:20:54,307][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:20:54,807][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10285 tokens. +[2026-03-26 01:20:55,534][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 60.65%, Block Peak % of device VRAM: 62.38%, ΔTime: 00:00:32 +[2026-03-26 01:20:56,279][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:20:56,281][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:20:56,283][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:20:57,141][__main__][INFO] - Iteration 288 took 53s (32.48% Gen, 65.90% Train). Generation: 17s, Training: 34s. Estimated remaining time: 39h 56m 32s. Estimated total time: 44h 11m 49s. Time estimates for 10 more iterations: 8m 50s, 100 more iterations: 1h 28m 23s, 500 more iterations: 7h 21m 58s. +[2026-03-26 01:20:57,143][__main__][INFO] - Starting iteration 288. +[2026-03-26 01:20:57,544][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 28 and human policies 1. +[2026-03-26 01:20:57,545][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:21:02,339][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 01:21:14,697][__main__][INFO] - Number of regex retries in iteration 288: 1 +[2026-03-26 01:21:14,698][__main__][INFO] - agents played in iteration 288 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:21:15,462][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:21:15,485][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:21:15,508][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:21:15,532][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:21:15,533][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:21:15,533][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:21:16,217][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:21:16,676][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:21:17,179][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:21:17,680][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:21:18,178][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:21:18,675][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:21:19,180][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:21:19,679][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:21:20,179][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:21:20,678][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:21:21,180][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:21:21,677][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:21:22,176][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:21:22,688][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:21:23,186][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:21:23,688][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:21:24,187][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:21:24,687][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:21:25,187][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:21:25,687][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:21:26,186][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:21:26,685][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:21:27,185][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:21:27,695][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:21:28,198][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:21:28,700][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:21:29,202][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:21:29,704][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:21:30,200][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:21:30,700][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:21:31,200][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:21:31,700][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:21:32,199][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:21:32,702][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:21:33,202][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:21:33,708][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:21:34,210][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:21:34,712][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:21:35,213][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:21:35,715][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:21:36,214][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:21:36,709][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:21:37,205][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:21:37,704][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:21:38,204][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:21:38,703][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:21:39,199][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:21:39,699][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:21:40,198][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:21:40,694][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:21:41,188][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:21:41,690][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:21:42,185][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:21:42,682][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:21:43,177][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:21:43,673][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:21:44,169][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:21:44,670][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:21:45,168][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:21:45,666][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:21:46,185][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:21:46,686][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:21:47,184][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:21:47,682][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:21:48,182][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10255 tokens. +[2026-03-26 01:21:48,885][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 60.61%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:32 +[2026-03-26 01:21:49,605][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:21:49,607][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:21:49,609][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:21:50,339][__main__][INFO] - Iteration 289 took 52s (32.49% Gen, 66.13% Train). Generation: 17s, Training: 34s. Estimated remaining time: 39h 43m 36s. Estimated total time: 43h 59m 46s. Time estimates for 10 more iterations: 8m 47s, 100 more iterations: 1h 27m 59s, 500 more iterations: 7h 19m 57s. +[2026-03-26 01:21:50,341][__main__][INFO] - Starting iteration 289. +[2026-03-26 01:21:50,738][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 28 and human policies 1. +[2026-03-26 01:21:50,739][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:21:55,740][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 01:21:58,312][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 01:22:12,408][__main__][INFO] - Number of regex retries in iteration 289: 2 +[2026-03-26 01:22:12,409][__main__][INFO] - agents played in iteration 289 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:22:13,622][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:22:13,646][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:22:13,668][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:22:13,691][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:22:13,692][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:22:13,692][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:22:14,446][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:22:15,183][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:22:15,687][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:22:16,190][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:22:16,689][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:22:17,189][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:22:17,688][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:22:18,186][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:22:18,685][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:22:19,185][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:22:19,684][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:22:20,179][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:22:20,686][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:22:21,182][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:22:21,685][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:22:22,188][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:22:22,685][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:22:23,188][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:22:23,687][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:22:24,185][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:22:24,683][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:22:25,186][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:22:25,686][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:22:26,184][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:22:26,686][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:22:27,188][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:22:27,690][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:22:28,191][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:22:28,692][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:22:29,193][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:22:29,690][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:22:30,192][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:22:30,692][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:22:31,191][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:22:31,688][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:22:32,186][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:22:32,687][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:22:33,182][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:22:33,678][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:22:34,175][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:22:34,670][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:22:35,168][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:22:35,673][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:22:36,172][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:22:36,671][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:22:37,169][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:22:37,674][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:22:38,172][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:22:38,669][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:22:39,169][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:22:39,669][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:22:40,165][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:22:40,660][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:22:41,159][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:22:41,657][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:22:42,159][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:22:42,658][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:22:43,157][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:22:43,655][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:22:44,149][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:22:44,645][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:22:45,145][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:22:45,640][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:22:46,136][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:22:46,636][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10399 tokens. +[2026-03-26 01:22:47,380][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.28%, ΔTime: 00:00:32 +[2026-03-26 01:22:48,075][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:22:48,077][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:22:48,079][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:22:48,836][__main__][INFO] - Iteration 290 took 58s (37.30% Gen, 61.40% Train). Generation: 21s, Training: 35s. Estimated remaining time: 44h 7m 46s. Estimated total time: 48h 24m 54s. Time estimates for 10 more iterations: 9m 40s, 100 more iterations: 1h 36m 49s, 500 more iterations: 8h 4m 9s. +[2026-03-26 01:22:48,840][__main__][INFO] - Starting iteration 290. +[2026-03-26 01:22:49,242][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 28 and human policies 1. +[2026-03-26 01:22:49,243][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:24:24,986][__main__][INFO] - Number of regex retries in iteration 290: 0 +[2026-03-26 01:24:24,987][__main__][INFO] - agents played in iteration 290 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:24:25,757][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:24:25,780][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:24:25,804][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:24:25,826][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:24:25,827][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:24:25,828][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:24:26,509][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:24:26,964][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:24:27,467][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:24:27,965][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:24:28,462][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:24:28,964][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:24:29,467][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:24:29,966][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:24:30,463][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:24:30,961][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:24:31,457][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:24:31,954][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:24:32,450][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:24:32,947][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:24:33,443][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:24:33,945][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:24:34,444][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:24:34,942][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:24:35,441][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:24:35,939][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:24:36,437][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:24:36,934][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:24:37,431][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:24:37,928][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:24:38,424][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:24:38,926][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:24:39,426][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:24:39,924][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:24:40,423][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:24:40,924][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:24:41,423][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:24:41,921][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:24:42,425][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:24:42,924][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:24:43,423][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:24:43,923][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:24:44,423][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:24:44,922][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:24:45,442][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:24:45,944][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:24:46,441][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:24:46,937][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:24:47,433][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:24:47,929][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:24:48,430][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:24:48,934][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:24:49,432][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:24:49,929][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:24:50,433][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:24:50,932][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:24:51,434][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:24:51,937][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:24:52,437][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:24:52,938][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:24:53,440][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:24:53,939][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:24:54,441][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:24:54,939][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:24:55,437][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:24:55,935][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:24:56,435][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:24:56,933][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:24:57,433][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:24:57,929][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:24:58,426][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10327 tokens. +[2026-03-26 01:24:59,128][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 60.56%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:32 +[2026-03-26 01:24:59,870][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:24:59,872][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:24:59,874][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:25:01,155][__main__][INFO] - Iteration 291 took 2m 11s (72.58% Gen, 26.45% Train). Generation: 1m 35s, Training: 34s. Estimated remaining time: 105h 36m 18s. Estimated total time: 109h 55m 39s. Time estimates for 10 more iterations: 21m 59s, 100 more iterations: 3h 39m 51s, 500 more iterations: 18h 19m 16s. +[2026-03-26 01:25:01,157][__main__][INFO] - Starting iteration 291. +[2026-03-26 01:25:01,556][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 29 and human policies 1. +[2026-03-26 01:25:01,557][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:25:38,671][__main__][INFO] - Number of regex retries in iteration 291: 0 +[2026-03-26 01:25:38,672][__main__][INFO] - agents played in iteration 291 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:25:39,441][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:25:39,464][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:25:39,487][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:25:39,510][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:25:39,511][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:25:39,511][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:25:40,178][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:25:40,636][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:25:41,137][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:25:41,642][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:25:42,142][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:25:42,643][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:25:43,142][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:25:43,642][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:25:44,140][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:25:44,640][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:25:45,137][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:25:45,634][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:25:46,134][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:25:46,638][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:25:47,134][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:25:47,629][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:25:48,125][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:25:48,626][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:25:49,130][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:25:49,633][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:25:50,132][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:25:50,630][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:25:51,126][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:25:51,622][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:25:52,118][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:25:52,614][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:25:53,112][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:25:53,608][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:25:54,105][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:25:54,608][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:25:55,104][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:25:55,600][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:25:56,100][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:25:56,596][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:25:57,092][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:25:57,587][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:25:58,083][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:25:58,580][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:25:59,080][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:25:59,579][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:26:00,075][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:26:00,575][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:26:01,075][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:26:01,575][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:26:02,078][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:26:02,577][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:26:03,075][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:26:03,572][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:26:04,073][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:26:04,571][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:26:05,067][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:26:05,562][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:26:06,060][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:26:06,558][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:26:07,053][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:26:07,551][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:26:08,044][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:26:08,542][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:26:09,038][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:26:09,539][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:26:10,038][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:26:10,537][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:26:11,034][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:26:11,533][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:26:12,033][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10201 tokens. +[2026-03-26 01:26:12,736][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 60.62%, Block Peak % of device VRAM: 62.21%, ΔTime: 00:00:32 +[2026-03-26 01:26:13,484][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:26:13,486][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:26:13,488][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:26:14,206][__main__][INFO] - Iteration 292 took 1m 12s (51.08% Gen, 47.92% Train). Generation: 37s, Training: 34s. Estimated remaining time: 56h 11m 58s. Estimated total time: 60h 32m 32s. Time estimates for 10 more iterations: 12m 6s, 100 more iterations: 2h 1m 5s, 500 more iterations: 10h 5m 25s. +[2026-03-26 01:26:14,208][__main__][INFO] - Starting iteration 292. +[2026-03-26 01:26:14,607][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 29 and human policies 1. +[2026-03-26 01:26:14,608][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:26:30,917][__main__][INFO] - Number of regex retries in iteration 292: 0 +[2026-03-26 01:26:30,918][__main__][INFO] - agents played in iteration 292 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:26:31,686][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:26:31,709][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:26:31,732][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:26:31,755][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:26:31,756][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:26:31,756][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:26:32,445][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:26:32,902][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:26:33,404][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:26:33,913][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:26:34,414][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:26:34,916][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:26:35,417][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:26:35,915][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:26:36,417][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:26:36,922][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:26:37,421][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:26:37,919][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:26:38,420][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:26:38,920][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:26:39,419][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:26:39,917][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:26:40,414][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:26:40,909][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:26:41,405][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:26:41,902][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:26:42,403][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:26:42,900][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:26:43,397][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:26:43,895][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:26:44,391][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:26:44,888][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:26:45,384][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:26:45,880][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:26:46,380][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:26:46,876][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:26:47,375][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:26:47,875][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:26:48,372][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:26:48,876][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:26:49,376][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:26:49,877][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:26:50,376][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:26:50,874][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:26:51,375][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:26:51,874][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:26:52,375][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:26:52,872][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:26:53,369][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:26:53,863][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:26:54,360][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:26:54,857][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:26:55,352][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:26:55,848][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:26:56,344][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:26:56,840][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:26:57,337][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:26:57,841][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:26:58,339][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:26:58,834][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:26:59,333][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:26:59,830][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:27:00,328][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:27:00,828][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:27:01,323][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:27:01,821][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:27:02,319][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:27:02,822][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:27:03,322][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:27:03,821][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:27:04,317][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10357 tokens. +[2026-03-26 01:27:05,009][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.13%, ΔTime: 00:00:32 +[2026-03-26 01:27:05,761][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:27:05,763][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:27:05,765][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:27:06,583][__main__][INFO] - Iteration 293 took 51s (31.38% Gen, 67.04% Train). Generation: 16s, Training: 34s. Estimated remaining time: 38h 57m 24s. Estimated total time: 43h 18m 51s. Time estimates for 10 more iterations: 8m 39s, 100 more iterations: 1h 26m 37s, 500 more iterations: 7h 13m 8s. +[2026-03-26 01:27:06,586][__main__][INFO] - Starting iteration 293. +[2026-03-26 01:27:06,987][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 29 and human policies 1. +[2026-03-26 01:27:06,988][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:27:14,323][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 01:27:25,356][__main__][INFO] - Number of regex retries in iteration 293: 1 +[2026-03-26 01:27:25,357][__main__][INFO] - agents played in iteration 293 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:27:26,156][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:27:26,180][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:27:26,204][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:27:26,228][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:27:26,228][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:27:26,229][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:27:26,887][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:27:27,344][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:27:27,848][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:27:28,352][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:27:28,850][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:27:29,351][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:27:29,851][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:27:30,347][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:27:30,845][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:27:31,344][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:27:31,849][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:27:32,345][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:27:32,842][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:27:33,339][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:27:33,835][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:27:34,335][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:27:34,834][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:27:35,334][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:27:35,831][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:27:36,334][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:27:36,835][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:27:37,331][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:27:37,830][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:27:38,332][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:27:38,835][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:27:39,332][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:27:39,831][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:27:40,332][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:27:40,831][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:27:41,327][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:27:41,826][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:27:42,323][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:27:42,819][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:27:43,316][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:27:43,817][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:27:44,318][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:27:44,813][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:27:45,311][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:27:45,807][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:27:46,306][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:27:46,803][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:27:47,299][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:27:47,796][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:27:48,293][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:27:48,790][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:27:49,289][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:27:49,788][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:27:50,284][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:27:50,780][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:27:51,279][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:27:51,780][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:27:52,277][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:27:52,777][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:27:53,275][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:27:53,773][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:27:54,273][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:27:54,775][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:27:55,275][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:27:55,773][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:27:56,274][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:27:56,773][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:27:57,271][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:27:57,766][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:27:58,270][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:27:58,769][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10241 tokens. +[2026-03-26 01:27:59,479][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.58%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:32 +[2026-03-26 01:28:00,234][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:28:00,236][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:28:00,237][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:28:00,990][__main__][INFO] - Iteration 294 took 54s (34.01% Gen, 64.59% Train). Generation: 18s, Training: 34s. Estimated remaining time: 40h 37m 49s. Estimated total time: 45h 0m 10s. Time estimates for 10 more iterations: 9m 0s, 100 more iterations: 1h 30m 0s, 500 more iterations: 7h 30m 1s. +[2026-03-26 01:28:00,992][__main__][INFO] - Starting iteration 294. +[2026-03-26 01:28:01,390][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 29 and human policies 1. +[2026-03-26 01:28:01,391][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:28:06,044][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 01:28:10,335][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 01:28:19,565][__main__][INFO] - Number of regex retries in iteration 294: 2 +[2026-03-26 01:28:19,566][__main__][INFO] - agents played in iteration 294 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:28:20,338][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:28:20,361][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:28:20,384][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:28:20,407][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:28:20,408][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:28:20,408][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:28:21,075][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:28:21,534][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:28:22,039][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:28:22,538][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:28:23,047][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:28:23,546][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:28:24,050][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:28:24,549][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:28:25,048][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:28:25,551][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:28:26,049][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:28:26,546][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:28:27,042][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:28:27,539][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:28:28,039][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:28:28,540][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:28:29,038][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:28:29,540][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:28:30,039][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:28:30,537][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:28:31,035][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:28:31,535][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:28:32,036][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:28:32,537][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:28:33,039][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:28:33,539][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:28:34,035][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:28:34,530][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:28:35,028][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:28:35,527][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:28:36,028][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:28:36,527][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:28:37,025][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:28:37,522][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:28:38,022][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:28:38,529][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:28:39,026][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:28:39,523][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:28:40,024][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:28:40,521][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:28:41,023][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:28:41,520][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:28:42,020][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:28:42,520][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:28:43,025][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:28:43,528][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:28:44,025][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:28:44,523][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:28:45,021][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:28:45,519][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:28:46,016][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:28:46,514][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:28:47,015][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:28:47,515][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:28:48,012][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:28:48,522][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:28:49,022][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:28:49,520][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:28:50,020][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:28:50,517][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:28:51,017][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:28:51,516][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:28:52,013][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:28:52,509][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:28:53,007][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10278 tokens. +[2026-03-26 01:28:53,714][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.09%, Current % of VRAM taken: 60.52%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 +[2026-03-26 01:28:54,466][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:28:54,468][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:28:54,470][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:28:55,235][__main__][INFO] - Iteration 295 took 53s (33.75% Gen, 64.82% Train). Generation: 18s, Training: 34s. Estimated remaining time: 40h 29m 1s. Estimated total time: 44h 52m 16s. Time estimates for 10 more iterations: 8m 58s, 100 more iterations: 1h 29m 44s, 500 more iterations: 7h 28m 42s. +[2026-03-26 01:28:55,237][__main__][INFO] - Starting iteration 295. +[2026-03-26 01:28:55,637][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 29 and human policies 1. +[2026-03-26 01:28:55,637][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:29:13,483][__main__][INFO] - Number of regex retries in iteration 295: 0 +[2026-03-26 01:29:13,484][__main__][INFO] - agents played in iteration 295 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:29:14,257][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:29:14,280][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:29:14,304][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:29:14,327][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:29:14,327][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:29:14,328][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:29:15,018][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:29:15,481][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:29:15,990][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:29:16,490][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:29:16,989][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:29:17,487][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:29:17,996][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:29:18,496][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:29:18,995][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:29:19,494][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:29:19,994][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:29:20,495][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:29:20,991][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:29:21,491][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:29:21,989][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:29:22,490][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:29:22,991][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:29:23,497][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:29:23,995][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:29:24,491][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:29:24,989][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:29:25,487][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:29:25,984][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:29:26,483][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:29:26,980][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:29:27,477][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:29:27,974][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:29:28,474][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:29:28,969][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:29:29,464][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:29:29,961][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:29:30,458][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:29:30,957][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:29:31,463][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:29:31,969][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:29:32,469][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:29:32,968][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:29:33,472][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:29:33,977][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:29:34,476][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:29:34,976][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:29:35,476][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:29:35,985][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:29:36,484][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:29:36,985][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:29:37,484][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:29:37,985][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:29:38,487][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:29:38,989][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:29:39,484][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:29:39,980][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:29:40,479][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:29:40,978][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:29:41,477][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:29:41,976][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:29:42,473][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:29:42,969][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:29:43,467][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:29:43,966][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:29:44,463][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:29:44,963][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:29:45,462][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:29:45,963][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:29:46,460][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:29:46,958][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10285 tokens. +[2026-03-26 01:29:47,660][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 60.54%, Block Peak % of device VRAM: 62.36%, ΔTime: 00:00:32 +[2026-03-26 01:29:48,419][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:29:48,421][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:29:48,423][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:29:49,143][__main__][INFO] - Iteration 296 took 53s (33.35% Gen, 65.30% Train). Generation: 17s, Training: 34s. Estimated remaining time: 40h 11m 13s. Estimated total time: 44h 35m 22s. Time estimates for 10 more iterations: 8m 55s, 100 more iterations: 1h 29m 10s, 500 more iterations: 7h 25m 53s. +[2026-03-26 01:29:49,145][__main__][INFO] - Starting iteration 296. +[2026-03-26 01:29:49,545][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 29 and human policies 1. +[2026-03-26 01:29:49,546][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:29:53,360][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 01:29:59,974][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 01:30:05,870][__main__][INFO] - Number of regex retries in iteration 296: 2 +[2026-03-26 01:30:05,871][__main__][INFO] - agents played in iteration 296 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:30:06,642][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:30:06,666][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:30:06,689][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:30:06,712][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:30:06,713][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:30:06,713][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:30:07,383][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:30:07,836][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:30:08,340][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:30:08,837][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:30:09,339][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:30:09,842][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:30:10,338][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:30:10,834][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:30:11,334][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:30:11,862][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:30:12,361][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:30:12,856][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:30:13,352][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:30:13,859][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:30:14,359][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:30:14,858][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:30:15,356][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:30:15,855][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:30:16,378][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:30:16,877][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:30:17,378][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:30:17,876][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:30:18,378][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:30:18,878][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:30:19,377][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:30:19,871][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:30:20,365][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:30:20,860][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:30:21,357][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:30:21,851][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:30:22,344][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:30:22,838][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:30:23,334][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:30:23,830][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:30:24,327][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:30:24,827][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:30:25,323][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:30:25,825][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:30:26,323][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:30:26,821][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:30:27,319][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:30:27,819][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:30:28,317][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:30:28,816][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:30:29,315][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:30:29,816][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:30:30,315][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:30:30,815][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:30:31,314][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:30:31,814][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:30:32,314][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:30:32,812][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:30:33,311][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:30:33,809][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:30:34,305][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:30:34,802][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:30:35,309][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:30:35,812][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:30:36,310][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:30:36,811][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:30:37,308][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:30:37,806][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:30:38,304][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:30:38,802][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:30:39,301][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10266 tokens. +[2026-03-26 01:30:39,994][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 60.67%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:32 +[2026-03-26 01:30:40,844][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:30:40,846][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:30:40,856][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:30:41,713][__main__][INFO] - Iteration 297 took 52s (31.29% Gen, 67.06% Train). Generation: 16s, Training: 34s. Estimated remaining time: 39h 3m 23s. Estimated total time: 43h 28m 25s. Time estimates for 10 more iterations: 8m 41s, 100 more iterations: 1h 26m 56s, 500 more iterations: 7h 14m 44s. +[2026-03-26 01:30:41,716][__main__][INFO] - Starting iteration 297. +[2026-03-26 01:30:42,114][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 29 and human policies 1. +[2026-03-26 01:30:42,115][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:30:58,933][__main__][INFO] - Number of regex retries in iteration 297: 0 +[2026-03-26 01:30:58,934][__main__][INFO] - agents played in iteration 297 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:30:59,719][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:30:59,742][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:30:59,765][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:30:59,788][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:30:59,789][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:30:59,789][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:31:00,452][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:31:00,906][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:31:01,432][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:31:01,932][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:31:02,433][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:31:02,937][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:31:03,438][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:31:03,937][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:31:04,436][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:31:04,942][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:31:05,441][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:31:05,941][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:31:06,440][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:31:06,936][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:31:07,436][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:31:07,939][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:31:08,438][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:31:08,932][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:31:09,432][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:31:09,930][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:31:10,428][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:31:10,925][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:31:11,429][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:31:11,930][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:31:12,427][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:31:12,923][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:31:13,422][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:31:13,920][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:31:14,416][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:31:14,913][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:31:15,415][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:31:15,915][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:31:16,417][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:31:16,917][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:31:17,418][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:31:17,921][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:31:18,421][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:31:18,922][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:31:19,421][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:31:19,921][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:31:20,421][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:31:20,919][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:31:21,415][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:31:21,911][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:31:22,411][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:31:22,907][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:31:23,404][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:31:23,903][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:31:24,401][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:31:24,900][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:31:25,401][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:31:25,897][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:31:26,398][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:31:26,900][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:31:27,401][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:31:27,902][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:31:28,402][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:31:28,900][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:31:29,397][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:31:29,896][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:31:30,396][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:31:30,895][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:31:31,391][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:31:31,892][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:31:32,388][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10489 tokens. +[2026-03-26 01:31:33,089][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 60.51%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:32 +[2026-03-26 01:31:33,866][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:31:33,868][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:31:33,870][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:31:34,632][__main__][INFO] - Iteration 298 took 52s (32.02% Gen, 66.52% Train). Generation: 16s, Training: 34s. Estimated remaining time: 39h 20m 1s. Estimated total time: 43h 45m 56s. Time estimates for 10 more iterations: 8m 45s, 100 more iterations: 1h 27m 31s, 500 more iterations: 7h 17m 39s. +[2026-03-26 01:31:34,634][__main__][INFO] - Starting iteration 298. +[2026-03-26 01:31:35,034][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 29 and human policies 1. +[2026-03-26 01:31:35,035][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:31:52,347][__main__][INFO] - Number of regex retries in iteration 298: 0 +[2026-03-26 01:31:52,348][__main__][INFO] - agents played in iteration 298 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:31:53,120][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:31:53,144][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:31:53,167][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:31:53,190][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:31:53,191][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:31:53,191][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:31:53,851][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:31:54,309][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:31:54,812][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:31:55,314][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:31:55,816][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:31:56,317][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:31:56,816][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:31:57,315][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:31:57,816][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:31:58,316][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:31:58,822][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:31:59,322][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:31:59,824][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:32:00,321][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:32:00,819][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:32:01,317][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:32:01,820][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:32:02,318][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:32:02,819][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:32:03,315][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:32:03,812][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:32:04,308][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:32:04,804][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:32:05,301][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:32:05,802][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:32:06,298][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:32:06,796][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:32:07,296][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:32:07,791][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:32:08,288][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:32:08,784][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:32:09,282][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:32:09,782][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:32:10,282][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:32:10,779][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:32:11,277][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:32:11,783][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:32:12,284][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:32:12,782][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:32:13,283][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:32:13,785][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:32:14,284][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:32:14,785][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:32:15,282][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:32:15,782][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:32:16,281][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:32:16,781][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:32:17,278][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:32:17,776][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:32:18,276][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:32:18,776][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:32:19,277][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:32:19,776][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:32:20,278][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:32:20,776][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:32:21,276][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:32:21,775][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:32:22,272][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:32:22,768][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:32:23,264][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:32:23,761][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:32:24,258][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:32:24,758][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:32:25,255][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:32:25,750][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10354 tokens. +[2026-03-26 01:32:26,443][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 60.52%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:32 +[2026-03-26 01:32:27,189][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:32:27,191][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:32:27,193][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:32:27,913][__main__][INFO] - Iteration 299 took 52s (32.74% Gen, 65.90% Train). Generation: 17s, Training: 34s. Estimated remaining time: 39h 37m 9s. Estimated total time: 44h 3m 57s. Time estimates for 10 more iterations: 8m 48s, 100 more iterations: 1h 28m 7s, 500 more iterations: 7h 20m 39s. +[2026-03-26 01:32:27,915][__main__][INFO] - Starting iteration 299. +[2026-03-26 01:32:28,313][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 29 and human policies 1. +[2026-03-26 01:32:28,314][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:32:45,736][__main__][INFO] - Number of regex retries in iteration 299: 0 +[2026-03-26 01:32:45,737][__main__][INFO] - agents played in iteration 299 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:32:46,507][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:32:46,530][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:32:46,554][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:32:46,577][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:32:46,578][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:32:46,578][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:32:47,247][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:32:47,699][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:32:48,203][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:32:48,700][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:32:49,198][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:32:49,697][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:32:50,209][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:32:50,706][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:32:51,207][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:32:51,704][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:32:52,202][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:32:52,700][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:32:53,198][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:32:53,696][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:32:54,194][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:32:54,692][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:32:55,189][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:32:55,689][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:32:56,188][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:32:56,687][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:32:57,187][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:32:57,686][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:32:58,185][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:32:58,687][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:32:59,184][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:32:59,681][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:33:00,178][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:33:00,677][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:33:01,173][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:33:01,670][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:33:02,166][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:33:02,662][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:33:03,159][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:33:03,654][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:33:04,150][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:33:04,649][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:33:05,144][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:33:05,640][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:33:06,135][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:33:06,635][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:33:07,132][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:33:07,626][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:33:08,130][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:33:08,627][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:33:09,124][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:33:09,620][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:33:10,117][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:33:10,617][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:33:11,117][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:33:11,613][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:33:12,114][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:33:12,613][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:33:13,110][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:33:13,612][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:33:14,111][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:33:14,610][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:33:15,110][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:33:15,627][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:33:16,125][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:33:16,622][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:33:17,119][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:33:17,615][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:33:18,119][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:33:18,617][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:33:19,114][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10305 tokens. +[2026-03-26 01:33:19,804][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.20%, ΔTime: 00:00:32 +[2026-03-26 01:33:20,553][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:33:20,555][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:33:20,557][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:33:21,282][__main__][INFO] - Iteration 300 took 52s (32.89% Gen, 65.74% Train). Generation: 17s, Training: 34s. Estimated remaining time: 39h 40m 48s. Estimated total time: 44h 8m 29s. Time estimates for 10 more iterations: 8m 49s, 100 more iterations: 1h 28m 16s, 500 more iterations: 7h 21m 24s. +[2026-03-26 01:33:21,285][__main__][INFO] - Starting iteration 300. +[2026-03-26 01:33:21,682][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 29 and human policies 1. +[2026-03-26 01:33:21,683][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:33:38,793][__main__][INFO] - Number of regex retries in iteration 300: 0 +[2026-03-26 01:33:38,794][__main__][INFO] - agents played in iteration 300 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:33:39,562][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:33:39,585][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:33:39,609][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:33:39,632][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:33:39,633][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:33:39,633][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:33:40,304][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:33:40,764][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:33:41,267][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:33:41,768][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:33:42,279][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:33:42,780][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:33:43,280][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:33:43,780][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:33:44,280][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:33:44,782][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:33:45,283][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:33:45,784][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:33:46,280][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:33:46,778][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:33:47,289][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:33:47,790][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:33:48,291][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:33:48,787][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:33:49,283][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:33:49,779][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:33:50,276][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:33:50,772][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:33:51,276][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:33:51,772][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:33:52,268][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:33:52,765][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:33:53,262][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:33:53,762][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:33:54,259][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:33:54,756][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:33:55,252][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:33:55,757][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:33:56,254][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:33:56,752][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:33:57,248][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:33:57,745][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:33:58,242][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:33:58,742][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:33:59,236][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:33:59,732][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:34:00,230][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:34:00,726][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:34:01,223][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:34:01,719][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:34:02,225][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:34:02,721][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:34:03,218][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:34:03,712][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:34:04,209][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:34:04,706][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:34:05,202][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:34:05,699][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:34:06,196][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:34:06,692][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:34:07,189][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:34:07,685][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:34:08,185][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:34:08,682][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:34:09,178][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:34:09,678][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:34:10,179][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:34:10,676][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:34:11,177][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:34:11,676][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:34:12,173][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10326 tokens. +[2026-03-26 01:34:12,866][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.57%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:32 +[2026-03-26 01:34:13,796][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:34:13,798][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:34:13,800][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:34:15,163][__main__][INFO] - Iteration 301 took 53s (31.99% Gen, 65.45% Train). Generation: 17s, Training: 35s. Estimated remaining time: 40h 5m 29s. Estimated total time: 44h 34m 3s. Time estimates for 10 more iterations: 8m 54s, 100 more iterations: 1h 29m 8s, 500 more iterations: 7h 25m 40s. +[2026-03-26 01:34:15,165][__main__][INFO] - Starting iteration 301. +[2026-03-26 01:34:15,563][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 30 and human policies 1. +[2026-03-26 01:34:15,564][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:34:32,523][__main__][INFO] - Number of regex retries in iteration 301: 0 +[2026-03-26 01:34:32,523][__main__][INFO] - agents played in iteration 301 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:34:33,301][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:34:33,324][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:34:33,348][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:34:33,371][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:34:33,371][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:34:33,372][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:34:34,035][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:34:34,496][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:34:35,004][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:34:35,504][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:34:36,003][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:34:36,515][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:34:37,018][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:34:37,518][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:34:38,017][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:34:38,513][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:34:39,010][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:34:39,507][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:34:40,005][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:34:40,503][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:34:40,999][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:34:41,500][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:34:42,000][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:34:42,498][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:34:42,999][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:34:43,498][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:34:43,997][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:34:44,496][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:34:44,995][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:34:45,496][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:34:45,994][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:34:46,490][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:34:46,987][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:34:47,485][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:34:47,986][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:34:48,485][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:34:48,983][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:34:49,481][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:34:49,981][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:34:50,481][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:34:50,980][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:34:51,479][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:34:51,978][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:34:52,482][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:34:52,982][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:34:53,481][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:34:53,979][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:34:54,475][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:34:54,972][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:34:55,468][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:34:55,968][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:34:56,469][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:34:56,966][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:34:57,467][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:34:57,967][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:34:58,463][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:34:58,972][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:34:59,470][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:34:59,968][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:35:00,464][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:35:00,961][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:35:01,458][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:35:01,955][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:35:02,451][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:35:02,950][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:35:03,455][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:35:03,957][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:35:04,458][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:35:04,957][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:35:05,456][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:35:05,954][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10392 tokens. +[2026-03-26 01:35:06,652][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 60.60%, Block Peak % of device VRAM: 62.28%, ΔTime: 00:00:32 +[2026-03-26 01:35:07,402][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:35:07,404][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:35:07,406][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:35:08,196][__main__][INFO] - Iteration 302 took 52s (32.22% Gen, 66.27% Train). Generation: 16s, Training: 34s. Estimated remaining time: 39h 22m 13s. Estimated total time: 43h 51m 41s. Time estimates for 10 more iterations: 8m 46s, 100 more iterations: 1h 27m 43s, 500 more iterations: 7h 18m 36s. +[2026-03-26 01:35:08,198][__main__][INFO] - Starting iteration 302. +[2026-03-26 01:35:08,599][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 30 and human policies 1. +[2026-03-26 01:35:08,600][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:35:25,745][__main__][INFO] - Number of regex retries in iteration 302: 0 +[2026-03-26 01:35:25,746][__main__][INFO] - agents played in iteration 302 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:35:26,521][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:35:26,545][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:35:26,568][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:35:26,591][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:35:26,592][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:35:26,592][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:35:27,253][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:35:27,704][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:35:28,208][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:35:28,704][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:35:29,202][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:35:29,708][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:35:30,205][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:35:30,702][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:35:31,198][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:35:31,709][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:35:32,214][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:35:32,712][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:35:33,214][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:35:33,715][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:35:34,212][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:35:34,719][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:35:35,220][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:35:35,719][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:35:36,216][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:35:36,713][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:35:37,209][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:35:37,709][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:35:38,214][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:35:38,714][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:35:39,210][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:35:39,707][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:35:40,204][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:35:40,702][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:35:41,200][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:35:41,698][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:35:42,199][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:35:42,707][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:35:43,207][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:35:43,705][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:35:44,226][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:35:44,729][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:35:45,229][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:35:45,727][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:35:46,226][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:35:46,725][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:35:47,224][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:35:47,720][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:35:48,217][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:35:48,722][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:35:49,220][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:35:49,718][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:35:50,219][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:35:50,720][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:35:51,217][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:35:51,715][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:35:52,210][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:35:52,707][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:35:53,204][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:35:53,701][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:35:54,199][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:35:54,695][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:35:55,193][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:35:55,691][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:35:56,188][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:35:56,693][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:35:57,190][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:35:57,687][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:35:58,187][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:35:58,683][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:35:59,183][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10249 tokens. +[2026-03-26 01:35:59,876][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 60.61%, Block Peak % of device VRAM: 62.17%, ΔTime: 00:00:32 +[2026-03-26 01:36:00,627][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:36:00,629][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:36:00,631][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:36:01,357][__main__][INFO] - Iteration 303 took 52s (32.50% Gen, 66.12% Train). Generation: 17s, Training: 34s. Estimated remaining time: 39h 27m 33s. Estimated total time: 43h 57m 54s. Time estimates for 10 more iterations: 8m 47s, 100 more iterations: 1h 27m 55s, 500 more iterations: 7h 19m 39s. +[2026-03-26 01:36:01,359][__main__][INFO] - Starting iteration 303. +[2026-03-26 01:36:01,763][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 30 and human policies 1. +[2026-03-26 01:36:01,764][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:36:12,660][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 01:36:19,217][__main__][INFO] - Number of regex retries in iteration 303: 1 +[2026-03-26 01:36:19,217][__main__][INFO] - agents played in iteration 303 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:36:20,002][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:36:20,026][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:36:20,049][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:36:20,072][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:36:20,072][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:36:20,073][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:36:20,744][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:36:21,200][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:36:21,706][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:36:22,205][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:36:22,708][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:36:23,212][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:36:23,714][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:36:24,212][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:36:24,711][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:36:25,210][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:36:25,708][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:36:26,207][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:36:26,706][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:36:27,203][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:36:27,705][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:36:28,205][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:36:28,702][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:36:29,199][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:36:29,698][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:36:30,202][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:36:30,701][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:36:31,200][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:36:31,699][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:36:32,198][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:36:32,695][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:36:33,201][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:36:33,700][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:36:34,198][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:36:34,697][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:36:35,198][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:36:35,695][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:36:36,194][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:36:36,690][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:36:37,192][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:36:37,698][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:36:38,198][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:36:38,695][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:36:39,196][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:36:39,692][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:36:40,189][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:36:40,686][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:36:41,187][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:36:41,685][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:36:42,187][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:36:42,688][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:36:43,193][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:36:43,695][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:36:44,194][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:36:44,694][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:36:45,194][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:36:45,695][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:36:46,194][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:36:46,692][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:36:47,191][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:36:47,688][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:36:48,186][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:36:48,685][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:36:49,185][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:36:49,681][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:36:50,179][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:36:50,684][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:36:51,183][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:36:51,683][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:36:52,184][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:36:52,685][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10408 tokens. +[2026-03-26 01:36:53,382][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 60.64%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 +[2026-03-26 01:36:54,151][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:36:54,153][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:36:54,155][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:36:54,900][__main__][INFO] - Iteration 304 took 53s (32.85% Gen, 65.75% Train). Generation: 17s, Training: 34s. Estimated remaining time: 39h 45m 37s. Estimated total time: 44h 16m 52s. Time estimates for 10 more iterations: 8m 51s, 100 more iterations: 1h 28m 33s, 500 more iterations: 7h 22m 48s. +[2026-03-26 01:36:54,902][__main__][INFO] - Starting iteration 304. +[2026-03-26 01:36:55,301][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 30 and human policies 1. +[2026-03-26 01:36:55,302][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:37:10,207][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 01:37:12,001][__main__][INFO] - Number of regex retries in iteration 304: 1 +[2026-03-26 01:37:12,002][__main__][INFO] - agents played in iteration 304 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:37:12,777][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:37:12,800][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:37:12,823][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:37:12,846][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:37:12,847][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:37:12,847][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:37:13,532][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:37:13,987][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:37:14,489][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:37:14,991][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:37:15,496][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:37:15,995][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:37:16,495][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:37:16,992][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:37:17,489][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:37:17,986][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:37:18,482][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:37:18,980][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:37:19,478][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:37:19,977][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:37:20,479][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:37:20,975][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:37:21,471][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:37:21,968][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:37:22,467][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:37:22,973][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:37:23,476][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:37:23,977][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:37:24,477][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:37:24,974][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:37:25,483][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:37:25,981][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:37:26,479][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:37:26,975][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:37:27,470][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:37:27,966][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:37:28,462][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:37:28,958][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:37:29,460][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:37:29,957][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:37:30,454][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:37:30,954][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:37:31,456][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:37:31,956][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:37:32,453][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:37:32,951][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:37:33,447][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:37:33,946][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:37:34,449][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:37:34,946][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:37:35,447][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:37:35,944][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:37:36,444][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:37:36,943][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:37:37,443][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:37:37,946][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:37:38,463][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:37:38,965][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:37:39,464][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:37:39,964][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:37:40,471][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:37:40,971][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:37:41,471][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:37:41,970][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:37:42,469][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:37:42,964][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:37:43,463][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:37:43,962][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:37:44,460][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:37:44,958][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:37:45,457][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10335 tokens. +[2026-03-26 01:37:46,163][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 60.65%, Block Peak % of device VRAM: 62.47%, ΔTime: 00:00:32 +[2026-03-26 01:37:46,931][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:37:46,933][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:37:46,935][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:37:47,673][__main__][INFO] - Iteration 305 took 52s (31.89% Gen, 66.70% Train). Generation: 16s, Training: 34s. Estimated remaining time: 39h 6m 29s. Estimated total time: 43h 38m 36s. Time estimates for 10 more iterations: 8m 43s, 100 more iterations: 1h 27m 17s, 500 more iterations: 7h 16m 26s. +[2026-03-26 01:37:47,676][__main__][INFO] - Starting iteration 305. +[2026-03-26 01:37:48,074][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 30 and human policies 1. +[2026-03-26 01:37:48,074][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:38:04,546][__main__][INFO] - Number of regex retries in iteration 305: 0 +[2026-03-26 01:38:04,547][__main__][INFO] - agents played in iteration 305 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:38:05,320][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.59%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:38:05,344][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.59%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:38:05,367][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.59%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:38:05,395][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.59%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:38:05,396][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:38:05,397][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:38:06,096][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:38:06,551][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:38:07,058][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:38:07,560][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:38:08,066][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:38:08,565][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:38:09,064][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:38:09,561][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:38:10,058][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:38:10,556][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:38:11,061][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:38:11,562][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:38:12,059][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:38:12,555][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:38:13,056][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:38:13,551][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:38:14,047][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:38:14,544][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:38:15,052][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:38:15,550][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:38:16,048][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:38:16,550][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:38:17,047][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:38:17,544][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:38:18,043][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:38:18,542][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:38:19,040][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:38:19,550][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:38:20,052][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:38:20,551][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:38:21,050][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:38:21,550][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:38:22,048][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:38:22,544][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:38:23,042][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:38:23,539][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:38:24,035][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:38:24,531][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:38:25,031][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:38:25,537][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:38:26,035][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:38:26,536][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:38:27,033][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:38:27,536][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:38:28,035][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:38:28,535][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:38:29,034][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:38:29,532][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:38:30,031][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:38:30,531][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:38:31,031][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:38:31,531][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:38:32,033][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:38:32,530][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:38:33,032][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:38:33,536][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:38:34,036][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:38:34,538][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:38:35,037][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:38:35,537][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:38:36,033][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:38:36,530][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:38:37,029][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:38:37,527][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:38:38,031][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10327 tokens. +[2026-03-26 01:38:38,741][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.25%, Current % of VRAM taken: 60.68%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:32 +[2026-03-26 01:38:39,500][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:38:39,502][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:38:39,504][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:38:40,238][__main__][INFO] - Iteration 306 took 52s (31.58% Gen, 67.01% Train). Generation: 16s, Training: 34s. Estimated remaining time: 38h 55m 13s. Estimated total time: 43h 28m 13s. Time estimates for 10 more iterations: 8m 41s, 100 more iterations: 1h 26m 56s, 500 more iterations: 7h 14m 42s. +[2026-03-26 01:38:40,240][__main__][INFO] - Starting iteration 306. +[2026-03-26 01:38:40,641][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 30 and human policies 1. +[2026-03-26 01:38:40,641][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:38:45,119][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 01:38:54,628][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 20 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 01:38:55,391][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 01:38:58,101][__main__][INFO] - Number of regex retries in iteration 306: 3 +[2026-03-26 01:38:58,101][__main__][INFO] - agents played in iteration 306 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:38:58,873][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:38:58,896][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:38:58,919][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:38:58,942][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:38:58,943][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:38:58,944][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:38:59,620][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:39:00,080][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:39:00,590][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:39:01,088][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:39:01,590][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:39:02,091][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:39:02,591][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:39:03,091][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:39:03,590][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:39:04,089][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:39:04,588][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:39:05,091][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:39:05,593][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:39:06,104][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:39:06,610][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:39:07,113][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:39:07,611][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:39:08,108][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:39:08,605][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:39:09,108][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:39:09,607][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:39:10,106][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:39:10,604][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:39:11,102][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:39:11,600][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:39:12,108][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:39:12,609][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:39:13,109][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:39:13,608][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:39:14,106][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:39:14,607][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:39:15,104][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:39:15,605][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:39:16,105][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:39:16,611][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:39:17,110][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:39:17,606][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:39:18,105][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:39:18,603][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:39:19,104][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:39:19,602][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:39:20,101][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:39:20,598][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:39:21,096][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:39:21,597][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:39:22,106][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:39:22,603][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:39:23,104][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:39:23,601][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:39:24,100][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:39:24,597][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:39:25,094][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:39:25,590][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:39:26,086][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:39:26,583][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:39:27,080][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:39:27,578][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:39:28,075][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:39:28,573][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:39:29,069][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:39:29,565][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:39:30,060][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:39:30,557][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:39:31,053][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:39:31,549][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10246 tokens. +[2026-03-26 01:39:32,274][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 60.49%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 +[2026-03-26 01:39:33,040][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:39:33,042][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:39:33,044][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:39:33,808][__main__][INFO] - Iteration 307 took 53s (32.84% Gen, 65.72% Train). Generation: 17s, Training: 34s. Estimated remaining time: 39h 44m 30s. Estimated total time: 44h 18m 23s. Time estimates for 10 more iterations: 8m 51s, 100 more iterations: 1h 28m 36s, 500 more iterations: 7h 23m 3s. +[2026-03-26 01:39:33,811][__main__][INFO] - Starting iteration 307. +[2026-03-26 01:39:34,215][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 30 and human policies 1. +[2026-03-26 01:39:34,215][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:39:52,336][__main__][INFO] - Number of regex retries in iteration 307: 0 +[2026-03-26 01:39:52,336][__main__][INFO] - agents played in iteration 307 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:39:53,111][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:39:53,134][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:39:53,157][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:39:53,181][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:39:53,181][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:39:53,182][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:39:53,868][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:39:54,321][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:39:54,828][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:39:55,326][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:39:55,827][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:39:56,331][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:39:56,829][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:39:57,329][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:39:57,832][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:39:58,329][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:39:58,826][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:39:59,323][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:39:59,823][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:40:00,320][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:40:00,820][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:40:01,320][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:40:01,819][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:40:02,320][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:40:02,820][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:40:03,319][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:40:03,818][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:40:04,318][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:40:04,827][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:40:05,327][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:40:05,826][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:40:06,326][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:40:06,823][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:40:07,320][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:40:07,818][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:40:08,318][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:40:08,819][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:40:09,317][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:40:09,813][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:40:10,311][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:40:10,811][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:40:11,307][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:40:11,804][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:40:12,303][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:40:12,800][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:40:13,297][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:40:13,797][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:40:14,293][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:40:14,789][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:40:15,284][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:40:15,781][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:40:16,281][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:40:16,778][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:40:17,279][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:40:17,777][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:40:18,274][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:40:18,771][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:40:19,266][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:40:19,762][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:40:20,260][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:40:20,757][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:40:21,256][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:40:21,751][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:40:22,247][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:40:22,745][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:40:23,245][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:40:23,741][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:40:24,237][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:40:24,738][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:40:25,235][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:40:25,732][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10186 tokens. +[2026-03-26 01:40:26,450][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 60.56%, Block Peak % of device VRAM: 62.12%, ΔTime: 00:00:32 +[2026-03-26 01:40:27,339][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:40:27,342][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:40:27,353][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:40:28,233][__main__][INFO] - Iteration 308 took 54s (33.55% Gen, 64.82% Train). Generation: 18s, Training: 35s. Estimated remaining time: 40h 26m 8s. Estimated total time: 45h 0m 56s. Time estimates for 10 more iterations: 9m 0s, 100 more iterations: 1h 30m 1s, 500 more iterations: 7h 30m 9s. +[2026-03-26 01:40:28,235][__main__][INFO] - Starting iteration 308. +[2026-03-26 01:40:28,633][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 30 and human policies 1. +[2026-03-26 01:40:28,634][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:40:43,755][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 01:40:45,510][__main__][INFO] - Number of regex retries in iteration 308: 1 +[2026-03-26 01:40:45,511][__main__][INFO] - agents played in iteration 308 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:40:46,292][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:40:46,315][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:40:46,339][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:40:46,363][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:40:46,364][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:40:46,364][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:40:47,051][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:40:47,504][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:40:48,009][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:40:48,507][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:40:49,006][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:40:49,508][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:40:50,010][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:40:50,510][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:40:51,008][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:40:51,515][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:40:52,018][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:40:52,521][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:40:53,019][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:40:53,516][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:40:54,014][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:40:54,512][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:40:55,010][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:40:55,507][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:40:56,010][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:40:56,507][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:40:57,013][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:40:57,514][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:40:58,013][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:40:58,512][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:40:59,012][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:40:59,510][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:41:00,007][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:41:00,504][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:41:01,004][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:41:01,501][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:41:02,000][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:41:02,497][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:41:02,996][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:41:03,507][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:41:04,008][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:41:04,505][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:41:05,007][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:41:05,509][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:41:06,009][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:41:06,510][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:41:07,010][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:41:07,507][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:41:08,005][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:41:08,502][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:41:08,999][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:41:09,497][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:41:09,995][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:41:10,492][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:41:10,990][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:41:11,490][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:41:11,989][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:41:12,488][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:41:12,987][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:41:13,486][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:41:13,986][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:41:14,486][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:41:14,982][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:41:15,480][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:41:15,975][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:41:16,469][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:41:16,965][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:41:17,462][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:41:17,959][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:41:18,458][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:41:18,956][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10279 tokens. +[2026-03-26 01:41:19,672][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 60.48%, Block Peak % of device VRAM: 62.15%, ΔTime: 00:00:32 +[2026-03-26 01:41:20,407][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:41:20,409][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:41:20,410][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:41:21,133][__main__][INFO] - Iteration 309 took 52s (32.15% Gen, 66.48% Train). Generation: 16s, Training: 34s. Estimated remaining time: 39h 9m 20s. Estimated total time: 43h 45m 0s. Time estimates for 10 more iterations: 8m 45s, 100 more iterations: 1h 27m 30s, 500 more iterations: 7h 17m 30s. +[2026-03-26 01:41:21,135][__main__][INFO] - Starting iteration 309. +[2026-03-26 01:41:21,534][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 30 and human policies 1. +[2026-03-26 01:41:21,535][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:41:38,588][__main__][INFO] - Number of regex retries in iteration 309: 0 +[2026-03-26 01:41:38,589][__main__][INFO] - agents played in iteration 309 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:41:39,370][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:41:39,394][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:41:39,417][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:41:39,440][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:41:39,441][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:41:39,441][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:41:40,118][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:41:40,577][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:41:41,084][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:41:41,582][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:41:42,078][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:41:42,585][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:41:43,089][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:41:43,584][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:41:44,080][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:41:44,577][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:41:45,076][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:41:45,595][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:41:46,090][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:41:46,588][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:41:47,089][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:41:47,587][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:41:48,088][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:41:48,585][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:41:49,083][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:41:49,581][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:41:50,085][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:41:50,586][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:41:51,084][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:41:51,581][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:41:52,079][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:41:52,586][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:41:53,085][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:41:53,583][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:41:54,081][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:41:54,580][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:41:55,078][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:41:55,575][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:41:56,080][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:41:56,578][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:41:57,074][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:41:57,573][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:41:58,071][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:41:58,570][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:41:59,067][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:41:59,567][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:42:00,063][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:42:00,559][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:42:01,054][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:42:01,549][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:42:02,046][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:42:02,542][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:42:03,037][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:42:03,534][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:42:04,030][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:42:04,532][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:42:05,030][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:42:05,529][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:42:06,030][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:42:06,533][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:42:07,030][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:42:07,527][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:42:08,026][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:42:08,523][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:42:09,028][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:42:09,529][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:42:10,030][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:42:10,525][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:42:11,023][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:42:11,523][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:42:12,022][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10258 tokens. +[2026-03-26 01:42:12,732][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 60.63%, Block Peak % of device VRAM: 62.10%, ΔTime: 00:00:32 +[2026-03-26 01:42:13,517][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:42:13,520][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:42:13,524][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:42:14,395][__main__][INFO] - Iteration 310 took 52s (32.26% Gen, 66.09% Train). Generation: 17s, Training: 34s. Estimated remaining time: 39h 26m 30s. Estimated total time: 44h 3m 4s. Time estimates for 10 more iterations: 8m 48s, 100 more iterations: 1h 28m 6s, 500 more iterations: 7h 20m 30s. +[2026-03-26 01:42:14,397][__main__][INFO] - Starting iteration 310. +[2026-03-26 01:42:14,795][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 30 and human policies 1. +[2026-03-26 01:42:14,796][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:42:19,003][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 01:42:31,169][__main__][INFO] - Number of regex retries in iteration 310: 1 +[2026-03-26 01:42:31,170][__main__][INFO] - agents played in iteration 310 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:42:31,941][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:42:31,964][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:42:31,987][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:42:32,010][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:42:32,011][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:42:32,011][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:42:32,697][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:42:33,153][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:42:33,667][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:42:34,165][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:42:34,663][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:42:35,164][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:42:35,662][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:42:36,160][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:42:36,659][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:42:37,158][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:42:37,655][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:42:38,154][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:42:38,655][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:42:39,153][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:42:39,652][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:42:40,161][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:42:40,664][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:42:41,160][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:42:41,656][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:42:42,155][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:42:42,655][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:42:43,151][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:42:43,647][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:42:44,145][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:42:44,641][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:42:45,147][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:42:45,644][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:42:46,141][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:42:46,637][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:42:47,133][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:42:47,631][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:42:48,127][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:42:48,631][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:42:49,128][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:42:49,625][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:42:50,121][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:42:50,617][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:42:51,114][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:42:51,609][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:42:52,104][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:42:52,604][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:42:53,114][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:42:53,616][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:42:54,117][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:42:54,617][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:42:55,112][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:42:55,614][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:42:56,110][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:42:56,612][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:42:57,108][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:42:57,604][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:42:58,104][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:42:58,602][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:42:59,108][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:42:59,607][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:43:00,106][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:43:00,604][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:43:01,104][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:43:01,604][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:43:02,105][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:43:02,604][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:43:03,103][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:43:03,602][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:43:04,100][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:43:04,598][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10304 tokens. +[2026-03-26 01:43:05,314][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 60.63%, Block Peak % of device VRAM: 62.28%, ΔTime: 00:00:32 +[2026-03-26 01:43:06,066][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:43:06,068][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:43:06,070][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:43:07,634][__main__][INFO] - Iteration 311 took 52s (30.99% Gen, 66.05% Train). Generation: 16s, Training: 34s. Estimated remaining time: 39h 24m 31s. Estimated total time: 44h 1m 59s. Time estimates for 10 more iterations: 8m 48s, 100 more iterations: 1h 28m 3s, 500 more iterations: 7h 20m 19s. +[2026-03-26 01:43:07,639][__main__][INFO] - Starting iteration 311. +[2026-03-26 01:43:08,036][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 31 and human policies 1. +[2026-03-26 01:43:08,037][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:43:26,093][__main__][INFO] - Number of regex retries in iteration 311: 0 +[2026-03-26 01:43:26,094][__main__][INFO] - agents played in iteration 311 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:43:26,859][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:43:26,882][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:43:26,906][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:43:26,929][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:43:26,930][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:43:26,930][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:43:27,612][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:43:28,070][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:43:28,574][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:43:29,076][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:43:29,574][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:43:30,072][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:43:30,576][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:43:31,077][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:43:31,581][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:43:32,082][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:43:32,579][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:43:33,075][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:43:33,570][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:43:34,068][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:43:34,564][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:43:35,059][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:43:35,554][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:43:36,050][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:43:36,548][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:43:37,044][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:43:37,539][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:43:38,035][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:43:38,529][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:43:39,023][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:43:39,518][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:43:40,013][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:43:40,508][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:43:41,004][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:43:41,502][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:43:41,997][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:43:42,493][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:43:42,989][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:43:43,484][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:43:43,981][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:43:44,476][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:43:44,974][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:43:45,474][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:43:45,970][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:43:46,464][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:43:46,960][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:43:47,459][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:43:47,955][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:43:48,459][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:43:48,957][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:43:49,456][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:43:49,959][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:43:50,456][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:43:50,956][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:43:51,456][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:43:51,954][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:43:52,452][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:43:52,974][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:43:53,472][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:43:53,977][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:43:54,481][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:43:54,981][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:43:55,480][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:43:55,977][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:43:56,474][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:43:56,974][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:43:57,472][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:43:57,971][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:43:58,470][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:43:58,968][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:43:59,465][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10292 tokens. +[2026-03-26 01:44:00,190][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 60.56%, Block Peak % of device VRAM: 62.28%, ΔTime: 00:00:32 +[2026-03-26 01:44:00,937][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:44:00,939][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:44:00,941][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:44:01,781][__main__][INFO] - Iteration 312 took 53s (33.60% Gen, 64.84% Train). Generation: 18s, Training: 34s. Estimated remaining time: 40h 8m 55s. Estimated total time: 44h 47m 16s. Time estimates for 10 more iterations: 8m 57s, 100 more iterations: 1h 29m 34s, 500 more iterations: 7h 27m 52s. +[2026-03-26 01:44:01,784][__main__][INFO] - Starting iteration 312. +[2026-03-26 01:44:02,181][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 31 and human policies 1. +[2026-03-26 01:44:02,182][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:44:20,411][__main__][INFO] - Number of regex retries in iteration 312: 0 +[2026-03-26 01:44:20,411][__main__][INFO] - agents played in iteration 312 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:44:21,180][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:44:21,204][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:44:21,227][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:44:21,250][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:44:21,251][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:44:21,251][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:44:21,933][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:44:22,388][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:44:22,893][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:44:23,392][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:44:23,892][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:44:24,981][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:44:25,566][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:44:27,958][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:44:28,852][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:44:29,491][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:44:30,119][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:44:30,628][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:44:31,141][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:44:31,646][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:44:32,741][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:44:33,370][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:44:34,194][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:44:34,930][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:44:35,849][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:44:36,935][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:44:37,775][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:44:38,772][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:44:39,269][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:44:39,765][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:44:40,263][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:44:40,763][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:44:41,259][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:44:41,754][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:44:42,250][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:44:42,745][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:44:43,239][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:44:43,734][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:44:44,229][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:44:44,727][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:44:45,226][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:44:45,724][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:44:46,223][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:44:46,721][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:44:47,221][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:44:47,719][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:44:48,227][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:44:48,727][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:44:49,224][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:44:49,721][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:44:50,222][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:44:50,722][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:44:51,218][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:44:51,714][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:44:52,213][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:44:52,709][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:44:53,207][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:44:53,703][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:44:54,202][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:44:54,698][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:44:55,193][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:44:55,688][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:44:56,185][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:44:56,684][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:44:57,183][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:44:57,681][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:44:58,182][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:44:58,681][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:44:59,178][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:44:59,678][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:45:00,176][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10323 tokens. +[2026-03-26 01:45:01,000][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 60.53%, Block Peak % of device VRAM: 62.22%, ΔTime: 00:00:39 +[2026-03-26 01:45:01,840][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:45:01,842][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:45:01,843][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:45:02,566][__main__][INFO] - Iteration 313 took 1m 0s (30.19% Gen, 68.61% Train). Generation: 18s, Training: 41s. Estimated remaining time: 45h 39m 54s. Estimated total time: 50h 19m 16s. Time estimates for 10 more iterations: 10m 3s, 100 more iterations: 1h 40m 38s, 500 more iterations: 8h 23m 12s. +[2026-03-26 01:45:02,575][__main__][INFO] - Starting iteration 313. +[2026-03-26 01:45:02,981][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 31 and human policies 1. +[2026-03-26 01:45:02,981][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:46:00,386][__main__][INFO] - Number of regex retries in iteration 313: 0 +[2026-03-26 01:46:00,387][__main__][INFO] - agents played in iteration 313 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:46:05,113][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:03 +[2026-03-26 01:46:05,138][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:46:05,161][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:46:05,184][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:46:05,185][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:46:05,185][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:46:07,912][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:46:08,364][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:46:08,865][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:46:09,366][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:46:09,861][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:46:10,358][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:46:10,867][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:46:11,366][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:46:11,866][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:46:12,364][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:46:12,872][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:46:13,370][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:46:13,875][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:46:14,396][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:46:14,896][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:46:15,392][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:46:15,900][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:46:16,398][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:46:16,897][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:46:17,394][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:46:17,891][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:46:18,392][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:46:18,898][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:46:19,397][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:46:19,896][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:46:20,394][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:46:20,894][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:46:21,390][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:46:21,888][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:46:22,384][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:46:22,884][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:46:23,385][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:46:23,886][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:46:24,389][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:46:24,892][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:46:25,392][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:46:25,893][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:46:26,394][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:46:26,894][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:46:27,395][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:46:27,896][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:46:28,396][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:46:28,904][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:46:29,405][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:46:29,902][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:46:30,400][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:46:30,901][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:46:31,400][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:46:31,903][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:46:32,400][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:46:32,900][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:46:33,399][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:46:33,895][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:46:34,396][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:46:34,894][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:46:35,390][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:46:35,889][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:46:36,388][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:46:36,887][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:46:37,386][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:46:37,883][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:46:38,381][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:46:38,878][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:46:39,374][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:46:39,870][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10334 tokens. +[2026-03-26 01:46:40,583][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 60.57%, Block Peak % of device VRAM: 62.24%, ΔTime: 00:00:32 +[2026-03-26 01:46:41,354][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:46:41,356][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:46:41,357][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:46:42,082][__main__][INFO] - Iteration 314 took 1m 39s (57.92% Gen, 41.34% Train). Generation: 57s, Training: 40s. Estimated remaining time: 77h 54m 5s. Estimated total time: 82h 35m 6s. Time estimates for 10 more iterations: 16m 31s, 100 more iterations: 2h 45m 10s, 500 more iterations: 13h 45m 51s. +[2026-03-26 01:46:42,084][__main__][INFO] - Starting iteration 314. +[2026-03-26 01:46:42,483][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 31 and human policies 1. +[2026-03-26 01:46:42,484][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:47:26,779][__main__][INFO] - Number of regex retries in iteration 314: 0 +[2026-03-26 01:47:26,779][__main__][INFO] - agents played in iteration 314 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:47:27,539][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:47:27,564][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:47:27,590][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:47:27,615][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:47:27,615][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:47:27,616][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:47:28,288][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:47:28,739][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:47:29,245][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:47:29,743][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:47:30,243][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:47:30,740][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:47:31,239][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:47:31,737][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:47:32,236][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:47:32,734][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:47:33,229][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:47:33,732][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:47:34,228][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:47:34,724][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:47:35,220][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:47:35,716][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:47:36,211][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:47:36,716][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:47:37,216][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:47:37,732][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:47:38,237][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:47:38,737][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:47:39,236][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:47:39,737][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:47:40,237][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:47:40,739][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:47:41,235][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:47:41,741][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:47:42,239][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:47:42,736][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:47:43,233][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:47:43,730][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:47:44,227][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:47:44,722][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:47:45,222][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:47:45,718][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:47:46,215][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:47:46,711][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:47:47,208][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:47:47,703][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:47:48,198][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:47:48,693][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:47:49,188][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:47:49,683][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:47:50,181][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:47:50,677][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:47:51,172][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:47:51,671][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:47:52,166][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:47:52,662][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:47:53,157][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:47:53,653][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:47:54,152][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:47:54,646][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:47:55,143][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:47:55,641][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:47:56,140][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:47:56,639][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:47:57,135][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:47:57,631][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:47:58,127][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:47:58,626][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:47:59,124][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:47:59,621][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:48:00,120][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10128 tokens. +[2026-03-26 01:48:00,814][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 60.56%, Block Peak % of device VRAM: 62.38%, ΔTime: 00:00:32 +[2026-03-26 01:48:01,584][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:48:01,586][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:48:01,588][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:48:02,463][__main__][INFO] - Iteration 315 took 1m 19s (55.38% Gen, 43.52% Train). Generation: 44s, Training: 34s. Estimated remaining time: 61h 56m 39s. Estimated total time: 66h 39m 1s. Time estimates for 10 more iterations: 13m 19s, 100 more iterations: 2h 13m 18s, 500 more iterations: 11h 6m 30s. +[2026-03-26 01:48:02,465][__main__][INFO] - Starting iteration 315. +[2026-03-26 01:48:02,862][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 31 and human policies 1. +[2026-03-26 01:48:02,863][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:48:21,108][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 01:48:22,326][__main__][INFO] - Number of regex retries in iteration 315: 1 +[2026-03-26 01:48:22,327][__main__][INFO] - agents played in iteration 315 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:48:23,086][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:48:23,110][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:48:23,133][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:48:23,156][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:48:23,157][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:48:23,157][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:48:23,829][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:48:24,283][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:48:24,782][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:48:25,278][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:48:25,773][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:48:26,268][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:48:26,765][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:48:27,262][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:48:27,761][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:48:28,263][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:48:28,760][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:48:29,261][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:48:29,757][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:48:30,254][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:48:30,752][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:48:31,248][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:48:31,745][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:48:32,243][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:48:32,743][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:48:33,244][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:48:33,742][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:48:34,239][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:48:34,739][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:48:35,238][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:48:35,735][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:48:36,234][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:48:36,733][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:48:37,232][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:48:37,728][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:48:38,229][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:48:38,728][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:48:39,227][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:48:39,723][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:48:40,225][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:48:40,724][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:48:41,221][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:48:41,720][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:48:42,222][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:48:42,717][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:48:43,214][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:48:43,710][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:48:44,206][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:48:44,704][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:48:45,204][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:48:45,704][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:48:46,202][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:48:46,701][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:48:47,197][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:48:47,693][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:48:48,190][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:48:48,686][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:48:49,183][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:48:49,682][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:48:50,182][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:48:50,682][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:48:51,178][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:48:51,675][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:48:52,174][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:48:52,669][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:48:53,164][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:48:53,661][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:48:54,158][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:48:54,654][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:48:55,150][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:48:55,646][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10256 tokens. +[2026-03-26 01:48:56,347][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 60.51%, Block Peak % of device VRAM: 62.13%, ΔTime: 00:00:32 +[2026-03-26 01:48:57,189][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:48:57,191][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:48:57,192][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:48:57,854][__main__][INFO] - Iteration 316 took 54s (35.39% Gen, 63.40% Train). Generation: 19s, Training: 34s. Estimated remaining time: 41h 6m 20s. Estimated total time: 45h 49m 38s. Time estimates for 10 more iterations: 9m 9s, 100 more iterations: 1h 31m 39s, 500 more iterations: 7h 38m 16s. +[2026-03-26 01:48:57,856][__main__][INFO] - Starting iteration 316. +[2026-03-26 01:48:58,256][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 31 and human policies 1. +[2026-03-26 01:48:58,257][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:49:15,177][__main__][INFO] - Number of regex retries in iteration 316: 0 +[2026-03-26 01:49:15,178][__main__][INFO] - agents played in iteration 316 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:49:15,935][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:49:15,958][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:49:15,981][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:49:16,005][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:49:16,006][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:49:16,006][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:49:16,671][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:49:17,122][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:49:17,626][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:49:18,128][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:49:18,624][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:49:19,121][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:49:19,618][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:49:20,114][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:49:20,611][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:49:21,108][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:49:21,605][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:49:22,102][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:49:22,600][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:49:23,099][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:49:23,595][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:49:24,091][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:49:24,588][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:49:25,087][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:49:25,584][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:49:26,083][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:49:26,589][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:49:27,090][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:49:27,590][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:49:28,087][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:49:28,588][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:49:29,086][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:49:29,586][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:49:30,083][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:49:30,582][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:49:31,079][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:49:31,577][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:49:32,076][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:49:32,574][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:49:33,071][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:49:33,574][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:49:34,073][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:49:34,572][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:49:35,071][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:49:35,567][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:49:36,066][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:49:36,566][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:49:37,064][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:49:37,560][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:49:38,060][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:49:38,556][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:49:39,051][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:49:39,549][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:49:40,044][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:49:40,541][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:49:41,038][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:49:41,534][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:49:42,030][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:49:42,527][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:49:43,024][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:49:43,524][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:49:44,023][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:49:44,521][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:49:45,020][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:49:45,539][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:49:46,043][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:49:46,541][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:49:47,039][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:49:47,537][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:49:48,036][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:49:48,534][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10208 tokens. +[2026-03-26 01:49:49,231][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 60.61%, Block Peak % of device VRAM: 62.21%, ΔTime: 00:00:32 +[2026-03-26 01:49:50,015][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:49:50,017][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:49:50,019][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:49:50,767][__main__][INFO] - Iteration 317 took 52s (32.22% Gen, 66.35% Train). Generation: 16s, Training: 34s. Estimated remaining time: 39h 1m 22s. Estimated total time: 43h 45m 32s. Time estimates for 10 more iterations: 8m 45s, 100 more iterations: 1h 27m 31s, 500 more iterations: 7h 17m 35s. +[2026-03-26 01:49:50,769][__main__][INFO] - Starting iteration 317. +[2026-03-26 01:49:51,165][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 31 and human policies 1. +[2026-03-26 01:49:51,166][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:49:59,619][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 01:50:07,241][__main__][INFO] - Number of regex retries in iteration 317: 1 +[2026-03-26 01:50:07,242][__main__][INFO] - agents played in iteration 317 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:50:08,000][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:50:08,024][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:50:08,047][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:50:08,070][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:50:08,070][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:50:08,071][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:50:08,745][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:50:09,199][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:50:09,702][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:50:10,216][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:50:10,722][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:50:11,219][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:50:11,721][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:50:12,221][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:50:12,719][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:50:13,216][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:50:13,713][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:50:14,210][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:50:14,709][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:50:15,205][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:50:15,702][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:50:16,201][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:50:16,712][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:50:17,213][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:50:17,711][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:50:18,208][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:50:18,708][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:50:19,205][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:50:19,703][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:50:20,202][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:50:20,700][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:50:21,205][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:50:21,706][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:50:22,204][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:50:22,702][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:50:23,200][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:50:23,698][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:50:24,199][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:50:24,698][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:50:25,195][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:50:25,692][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:50:26,188][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:50:26,685][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:50:27,183][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:50:27,681][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:50:28,179][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:50:28,677][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:50:29,178][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:50:29,676][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:50:30,174][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:50:30,670][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:50:31,178][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:50:31,676][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:50:32,177][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:50:32,675][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:50:33,174][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:50:33,698][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:50:34,196][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:50:34,698][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:50:35,204][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:50:35,703][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:50:36,201][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:50:36,700][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:50:37,202][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:50:37,703][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:50:38,204][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:50:38,704][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:50:39,201][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:50:39,701][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:50:40,201][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:50:40,701][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10243 tokens. +[2026-03-26 01:50:41,398][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 60.62%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:32 +[2026-03-26 01:50:42,172][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:50:42,174][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:50:42,175][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:50:42,897][__main__][INFO] - Iteration 318 took 51s (31.07% Gen, 67.53% Train). Generation: 16s, Training: 34s. Estimated remaining time: 38h 21m 32s. Estimated total time: 43h 6m 35s. Time estimates for 10 more iterations: 8m 37s, 100 more iterations: 1h 26m 13s, 500 more iterations: 7h 11m 5s. +[2026-03-26 01:50:42,899][__main__][INFO] - Starting iteration 318. +[2026-03-26 01:50:43,297][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 31 and human policies 1. +[2026-03-26 01:50:43,297][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:51:01,308][__main__][INFO] - Number of regex retries in iteration 318: 0 +[2026-03-26 01:51:01,309][__main__][INFO] - agents played in iteration 318 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:51:02,080][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:51:02,103][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:51:02,126][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:51:02,150][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:51:02,151][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:51:02,151][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:51:02,814][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:51:03,270][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:51:03,772][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:51:04,270][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:51:04,769][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:51:05,272][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:51:05,770][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:51:06,271][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:51:06,771][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:51:07,273][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:51:07,772][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:51:08,272][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:51:08,776][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:51:09,277][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:51:09,777][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:51:10,297][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:51:10,799][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:51:11,296][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:51:11,793][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:51:12,301][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:51:12,801][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:51:13,297][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:51:13,798][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:51:14,301][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:51:14,801][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:51:15,297][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:51:15,794][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:51:16,292][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:51:16,788][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:51:17,286][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:51:17,785][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:51:18,282][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:51:18,781][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:51:19,281][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:51:19,777][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:51:20,274][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:51:20,770][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:51:21,266][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:51:21,765][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:51:22,265][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:51:22,764][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:51:23,260][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:51:23,755][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:51:24,250][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:51:24,745][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:51:25,243][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:51:25,738][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:51:26,232][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:51:26,727][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:51:27,221][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:51:27,716][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:51:28,210][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:51:28,709][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:51:29,207][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:51:29,703][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:51:30,202][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:51:30,698][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:51:31,195][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:51:31,691][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:51:32,188][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:51:32,686][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:51:33,183][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:51:33,683][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:51:34,182][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:51:34,677][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10230 tokens. +[2026-03-26 01:51:35,375][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 60.50%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:32 +[2026-03-26 01:51:36,161][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:51:36,164][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:51:36,165][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:51:36,879][__main__][INFO] - Iteration 319 took 53s (33.61% Gen, 65.05% Train). Generation: 18s, Training: 34s. Estimated remaining time: 39h 53m 13s. Estimated total time: 44h 39m 10s. Time estimates for 10 more iterations: 8m 55s, 100 more iterations: 1h 29m 18s, 500 more iterations: 7h 26m 31s. +[2026-03-26 01:51:36,882][__main__][INFO] - Starting iteration 319. +[2026-03-26 01:51:37,280][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 31 and human policies 1. +[2026-03-26 01:51:37,281][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:51:55,979][__main__][INFO] - Number of regex retries in iteration 319: 0 +[2026-03-26 01:51:55,980][__main__][INFO] - agents played in iteration 319 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:51:56,743][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:51:56,766][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:51:56,790][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:51:56,813][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:51:56,813][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:51:56,814][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:51:57,489][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:51:57,942][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:51:58,444][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:51:58,939][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:51:59,432][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:51:59,927][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:52:00,422][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:52:00,918][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:52:01,414][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:52:01,926][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:52:02,427][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:52:02,931][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:52:03,430][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:52:03,930][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:52:04,429][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:52:04,929][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:52:05,449][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:52:05,948][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:52:06,444][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:52:06,942][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:52:07,441][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:52:07,938][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:52:08,434][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:52:08,932][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:52:09,431][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:52:09,931][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:52:10,428][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:52:10,926][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:52:11,426][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:52:11,924][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:52:12,421][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:52:12,920][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:52:13,415][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:52:13,914][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:52:14,412][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:52:14,910][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:52:15,414][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:52:15,912][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:52:16,412][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:52:16,911][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:52:17,410][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:52:17,907][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:52:18,405][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:52:18,902][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:52:19,399][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:52:19,896][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:52:20,392][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:52:20,892][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:52:21,388][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:52:21,884][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:52:22,379][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:52:22,874][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:52:23,368][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:52:23,865][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:52:24,362][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:52:24,862][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:52:25,358][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:52:25,852][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:52:26,351][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:52:26,845][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:52:27,341][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:52:27,839][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:52:28,335][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:52:28,832][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:52:29,326][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10243 tokens. +[2026-03-26 01:52:30,029][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 60.54%, Block Peak % of device VRAM: 62.36%, ΔTime: 00:00:32 +[2026-03-26 01:52:30,800][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:52:30,802][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:52:30,803][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:52:31,563][__main__][INFO] - Iteration 320 took 54s (34.45% Gen, 64.15% Train). Generation: 18s, Training: 34s. Estimated remaining time: 40h 27m 19s. Estimated total time: 45h 14m 10s. Time estimates for 10 more iterations: 9m 2s, 100 more iterations: 1h 30m 28s, 500 more iterations: 7h 32m 21s. +[2026-03-26 01:52:31,565][__main__][INFO] - Starting iteration 320. +[2026-03-26 01:52:31,963][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 31 and human policies 1. +[2026-03-26 01:52:31,964][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:52:39,813][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 01:52:41,404][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 01:52:44,928][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 1 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 01:52:49,564][__main__][INFO] - Number of regex retries in iteration 320: 3 +[2026-03-26 01:52:49,565][__main__][INFO] - agents played in iteration 320 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:52:50,324][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:52:50,348][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:52:50,372][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:52:50,395][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:52:50,396][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:52:50,396][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:52:51,083][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:52:51,539][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:52:52,046][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:52:52,544][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:52:53,050][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:52:53,549][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:52:54,049][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:52:54,549][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:52:55,046][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:52:55,556][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:52:56,054][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:52:56,551][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:52:57,049][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:52:57,550][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:52:58,049][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:52:58,559][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:52:59,058][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:52:59,556][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:53:00,057][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:53:00,554][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:53:01,051][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:53:01,548][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:53:02,048][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:53:02,554][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:53:03,052][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:53:03,551][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:53:04,048][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:53:04,547][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:53:05,046][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:53:05,544][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:53:06,045][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:53:06,542][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:53:07,049][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:53:07,546][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:53:08,042][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:53:08,540][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:53:09,036][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:53:09,531][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:53:10,030][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:53:10,526][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:53:11,024][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:53:11,521][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:53:12,019][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:53:12,518][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:53:13,022][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:53:13,521][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:53:14,021][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:53:14,522][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:53:15,022][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:53:15,522][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:53:16,020][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:53:16,526][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:53:17,026][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:53:17,525][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:53:18,024][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:53:18,521][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:53:19,022][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:53:19,518][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:53:20,031][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:53:20,529][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:53:21,027][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:53:21,523][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:53:22,021][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:53:22,518][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:53:23,015][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10258 tokens. +[2026-03-26 01:53:23,721][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.22%, ΔTime: 00:00:32 +[2026-03-26 01:53:24,489][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:53:24,491][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:53:24,493][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:53:26,018][__main__][INFO] - Iteration 321 took 54s (32.56% Gen, 64.61% Train). Generation: 17s, Training: 34s. Estimated remaining time: 40h 15m 1s. Estimated total time: 45h 2m 46s. Time estimates for 10 more iterations: 9m 0s, 100 more iterations: 1h 30m 5s, 500 more iterations: 7h 30m 27s. +[2026-03-26 01:53:26,021][__main__][INFO] - Starting iteration 321. +[2026-03-26 01:53:26,421][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 32 and human policies 1. +[2026-03-26 01:53:26,421][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:53:44,021][__main__][INFO] - Number of regex retries in iteration 321: 0 +[2026-03-26 01:53:44,022][__main__][INFO] - agents played in iteration 321 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:53:44,797][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:53:44,821][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:53:44,844][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:53:44,867][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:53:44,868][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:53:44,868][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:53:45,555][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:53:46,012][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:53:46,513][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:53:47,011][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:53:47,518][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:53:48,019][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:53:48,521][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:53:49,018][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:53:49,513][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:53:50,011][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:53:50,508][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:53:51,006][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:53:51,504][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:53:52,001][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:53:52,500][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:53:52,999][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:53:53,496][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:53:53,993][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:53:54,490][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:53:54,989][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:53:55,485][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:53:55,985][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:53:56,484][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:53:56,982][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:53:57,481][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:53:57,985][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:53:58,507][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:53:59,005][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:53:59,502][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:54:00,000][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:54:00,499][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:54:00,997][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:54:01,495][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:54:01,995][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:54:02,493][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:54:02,989][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:54:03,486][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:54:03,985][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:54:04,481][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:54:04,979][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:54:05,480][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:54:05,980][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:54:06,483][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:54:06,983][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:54:07,483][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:54:07,981][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:54:08,478][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:54:08,975][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:54:09,473][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:54:09,969][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:54:10,467][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:54:10,970][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:54:11,467][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:54:11,964][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:54:12,464][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:54:12,964][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:54:13,466][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:54:13,967][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:54:14,463][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:54:14,962][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:54:15,463][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:54:15,962][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:54:16,458][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:54:16,958][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:54:17,457][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10275 tokens. +[2026-03-26 01:54:18,193][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 60.66%, Block Peak % of device VRAM: 62.21%, ΔTime: 00:00:32 +[2026-03-26 01:54:19,108][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:54:19,111][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:54:19,113][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:54:20,695][__main__][INFO] - Iteration 322 took 54s (32.43% Gen, 64.65% Train). Generation: 17s, Training: 35s. Estimated remaining time: 40h 25m 4s. Estimated total time: 45h 13m 44s. Time estimates for 10 more iterations: 9m 2s, 100 more iterations: 1h 30m 27s, 500 more iterations: 7h 32m 17s. +[2026-03-26 01:54:20,711][__main__][INFO] - Starting iteration 322. +[2026-03-26 01:54:21,333][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 32 and human policies 1. +[2026-03-26 01:54:21,334][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:55:10,299][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 01:55:21,111][__main__][INFO] - Number of regex retries in iteration 322: 1 +[2026-03-26 01:55:21,112][__main__][INFO] - agents played in iteration 322 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:55:21,989][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:55:22,012][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:55:22,035][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:55:22,058][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:55:22,059][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:55:22,059][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:55:22,751][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:55:23,344][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:55:23,853][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:55:24,352][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:55:24,849][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:55:25,345][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:55:25,851][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:55:26,349][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:55:26,846][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:55:27,347][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:55:27,846][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:55:28,346][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:55:28,855][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:55:29,355][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:55:29,856][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:55:30,356][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:55:30,861][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:55:31,358][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:55:31,862][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:55:32,369][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:55:32,867][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:55:33,366][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:55:33,864][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:55:34,367][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:55:34,864][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:55:35,362][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:55:35,859][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:55:36,359][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:55:36,856][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:55:37,352][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:55:37,849][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:55:38,352][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:55:38,853][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:55:39,350][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:55:39,848][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:55:40,347][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:55:40,843][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:55:41,340][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:55:41,838][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:55:42,335][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:55:42,831][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:55:43,328][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:55:43,827][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:55:44,324][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:55:44,821][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:55:45,321][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:55:45,820][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:55:46,317][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:55:46,813][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:55:47,310][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:55:47,816][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:55:48,318][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:55:48,818][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:55:49,315][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:55:49,813][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:55:50,313][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:55:50,812][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:55:51,313][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:55:51,811][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:55:52,308][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:55:52,807][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:55:53,306][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:55:53,806][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:55:54,306][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:55:54,803][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10275 tokens. +[2026-03-26 01:55:55,577][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.18%, ΔTime: 00:00:32 +[2026-03-26 01:55:56,386][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:55:56,388][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:55:56,389][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:55:57,098][__main__][INFO] - Iteration 323 took 1m 35s (62.42% Gen, 36.84% Train). Generation: 59s, Training: 35s. Estimated remaining time: 74h 58m 7s. Estimated total time: 79h 48m 24s. Time estimates for 10 more iterations: 15m 57s, 100 more iterations: 2h 39m 36s, 500 more iterations: 13h 18m 4s. +[2026-03-26 01:55:57,105][__main__][INFO] - Starting iteration 323. +[2026-03-26 01:55:57,508][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 32 and human policies 1. +[2026-03-26 01:55:57,509][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:56:25,329][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 01:56:28,258][__main__][INFO] - Number of regex retries in iteration 323: 1 +[2026-03-26 01:56:28,259][__main__][INFO] - agents played in iteration 323 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:56:29,018][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:56:29,042][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:56:29,065][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:56:29,088][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:56:29,089][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:56:29,089][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:56:29,768][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:56:30,228][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:56:30,730][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:56:31,227][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:56:31,724][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:56:32,223][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:56:32,718][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:56:33,214][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:56:33,710][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:56:34,211][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:56:34,710][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:56:35,208][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:56:35,706][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:56:36,215][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:56:36,713][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:56:37,213][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:56:37,712][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:56:38,209][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:56:38,710][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:56:39,212][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:56:39,714][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:56:40,214][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:56:40,711][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:56:41,211][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:56:41,713][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:56:42,215][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:56:42,713][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:56:43,209][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:56:43,707][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:56:44,203][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:56:44,701][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:56:45,199][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:56:45,705][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:56:46,203][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:56:46,700][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:56:47,196][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:56:47,692][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:56:48,191][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:56:48,686][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:56:49,183][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:56:49,680][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:56:50,186][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:56:50,684][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:56:51,181][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:56:51,677][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:56:52,176][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:56:52,670][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:56:53,169][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:56:53,665][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:56:54,161][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:56:54,657][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:56:55,152][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:56:55,647][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:56:56,142][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:56:56,636][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:56:57,132][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:56:57,628][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:56:58,127][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:56:58,623][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:56:59,127][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:56:59,627][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:57:00,126][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:57:00,627][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:57:01,126][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:57:01,625][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10208 tokens. +[2026-03-26 01:57:02,335][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 60.61%, Block Peak % of device VRAM: 62.27%, ΔTime: 00:00:32 +[2026-03-26 01:57:03,114][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:57:03,116][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:57:03,118][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:57:03,836][__main__][INFO] - Iteration 324 took 1m 6s (46.36% Gen, 52.56% Train). Generation: 30s, Training: 34s. Estimated remaining time: 50h 24m 59s. Estimated total time: 55h 16m 23s. Time estimates for 10 more iterations: 11m 3s, 100 more iterations: 1h 50m 32s, 500 more iterations: 9h 12m 43s. +[2026-03-26 01:57:03,838][__main__][INFO] - Starting iteration 324. +[2026-03-26 01:57:04,238][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 32 and human policies 1. +[2026-03-26 01:57:04,238][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:57:34,458][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 01:57:35,663][__main__][INFO] - Number of regex retries in iteration 324: 1 +[2026-03-26 01:57:35,664][__main__][INFO] - agents played in iteration 324 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:57:36,433][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:57:36,457][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:57:36,480][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:57:36,504][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:57:36,504][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:57:36,505][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:57:37,181][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:57:37,633][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:57:38,137][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:57:38,634][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:57:39,142][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:57:39,642][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:57:40,143][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:57:40,642][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:57:41,141][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:57:41,638][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:57:42,134][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:57:42,638][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:57:43,157][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:57:43,653][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:57:44,149][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:57:44,645][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:57:45,142][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:57:45,649][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:57:46,148][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:57:46,645][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:57:47,147][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:57:47,645][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:57:48,143][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:57:48,641][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:57:49,137][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:57:49,634][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:57:50,131][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:57:50,627][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:57:51,127][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:57:51,623][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:57:52,123][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:57:52,620][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:57:53,117][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:57:53,616][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:57:54,116][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:57:54,614][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:57:55,121][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:57:55,622][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:57:56,122][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:57:56,621][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:57:57,121][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:57:57,619][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:57:58,115][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:57:58,612][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:57:59,109][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:57:59,607][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:58:00,103][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:58:00,600][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:58:01,096][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:58:01,596][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:58:02,098][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:58:02,595][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:58:03,096][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:58:03,592][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:58:04,090][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:58:04,591][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:58:05,100][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:58:05,600][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:58:06,100][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:58:06,602][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:58:07,101][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:58:07,601][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:58:08,101][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:58:08,600][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:58:09,099][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10173 tokens. +[2026-03-26 01:58:09,817][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 60.57%, Block Peak % of device VRAM: 62.17%, ΔTime: 00:00:32 +[2026-03-26 01:58:10,564][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:58:10,567][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:58:10,569][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:58:11,296][__main__][INFO] - Iteration 325 took 1m 7s (46.86% Gen, 52.05% Train). Generation: 31s, Training: 34s. Estimated remaining time: 51h 0m 25s. Estimated total time: 55h 52m 56s. Time estimates for 10 more iterations: 11m 10s, 100 more iterations: 1h 51m 45s, 500 more iterations: 9h 18m 49s. +[2026-03-26 01:58:11,298][__main__][INFO] - Starting iteration 325. +[2026-03-26 01:58:11,696][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 32 and human policies 1. +[2026-03-26 01:58:11,697][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:58:16,360][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 01:58:27,026][__main__][INFO] - Number of regex retries in iteration 325: 1 +[2026-03-26 01:58:27,027][__main__][INFO] - agents played in iteration 325 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:58:27,783][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:58:27,807][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:58:27,830][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:58:27,853][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:58:27,854][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:58:27,854][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:58:28,528][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:58:28,979][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:58:29,481][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:58:29,976][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:58:30,471][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:58:30,968][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:58:31,463][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:58:31,961][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:58:32,456][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:58:32,954][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:58:33,465][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:58:33,966][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:58:34,466][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:58:34,964][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:58:35,466][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:58:35,963][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:58:36,463][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:58:36,982][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:58:37,484][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:58:37,986][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:58:38,488][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:58:38,986][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:58:39,486][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:58:39,987][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:58:40,498][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:58:40,998][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:58:41,496][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:58:41,998][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:58:42,498][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:58:43,001][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:58:43,500][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:58:44,009][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:58:44,511][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:58:45,013][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:58:45,509][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:58:46,008][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:58:46,505][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:58:47,002][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:58:47,500][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:58:48,007][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:58:48,507][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:58:49,004][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:58:49,501][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:58:50,000][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:58:50,497][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:58:50,994][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:58:51,491][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:58:51,987][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:58:52,485][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:58:52,984][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:58:53,485][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:58:53,983][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:58:54,482][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:58:54,983][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:58:55,483][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:58:55,982][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:58:56,478][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:58:56,990][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:58:57,490][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:58:57,988][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:58:58,487][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:58:58,986][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:58:59,486][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:58:59,985][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:59:00,486][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10212 tokens. +[2026-03-26 01:59:01,203][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 60.61%, Block Peak % of device VRAM: 62.23%, ΔTime: 00:00:32 +[2026-03-26 01:59:01,973][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:59:01,975][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:59:01,977][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:59:02,732][__main__][INFO] - Iteration 326 took 51s (30.04% Gen, 68.48% Train). Generation: 15s, Training: 34s. Estimated remaining time: 37h 38m 26s. Estimated total time: 42h 31m 48s. Time estimates for 10 more iterations: 8m 30s, 100 more iterations: 1h 25m 3s, 500 more iterations: 7h 5m 18s. +[2026-03-26 01:59:02,734][__main__][INFO] - Starting iteration 326. +[2026-03-26 01:59:03,133][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 32 and human policies 1. +[2026-03-26 01:59:03,134][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 01:59:15,602][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 01:59:21,717][__main__][INFO] - Number of regex retries in iteration 326: 1 +[2026-03-26 01:59:21,718][__main__][INFO] - agents played in iteration 326 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 01:59:22,476][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:59:22,499][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:59:22,523][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:59:22,546][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 01:59:22,546][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 01:59:22,547][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 01:59:23,223][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 01:59:23,680][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 01:59:24,194][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 01:59:24,692][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 01:59:25,189][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 01:59:25,690][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 01:59:26,188][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 01:59:26,689][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 01:59:27,189][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 01:59:27,689][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 01:59:28,189][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 01:59:28,689][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 01:59:29,189][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 01:59:29,690][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 01:59:30,188][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 01:59:30,685][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 01:59:31,183][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 01:59:31,682][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 01:59:32,183][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 01:59:32,680][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 01:59:33,186][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 01:59:33,687][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 01:59:34,189][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 01:59:34,688][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 01:59:35,189][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 01:59:35,693][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 01:59:36,193][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 01:59:36,694][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 01:59:37,193][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 01:59:37,692][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 01:59:38,192][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 01:59:38,691][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 01:59:39,189][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 01:59:39,687][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 01:59:40,185][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 01:59:40,683][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 01:59:41,184][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 01:59:41,680][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 01:59:42,176][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 01:59:42,673][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 01:59:43,173][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 01:59:43,671][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 01:59:44,166][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 01:59:44,665][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 01:59:45,163][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 01:59:45,661][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 01:59:46,160][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 01:59:46,659][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 01:59:47,157][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 01:59:47,655][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 01:59:48,153][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 01:59:48,647][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 01:59:49,143][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 01:59:49,643][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 01:59:50,138][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 01:59:50,632][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 01:59:51,129][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 01:59:51,625][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 01:59:52,119][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 01:59:52,615][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 01:59:53,110][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 01:59:53,606][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 01:59:54,102][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 01:59:54,597][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 01:59:55,092][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10311 tokens. +[2026-03-26 01:59:55,805][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 60.54%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 +[2026-03-26 01:59:56,566][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 01:59:56,568][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 01:59:56,599][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 01:59:57,312][__main__][INFO] - Iteration 327 took 54s (34.30% Gen, 64.38% Train). Generation: 18s, Training: 34s. Estimated remaining time: 40h 14m 40s. Estimated total time: 45h 8m 57s. Time estimates for 10 more iterations: 9m 1s, 100 more iterations: 1h 30m 17s, 500 more iterations: 7h 31m 29s. +[2026-03-26 01:59:57,314][__main__][INFO] - Starting iteration 327. +[2026-03-26 01:59:57,714][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 32 and human policies 1. +[2026-03-26 01:59:57,715][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:00:17,677][__main__][INFO] - Number of regex retries in iteration 327: 0 +[2026-03-26 02:00:17,678][__main__][INFO] - agents played in iteration 327 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:00:18,446][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:00:18,470][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:00:18,493][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:00:18,516][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:00:18,516][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:00:18,517][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:00:19,190][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:00:19,660][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:00:20,162][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:00:20,662][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:00:21,161][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:00:21,658][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:00:22,160][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:00:22,659][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:00:23,157][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:00:23,653][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:00:24,152][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:00:24,649][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:00:25,146][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:00:25,647][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:00:26,143][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:00:26,639][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:00:27,136][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:00:27,637][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:00:28,139][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:00:28,636][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:00:29,134][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:00:29,631][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:00:30,132][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:00:30,628][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:00:31,125][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:00:31,622][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:00:32,124][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:00:32,624][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:00:33,121][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:00:33,621][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:00:34,117][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:00:34,623][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:00:35,124][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:00:35,621][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:00:36,119][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:00:36,617][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:00:37,117][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:00:37,616][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:00:38,115][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:00:38,615][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:00:39,114][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:00:39,609][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:00:40,105][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:00:40,603][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:00:41,102][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:00:41,601][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:00:42,100][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:00:42,598][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:00:43,093][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:00:43,588][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:00:44,087][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:00:44,586][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:00:45,082][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:00:45,581][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:00:46,078][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:00:46,577][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:00:47,075][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:00:47,573][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:00:48,074][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:00:48,572][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:00:49,068][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:00:49,567][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:00:50,068][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:00:50,563][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:00:51,061][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10379 tokens. +[2026-03-26 02:00:51,773][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 60.67%, Block Peak % of device VRAM: 62.28%, ΔTime: 00:00:32 +[2026-03-26 02:00:52,534][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:00:52,536][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:00:52,538][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:00:53,264][__main__][INFO] - Iteration 328 took 55s (35.94% Gen, 62.75% Train). Generation: 19s, Training: 34s. Estimated remaining time: 41h 22m 19s. Estimated total time: 46h 17m 32s. Time estimates for 10 more iterations: 9m 15s, 100 more iterations: 1h 32m 35s, 500 more iterations: 7h 42m 55s. +[2026-03-26 02:00:53,268][__main__][INFO] - Starting iteration 328. +[2026-03-26 02:00:53,672][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 32 and human policies 1. +[2026-03-26 02:00:53,672][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:01:15,222][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 02:01:15,852][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 02:01:22,307][__main__][INFO] - Number of regex retries in iteration 328: 2 +[2026-03-26 02:01:22,308][__main__][INFO] - agents played in iteration 328 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:01:23,073][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:01:23,097][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:01:23,120][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:01:23,143][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:01:23,144][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:01:23,144][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:01:23,815][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:01:24,266][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:01:24,767][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:01:25,265][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:01:25,767][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:01:26,263][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:01:26,760][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:01:27,259][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:01:27,756][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:01:28,252][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:01:28,748][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:01:29,245][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:01:29,744][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:01:30,241][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:01:30,736][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:01:31,235][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:01:31,731][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:01:32,235][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:01:32,731][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:01:33,227][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:01:33,725][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:01:34,223][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:01:34,721][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:01:35,221][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:01:35,721][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:01:36,257][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:01:36,760][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:01:37,264][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:01:37,768][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:01:38,268][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:01:38,770][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:01:39,269][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:01:39,772][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:01:40,271][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:01:40,771][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:01:41,271][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:01:41,770][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:01:42,270][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:01:42,769][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:01:43,270][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:01:43,769][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:01:44,270][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:01:44,767][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:01:45,266][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:01:45,763][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:01:46,265][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:01:46,766][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:01:47,265][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:01:47,764][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:01:48,260][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:01:48,760][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:01:49,256][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:01:49,757][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:01:50,253][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:01:50,750][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:01:51,248][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:01:51,745][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:01:52,246][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:01:52,745][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:01:53,243][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:01:53,750][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:01:54,250][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:01:54,750][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:01:55,249][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:01:55,750][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10387 tokens. +[2026-03-26 02:01:56,461][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.27%, Current % of VRAM taken: 60.69%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 +[2026-03-26 02:01:57,244][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:01:57,247][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:01:57,248][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:01:57,969][__main__][INFO] - Iteration 329 took 1m 4s (44.53% Gen, 54.34% Train). Generation: 28s, Training: 34s. Estimated remaining time: 48h 38m 35s. Estimated total time: 53h 34m 53s. Time estimates for 10 more iterations: 10m 42s, 100 more iterations: 1h 47m 9s, 500 more iterations: 8h 55m 48s. +[2026-03-26 02:01:57,971][__main__][INFO] - Starting iteration 329. +[2026-03-26 02:01:58,370][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 32 and human policies 1. +[2026-03-26 02:01:58,370][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:02:18,067][__main__][INFO] - Number of regex retries in iteration 329: 0 +[2026-03-26 02:02:18,068][__main__][INFO] - agents played in iteration 329 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:02:18,830][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:02:18,854][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:02:18,877][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:02:18,900][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:02:18,900][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:02:18,901][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:02:19,577][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:02:20,030][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:02:20,535][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:02:21,032][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:02:21,528][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:02:22,029][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:02:22,525][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:02:23,021][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:02:23,519][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:02:24,019][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:02:24,515][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:02:25,014][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:02:25,513][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:02:26,011][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:02:26,509][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:02:27,005][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:02:27,501][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:02:27,997][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:02:28,494][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:02:28,991][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:02:29,488][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:02:29,985][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:02:30,481][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:02:30,978][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:02:31,478][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:02:31,973][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:02:32,468][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:02:32,965][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:02:33,462][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:02:33,957][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:02:34,463][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:02:34,963][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:02:35,459][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:02:35,963][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:02:36,467][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:02:36,968][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:02:37,464][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:02:37,961][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:02:38,463][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:02:38,962][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:02:39,459][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:02:39,960][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:02:40,455][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:02:40,952][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:02:41,447][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:02:41,946][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:02:42,444][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:02:42,938][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:02:43,433][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:02:43,932][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:02:44,428][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:02:44,925][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:02:45,425][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:02:45,922][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:02:46,419][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:02:46,915][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:02:47,413][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:02:47,911][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:02:48,407][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:02:48,905][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:02:49,404][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:02:49,902][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:02:50,402][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:02:50,900][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:02:51,396][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10302 tokens. +[2026-03-26 02:02:52,112][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 60.54%, Block Peak % of device VRAM: 62.13%, ΔTime: 00:00:32 +[2026-03-26 02:02:52,873][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:02:52,876][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:02:52,877][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:02:53,602][__main__][INFO] - Iteration 330 took 55s (35.66% Gen, 63.02% Train). Generation: 19s, Training: 34s. Estimated remaining time: 41h 4m 26s. Estimated total time: 46h 1m 39s. Time estimates for 10 more iterations: 9m 12s, 100 more iterations: 1h 32m 3s, 500 more iterations: 7h 40m 16s. +[2026-03-26 02:02:53,604][__main__][INFO] - Starting iteration 330. +[2026-03-26 02:02:54,004][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 32 and human policies 1. +[2026-03-26 02:02:54,005][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:03:14,163][__main__][INFO] - Number of regex retries in iteration 330: 0 +[2026-03-26 02:03:14,164][__main__][INFO] - agents played in iteration 330 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:03:14,924][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:03:14,948][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:03:14,971][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:03:14,995][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:03:14,996][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:03:14,996][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:03:15,674][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:03:16,131][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:03:16,631][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:03:17,128][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:03:17,624][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:03:18,123][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:03:18,620][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:03:19,119][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:03:19,626][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:03:20,127][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:03:20,631][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:03:21,137][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:03:21,636][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:03:22,137][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:03:22,641][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:03:23,143][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:03:23,646][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:03:24,146][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:03:24,648][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:03:25,150][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:03:25,649][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:03:26,155][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:03:26,655][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:03:27,157][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:03:27,656][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:03:28,155][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:03:28,654][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:03:29,151][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:03:29,651][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:03:30,159][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:03:30,657][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:03:31,157][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:03:31,657][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:03:32,157][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:03:32,651][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:03:33,147][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:03:33,643][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:03:34,142][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:03:34,638][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:03:35,135][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:03:35,631][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:03:36,128][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:03:36,628][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:03:37,128][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:03:37,621][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:03:38,116][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:03:38,612][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:03:39,110][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:03:39,610][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:03:40,107][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:03:40,603][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:03:41,100][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:03:41,597][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:03:42,092][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:03:42,588][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:03:43,085][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:03:43,582][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:03:44,082][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:03:44,582][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:03:45,081][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:03:45,580][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:03:46,078][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:03:46,578][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:03:47,077][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:03:47,577][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10244 tokens. +[2026-03-26 02:03:48,289][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.37%, ΔTime: 00:00:32 +[2026-03-26 02:03:49,060][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:03:49,063][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:03:49,065][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:03:50,546][__main__][INFO] - Iteration 331 took 56s (35.65% Gen, 61.72% Train). Generation: 20s, Training: 34s. Estimated remaining time: 42h 8m 57s. Estimated total time: 47h 7m 7s. Time estimates for 10 more iterations: 9m 25s, 100 more iterations: 1h 34m 14s, 500 more iterations: 7h 51m 11s. +[2026-03-26 02:03:50,549][__main__][INFO] - Starting iteration 331. +[2026-03-26 02:03:50,949][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 33 and human policies 1. +[2026-03-26 02:03:50,949][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:03:57,397][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 02:04:08,931][__main__][INFO] - Number of regex retries in iteration 331: 1 +[2026-03-26 02:04:08,932][__main__][INFO] - agents played in iteration 331 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:04:09,696][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:04:09,720][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:04:09,743][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:04:09,767][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:04:09,768][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:04:09,768][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:04:10,452][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:04:10,907][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:04:11,411][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:04:11,912][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:04:12,413][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:04:12,911][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:04:13,419][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:04:13,920][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:04:14,418][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:04:14,920][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:04:15,417][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:04:15,918][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:04:16,415][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:04:16,915][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:04:17,415][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:04:17,914][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:04:18,413][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:04:18,911][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:04:19,409][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:04:19,906][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:04:20,407][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:04:20,905][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:04:21,405][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:04:21,903][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:04:22,410][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:04:22,909][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:04:23,406][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:04:23,906][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:04:24,404][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:04:24,906][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:04:25,404][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:04:25,903][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:04:26,400][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:04:26,897][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:04:27,394][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:04:27,892][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:04:28,389][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:04:28,890][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:04:29,387][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:04:29,885][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:04:30,385][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:04:30,884][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:04:31,380][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:04:31,883][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:04:32,383][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:04:32,879][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:04:33,376][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:04:33,874][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:04:34,373][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:04:34,869][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:04:35,369][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:04:35,867][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:04:36,370][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:04:36,868][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:04:37,365][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:04:37,865][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:04:38,362][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:04:38,856][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:04:39,352][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:04:39,847][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:04:40,343][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:04:40,840][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:04:41,336][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:04:41,832][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:04:42,329][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10286 tokens. +[2026-03-26 02:04:43,045][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 60.54%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:32 +[2026-03-26 02:04:43,789][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:04:43,791][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:04:43,792][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:04:44,631][__main__][INFO] - Iteration 332 took 53s (33.50% Gen, 64.94% Train). Generation: 17s, Training: 34s. Estimated remaining time: 39h 45m 5s. Estimated total time: 44h 44m 9s. Time estimates for 10 more iterations: 8m 56s, 100 more iterations: 1h 29m 28s, 500 more iterations: 7h 27m 21s. +[2026-03-26 02:04:44,633][__main__][INFO] - Starting iteration 332. +[2026-03-26 02:04:45,032][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 33 and human policies 1. +[2026-03-26 02:04:45,033][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:05:02,117][__main__][INFO] - Number of regex retries in iteration 332: 0 +[2026-03-26 02:05:02,118][__main__][INFO] - agents played in iteration 332 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:05:02,876][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:05:02,899][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:05:02,923][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:05:02,946][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:05:02,947][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:05:02,947][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:05:03,624][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:05:04,077][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:05:04,583][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:05:05,089][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:05:05,592][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:05:06,089][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:05:06,587][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:05:07,084][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:05:07,582][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:05:08,083][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:05:08,580][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:05:09,077][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:05:09,577][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:05:10,073][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:05:10,570][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:05:11,067][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:05:11,563][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:05:12,060][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:05:12,555][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:05:13,053][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:05:13,548][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:05:14,043][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:05:14,540][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:05:15,034][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:05:15,527][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:05:16,022][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:05:16,516][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:05:17,013][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:05:17,510][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:05:18,004][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:05:18,503][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:05:19,000][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:05:19,494][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:05:19,993][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:05:20,489][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:05:20,988][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:05:21,485][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:05:21,984][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:05:22,481][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:05:22,977][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:05:23,476][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:05:23,970][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:05:24,471][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:05:24,970][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:05:25,466][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:05:25,971][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:05:26,468][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:05:26,969][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:05:27,468][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:05:27,964][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:05:28,460][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:05:28,956][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:05:29,453][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:05:29,948][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:05:30,444][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:05:30,944][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:05:31,438][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:05:31,935][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:05:32,432][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:05:32,931][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:05:33,427][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:05:33,923][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:05:34,422][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:05:34,918][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:05:35,418][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10232 tokens. +[2026-03-26 02:05:36,127][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 60.60%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:32 +[2026-03-26 02:05:36,887][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:05:36,889][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:05:36,891][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:05:37,640][__main__][INFO] - Iteration 333 took 52s (32.48% Gen, 66.10% Train). Generation: 17s, Training: 34s. Estimated remaining time: 38h 50m 27s. Estimated total time: 43h 50m 24s. Time estimates for 10 more iterations: 8m 46s, 100 more iterations: 1h 27m 40s, 500 more iterations: 7h 18m 24s. +[2026-03-26 02:05:37,642][__main__][INFO] - Starting iteration 333. +[2026-03-26 02:05:38,042][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 33 and human policies 1. +[2026-03-26 02:05:38,043][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:05:44,318][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 02:05:57,162][__main__][INFO] - Number of regex retries in iteration 333: 1 +[2026-03-26 02:05:57,162][__main__][INFO] - agents played in iteration 333 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:05:58,044][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:05:58,067][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:05:58,091][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:05:58,115][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:05:58,115][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:05:58,116][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:05:58,794][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:05:59,245][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:05:59,755][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:06:00,256][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:06:00,757][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:06:01,257][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:06:01,760][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:06:02,262][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:06:02,764][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:06:03,264][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:06:03,769][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:06:04,271][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:06:04,769][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:06:05,266][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:06:05,767][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:06:06,269][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:06:06,776][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:06:07,275][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:06:07,776][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:06:08,272][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:06:08,769][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:06:09,267][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:06:09,764][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:06:10,264][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:06:10,761][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:06:11,261][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:06:11,762][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:06:12,263][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:06:12,763][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:06:13,262][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:06:13,761][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:06:14,259][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:06:14,759][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:06:15,259][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:06:15,754][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:06:16,253][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:06:16,750][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:06:17,250][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:06:17,748][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:06:18,247][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:06:18,746][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:06:19,245][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:06:19,744][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:06:20,241][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:06:20,741][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:06:21,239][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:06:21,738][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:06:22,239][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:06:22,739][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:06:23,233][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:06:23,731][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:06:24,227][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:06:24,860][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:06:27,299][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:06:27,800][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:06:28,297][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:06:28,794][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:06:29,292][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:06:29,792][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:06:30,288][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:06:30,786][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:06:31,284][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:06:31,782][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:06:32,281][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:06:32,777][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10391 tokens. +[2026-03-26 02:06:33,560][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 60.49%, Block Peak % of device VRAM: 62.39%, ΔTime: 00:00:34 +[2026-03-26 02:06:34,322][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:06:34,324][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:06:34,326][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:06:35,081][__main__][INFO] - Iteration 334 took 57s (33.52% Gen, 65.15% Train). Generation: 19s, Training: 37s. Estimated remaining time: 42h 31m 4s. Estimated total time: 47h 31m 59s. Time estimates for 10 more iterations: 9m 30s, 100 more iterations: 1h 35m 3s, 500 more iterations: 7h 55m 19s. +[2026-03-26 02:06:35,089][__main__][INFO] - Starting iteration 334. +[2026-03-26 02:06:35,494][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 33 and human policies 1. +[2026-03-26 02:06:35,494][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:07:11,709][__main__][INFO] - Number of regex retries in iteration 334: 0 +[2026-03-26 02:07:11,710][__main__][INFO] - agents played in iteration 334 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:07:12,835][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:07:12,859][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:07:12,883][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:07:12,907][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:07:12,908][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:07:12,908][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:07:13,696][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:07:14,151][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:07:14,654][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:07:15,151][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:07:15,654][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:07:16,153][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:07:16,650][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:07:17,148][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:07:17,645][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:07:18,153][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:07:18,657][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:07:19,159][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:07:19,655][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:07:20,155][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:07:20,659][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:07:21,162][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:07:21,658][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:07:22,158][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:07:22,655][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:07:23,163][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:07:23,662][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:07:24,162][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:07:24,660][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:07:25,158][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:07:25,655][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:07:26,152][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:07:26,658][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:07:27,164][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:07:27,664][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:07:28,165][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:07:28,665][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:07:29,161][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:07:29,664][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:07:30,163][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:07:30,665][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:07:31,173][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:07:31,672][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:07:32,176][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:07:32,676][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:07:33,176][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:07:33,675][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:07:34,175][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:07:34,675][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:07:35,175][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:07:35,674][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:07:36,173][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:07:36,687][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:07:37,189][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:07:37,687][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:07:38,186][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:07:38,685][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:07:39,183][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:07:39,684][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:07:40,180][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:07:40,681][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:07:41,181][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:07:41,678][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:07:42,176][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:07:42,673][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:07:43,173][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:07:43,672][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:07:44,171][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:07:44,670][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:07:45,171][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:07:45,678][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10363 tokens. +[2026-03-26 02:07:46,417][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 60.65%, Block Peak % of device VRAM: 62.23%, ΔTime: 00:00:32 +[2026-03-26 02:07:47,187][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:07:47,189][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:07:47,191][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:07:47,956][__main__][INFO] - Iteration 335 took 1m 12s (49.98% Gen, 48.96% Train). Generation: 36s, Training: 35s. Estimated remaining time: 55h 21m 1s. Estimated total time: 60h 23m 9s. Time estimates for 10 more iterations: 12m 4s, 100 more iterations: 2h 0m 46s, 500 more iterations: 10h 3m 51s. +[2026-03-26 02:07:47,958][__main__][INFO] - Starting iteration 335. +[2026-03-26 02:07:48,359][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 33 and human policies 1. +[2026-03-26 02:07:48,360][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:08:12,912][__main__][INFO] - Number of regex retries in iteration 335: 0 +[2026-03-26 02:08:12,913][__main__][INFO] - agents played in iteration 335 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:08:13,681][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:08:13,705][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:08:13,728][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:08:13,751][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:08:13,752][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:08:13,753][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:08:14,451][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:08:14,909][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:08:15,414][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:08:15,912][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:08:16,410][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:08:16,907][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:08:17,403][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:08:17,900][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:08:18,397][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:08:18,896][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:08:19,393][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:08:19,891][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:08:20,394][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:08:20,891][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:08:21,389][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:08:21,887][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:08:22,385][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:08:22,885][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:08:23,384][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:08:23,882][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:08:24,380][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:08:24,877][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:08:25,377][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:08:25,893][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:08:26,395][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:08:26,894][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:08:27,395][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:08:27,891][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:08:28,389][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:08:28,888][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:08:29,390][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:08:29,891][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:08:30,392][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:08:30,889][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:08:31,389][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:08:31,889][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:08:32,387][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:08:32,885][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:08:33,385][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:08:33,884][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:08:34,386][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:08:34,888][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:08:35,386][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:08:35,885][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:08:36,384][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:08:36,884][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:08:37,384][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:08:37,883][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:08:38,383][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:08:38,880][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:08:39,377][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:08:39,873][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:08:40,371][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:08:40,867][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:08:41,365][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:08:41,864][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:08:42,361][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:08:42,857][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:08:43,356][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:08:43,850][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:08:44,346][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:08:44,842][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:08:45,343][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:08:45,838][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:08:46,334][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10231 tokens. +[2026-03-26 02:08:47,054][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.09%, Current % of VRAM taken: 60.52%, Block Peak % of device VRAM: 62.22%, ΔTime: 00:00:32 +[2026-03-26 02:08:47,820][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:08:47,822][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:08:47,823][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:08:48,582][__main__][INFO] - Iteration 336 took 1m 0s (40.77% Gen, 57.97% Train). Generation: 24s, Training: 34s. Estimated remaining time: 45h 8m 0s. Estimated total time: 50h 11m 9s. Time estimates for 10 more iterations: 10m 2s, 100 more iterations: 1h 40m 22s, 500 more iterations: 8h 21m 51s. +[2026-03-26 02:08:48,584][__main__][INFO] - Starting iteration 336. +[2026-03-26 02:08:49,051][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 33 and human policies 1. +[2026-03-26 02:08:49,052][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:09:05,315][__main__][INFO] - Number of regex retries in iteration 336: 0 +[2026-03-26 02:09:05,316][__main__][INFO] - agents played in iteration 336 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:09:06,079][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:09:06,103][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:09:06,126][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:09:06,150][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:09:06,151][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:09:06,151][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:09:06,836][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:09:07,288][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:09:07,801][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:09:08,299][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:09:08,797][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:09:09,295][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:09:09,797][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:09:10,294][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:09:10,791][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:09:11,289][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:09:11,788][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:09:12,287][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:09:12,796][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:09:13,294][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:09:13,799][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:09:14,297][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:09:14,794][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:09:15,292][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:09:15,790][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:09:16,290][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:09:16,788][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:09:17,291][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:09:17,791][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:09:18,288][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:09:18,787][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:09:19,286][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:09:19,785][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:09:20,284][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:09:20,783][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:09:21,283][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:09:21,782][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:09:22,278][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:09:22,774][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:09:23,270][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:09:23,768][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:09:24,266][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:09:24,764][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:09:25,263][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:09:25,764][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:09:26,260][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:09:26,779][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:09:27,281][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:09:27,782][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:09:28,281][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:09:28,779][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:09:29,277][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:09:29,775][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:09:30,272][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:09:30,769][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:09:31,271][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:09:31,771][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:09:32,272][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:09:32,771][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:09:33,270][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:09:33,769][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:09:34,267][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:09:34,767][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:09:35,265][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:09:35,763][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:09:36,263][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:09:36,763][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:09:37,261][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:09:37,756][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:09:38,251][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:09:38,748][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10341 tokens. +[2026-03-26 02:09:39,457][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.58%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 +[2026-03-26 02:09:40,225][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:09:40,227][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:09:40,229][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:09:40,952][__main__][INFO] - Iteration 337 took 51s (31.34% Gen, 67.27% Train). Generation: 16s, Training: 34s. Estimated remaining time: 38h 11m 2s. Estimated total time: 43h 15m 2s. Time estimates for 10 more iterations: 8m 39s, 100 more iterations: 1h 26m 30s, 500 more iterations: 7h 12m 30s. +[2026-03-26 02:09:40,954][__main__][INFO] - Starting iteration 337. +[2026-03-26 02:09:41,351][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 33 and human policies 1. +[2026-03-26 02:09:41,352][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:09:57,245][__main__][INFO] - Number of regex retries in iteration 337: 0 +[2026-03-26 02:09:57,247][__main__][INFO] - agents played in iteration 337 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:09:58,027][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:09:58,050][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:09:58,074][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:09:58,098][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:09:58,098][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:09:58,099][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:09:58,788][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:09:59,247][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:09:59,753][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:10:00,253][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:10:00,751][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:10:01,250][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:10:01,749][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:10:02,248][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:10:02,752][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:10:03,250][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:10:03,748][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:10:04,246][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:10:04,744][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:10:05,240][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:10:05,736][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:10:06,234][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:10:06,739][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:10:07,237][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:10:07,735][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:10:08,238][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:10:08,736][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:10:09,234][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:10:09,730][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:10:10,228][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:10:10,724][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:10:11,222][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:10:11,728][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:10:12,227][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:10:12,726][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:10:13,225][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:10:13,722][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:10:14,219][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:10:14,716][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:10:15,212][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:10:15,709][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:10:16,206][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:10:16,705][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:10:17,202][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:10:17,698][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:10:18,195][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:10:18,690][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:10:19,188][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:10:19,685][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:10:20,182][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:10:20,684][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:10:21,184][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:10:21,711][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:10:22,216][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:10:22,712][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:10:23,210][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:10:23,707][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:10:24,204][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:10:24,701][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:10:25,199][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:10:25,695][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:10:26,191][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:10:26,688][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:10:27,186][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:10:27,685][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:10:28,183][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:10:28,681][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:10:29,179][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:10:29,678][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:10:30,180][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:10:30,677][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10092 tokens. +[2026-03-26 02:10:31,411][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.13%, ΔTime: 00:00:32 +[2026-03-26 02:10:32,183][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:10:32,185][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:10:32,187][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:10:32,904][__main__][INFO] - Iteration 338 took 51s (30.83% Gen, 67.78% Train). Generation: 15s, Training: 34s. Estimated remaining time: 37h 52m 46s. Estimated total time: 42h 57m 38s. Time estimates for 10 more iterations: 8m 35s, 100 more iterations: 1h 25m 55s, 500 more iterations: 7h 9m 36s. +[2026-03-26 02:10:32,906][__main__][INFO] - Starting iteration 338. +[2026-03-26 02:10:33,303][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 33 and human policies 1. +[2026-03-26 02:10:33,304][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:10:53,770][__main__][INFO] - Number of regex retries in iteration 338: 0 +[2026-03-26 02:10:53,771][__main__][INFO] - agents played in iteration 338 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:10:54,538][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:10:54,562][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:10:54,589][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:10:54,616][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:10:54,617][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:10:54,618][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:10:55,319][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:10:55,776][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:10:56,280][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:10:56,777][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:10:57,274][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:10:57,782][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:10:58,279][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:10:58,778][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:10:59,274][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:10:59,772][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:11:00,269][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:11:00,765][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:11:01,261][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:11:01,771][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:11:02,271][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:11:02,769][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:11:03,267][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:11:03,764][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:11:04,266][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:11:04,760][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:11:05,258][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:11:05,754][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:11:06,249][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:11:06,746][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:11:07,254][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:11:07,751][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:11:08,248][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:11:08,747][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:11:09,245][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:11:09,744][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:11:10,242][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:11:10,740][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:11:11,246][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:11:11,744][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:11:12,243][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:11:12,744][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:11:13,241][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:11:13,766][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:11:14,272][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:11:14,770][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:11:15,270][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:11:15,768][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:11:16,266][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:11:16,763][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:11:17,260][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:11:17,761][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:11:18,257][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:11:18,757][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:11:19,253][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:11:19,749][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:11:20,246][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:11:20,743][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:11:21,240][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:11:21,735][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:11:22,231][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:11:22,727][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:11:23,233][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:11:23,734][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:11:24,236][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:11:24,739][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:11:25,238][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:11:25,740][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:11:26,240][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:11:26,739][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:11:27,240][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10102 tokens. +[2026-03-26 02:11:27,961][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.26%, Current % of VRAM taken: 60.69%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 +[2026-03-26 02:11:28,724][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:11:28,726][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:11:28,728][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:11:29,491][__main__][INFO] - Iteration 339 took 56s (36.43% Gen, 62.21% Train). Generation: 20s, Training: 34s. Estimated remaining time: 41h 43m 37s. Estimated total time: 46h 49m 26s. Time estimates for 10 more iterations: 9m 21s, 100 more iterations: 1h 33m 38s, 500 more iterations: 7h 48m 14s. +[2026-03-26 02:11:29,493][__main__][INFO] - Starting iteration 339. +[2026-03-26 02:11:29,894][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 33 and human policies 1. +[2026-03-26 02:11:29,895][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:11:36,031][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 02:11:47,083][__main__][INFO] - Number of regex retries in iteration 339: 1 +[2026-03-26 02:11:47,084][__main__][INFO] - agents played in iteration 339 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:11:47,843][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:11:47,867][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:11:47,890][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:11:47,913][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:11:47,914][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:11:47,914][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:11:48,595][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:11:49,049][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:11:49,550][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:11:50,047][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:11:50,543][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:11:51,044][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:11:51,538][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:11:52,033][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:11:52,536][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:11:53,034][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:11:53,532][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:11:54,029][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:11:54,528][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:11:55,047][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:11:55,547][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:11:56,046][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:11:56,544][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:11:57,044][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:11:57,540][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:11:58,038][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:11:58,534][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:11:59,040][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:11:59,539][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:12:00,039][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:12:00,537][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:12:01,035][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:12:01,532][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:12:02,032][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:12:02,532][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:12:03,031][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:12:03,529][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:12:04,027][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:12:04,527][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:12:05,023][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:12:05,519][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:12:06,015][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:12:06,514][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:12:07,011][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:12:07,508][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:12:08,005][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:12:08,501][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:12:09,002][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:12:09,499][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:12:09,996][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:12:10,492][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:12:10,989][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:12:11,489][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:12:11,987][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:12:12,486][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:12:12,990][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:12:13,488][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:12:13,987][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:12:14,486][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:12:14,983][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:12:15,482][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:12:15,983][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:12:16,480][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:12:16,981][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:12:17,483][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:12:17,979][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:12:18,478][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:12:18,977][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:12:19,474][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:12:19,972][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:12:20,471][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10216 tokens. +[2026-03-26 02:12:21,184][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.58%, Block Peak % of device VRAM: 62.14%, ΔTime: 00:00:32 +[2026-03-26 02:12:21,952][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:12:21,955][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:12:21,956][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:12:22,684][__main__][INFO] - Iteration 340 took 52s (32.56% Gen, 66.06% Train). Generation: 17s, Training: 34s. Estimated remaining time: 38h 52m 49s. Estimated total time: 43h 59m 31s. Time estimates for 10 more iterations: 8m 47s, 100 more iterations: 1h 27m 59s, 500 more iterations: 7h 19m 55s. +[2026-03-26 02:12:22,687][__main__][INFO] - Starting iteration 340. +[2026-03-26 02:12:23,085][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 33 and human policies 1. +[2026-03-26 02:12:23,086][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:12:32,549][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 02:12:45,245][__main__][INFO] - Number of regex retries in iteration 340: 1 +[2026-03-26 02:12:45,246][__main__][INFO] - agents played in iteration 340 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:12:46,011][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:12:46,034][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:12:46,058][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:12:46,081][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:12:46,082][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:12:46,082][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:12:46,766][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:12:47,229][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:12:47,735][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:12:48,235][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:12:48,734][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:12:49,232][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:12:49,732][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:12:50,239][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:12:50,741][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:12:51,240][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:12:51,740][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:12:52,244][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:12:52,746][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:12:53,257][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:12:53,757][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:12:54,260][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:12:54,762][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:12:55,260][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:12:55,757][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:12:56,255][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:12:56,752][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:12:57,260][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:12:57,758][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:12:58,260][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:12:58,758][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:12:59,254][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:12:59,753][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:13:00,252][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:13:00,751][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:13:01,250][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:13:01,746][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:13:02,257][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:13:02,753][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:13:03,251][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:13:03,750][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:13:04,251][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:13:04,749][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:13:05,247][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:13:05,745][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:13:06,246][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:13:06,744][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:13:07,243][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:13:07,744][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:13:08,241][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:13:08,746][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:13:09,247][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:13:09,746][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:13:10,247][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:13:10,747][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:13:11,246][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:13:11,743][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:13:12,242][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:13:12,737][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:13:13,238][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:13:13,734][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:13:14,230][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:13:14,730][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:13:15,228][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:13:15,725][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:13:16,220][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:13:16,715][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:13:17,210][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:13:17,706][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:13:18,205][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:13:18,702][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10308 tokens. +[2026-03-26 02:13:19,427][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.09%, Current % of VRAM taken: 60.51%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 +[2026-03-26 02:13:20,236][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:13:20,239][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:13:20,240][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:13:21,711][__main__][INFO] - Iteration 341 took 58s (37.80% Gen, 59.69% Train). Generation: 22s, Training: 34s. Estimated remaining time: 43h 43m 39s. Estimated total time: 48h 51m 21s. Time estimates for 10 more iterations: 9m 46s, 100 more iterations: 1h 37m 42s, 500 more iterations: 8h 8m 33s. +[2026-03-26 02:13:21,714][__main__][INFO] - Starting iteration 341. +[2026-03-26 02:13:22,111][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 34 and human policies 1. +[2026-03-26 02:13:22,112][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:13:39,269][__main__][INFO] - Number of regex retries in iteration 341: 0 +[2026-03-26 02:13:39,270][__main__][INFO] - agents played in iteration 341 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:13:40,029][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:13:40,053][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:13:40,076][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:13:40,099][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:13:40,100][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:13:40,100][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:13:40,784][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:13:41,246][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:13:41,751][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:13:42,249][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:13:42,746][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:13:43,257][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:13:43,754][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:13:44,251][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:13:44,752][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:13:45,252][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:13:45,750][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:13:46,248][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:13:46,746][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:13:47,244][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:13:47,744][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:13:48,241][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:13:48,739][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:13:49,236][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:13:49,739][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:13:50,235][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:13:50,732][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:13:51,232][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:13:51,729][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:13:52,229][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:13:52,727][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:13:53,227][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:13:53,722][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:13:54,219][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:13:54,716][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:13:55,212][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:13:55,711][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:13:56,208][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:13:56,705][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:13:57,200][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:13:57,695][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:13:58,191][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:13:58,687][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:13:59,183][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:13:59,678][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:14:00,174][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:14:00,669][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:14:01,166][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:14:01,666][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:14:02,167][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:14:02,664][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:14:03,161][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:14:03,659][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:14:04,156][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:14:04,655][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:14:05,151][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:14:05,650][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:14:06,146][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:14:06,642][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:14:07,138][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:14:07,635][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:14:08,132][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:14:08,628][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:14:09,125][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:14:09,624][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:14:10,120][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:14:10,619][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:14:11,117][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:14:11,612][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:14:12,111][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:14:12,607][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10212 tokens. +[2026-03-26 02:14:13,332][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 60.50%, Block Peak % of device VRAM: 62.12%, ΔTime: 00:00:32 +[2026-03-26 02:14:14,091][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:14:14,094][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:14:14,095][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:14:14,833][__main__][INFO] - Iteration 342 took 52s (32.54% Gen, 66.05% Train). Generation: 17s, Training: 34s. Estimated remaining time: 38h 47m 34s. Estimated total time: 43h 56m 9s. Time estimates for 10 more iterations: 8m 47s, 100 more iterations: 1h 27m 52s, 500 more iterations: 7h 19m 21s. +[2026-03-26 02:14:14,836][__main__][INFO] - Starting iteration 342. +[2026-03-26 02:14:15,237][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 34 and human policies 1. +[2026-03-26 02:14:15,238][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:14:19,610][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 02:14:19,770][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 02:14:31,134][__main__][INFO] - Number of regex retries in iteration 342: 2 +[2026-03-26 02:14:31,135][__main__][INFO] - agents played in iteration 342 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:14:31,893][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:14:31,916][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:14:31,940][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:14:31,963][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:14:31,964][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:14:31,965][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:14:32,644][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:14:33,107][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:14:33,612][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:14:34,116][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:14:34,619][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:14:35,121][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:14:35,626][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:14:36,128][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:14:36,627][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:14:37,127][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:14:37,628][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:14:38,134][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:14:38,634][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:14:39,132][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:14:39,629][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:14:40,127][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:14:40,625][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:14:41,122][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:14:41,627][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:14:42,126][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:14:42,623][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:14:43,120][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:14:43,617][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:14:44,113][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:14:44,610][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:14:45,106][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:14:45,601][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:14:46,098][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:14:46,593][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:14:47,089][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:14:47,584][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:14:48,077][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:14:48,571][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:14:49,067][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:14:49,564][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:14:50,070][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:14:50,568][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:14:51,064][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:14:51,560][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:14:52,057][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:14:52,553][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:14:53,050][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:14:53,548][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:14:54,048][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:14:54,548][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:14:55,046][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:14:55,550][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:14:56,048][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:14:56,544][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:14:57,039][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:14:57,534][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:14:58,030][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:14:58,526][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:14:59,021][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:14:59,516][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:15:00,012][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:15:00,507][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:15:01,007][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:15:01,506][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:15:02,014][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:15:02,514][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:15:03,013][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:15:03,512][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:15:04,008][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:15:04,508][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10070 tokens. +[2026-03-26 02:15:05,227][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.22%, ΔTime: 00:00:32 +[2026-03-26 02:15:05,986][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:15:05,989][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:15:05,990][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:15:06,736][__main__][INFO] - Iteration 343 took 51s (30.87% Gen, 67.68% Train). Generation: 15s, Training: 34s. Estimated remaining time: 37h 45m 32s. Estimated total time: 42h 54m 58s. Time estimates for 10 more iterations: 8m 34s, 100 more iterations: 1h 25m 49s, 500 more iterations: 7h 9m 9s. +[2026-03-26 02:15:06,738][__main__][INFO] - Starting iteration 343. +[2026-03-26 02:15:07,139][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 34 and human policies 1. +[2026-03-26 02:15:07,139][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:15:25,077][__main__][INFO] - Number of regex retries in iteration 343: 0 +[2026-03-26 02:15:25,078][__main__][INFO] - agents played in iteration 343 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:15:25,839][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:15:25,863][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:15:25,886][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:15:25,909][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:15:25,910][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:15:25,911][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:15:26,591][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:15:27,055][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:15:27,558][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:15:28,057][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:15:28,556][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:15:29,055][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:15:29,553][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:15:30,052][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:15:30,550][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:15:31,049][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:15:31,547][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:15:32,050][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:15:32,548][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:15:33,049][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:15:33,546][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:15:34,047][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:15:34,547][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:15:35,046][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:15:35,548][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:15:36,046][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:15:36,562][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:15:37,066][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:15:37,566][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:15:38,068][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:15:38,566][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:15:39,068][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:15:39,568][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:15:40,068][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:15:40,567][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:15:41,066][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:15:41,567][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:15:42,067][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:15:42,566][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:15:43,067][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:15:43,565][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:15:44,062][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:15:44,559][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:15:45,059][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:15:45,560][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:15:46,060][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:15:46,556][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:15:47,057][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:15:47,555][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:15:48,054][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:15:48,552][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:15:49,050][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:15:49,549][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:15:50,049][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:15:50,548][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:15:51,048][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:15:51,546][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:15:52,046][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:15:52,545][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:15:53,044][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:15:53,545][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:15:54,045][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:15:54,544][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:15:55,042][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:15:55,542][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:15:56,040][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:15:56,540][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:15:57,036][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:15:57,536][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:15:58,035][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:15:58,533][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10376 tokens. +[2026-03-26 02:15:59,248][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.24%, ΔTime: 00:00:32 +[2026-03-26 02:16:00,044][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:16:00,046][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:16:00,048][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:16:00,770][__main__][INFO] - Iteration 344 took 53s (33.45% Gen, 65.20% Train). Generation: 17s, Training: 34s. Estimated remaining time: 39h 31m 14s. Estimated total time: 44h 41m 35s. Time estimates for 10 more iterations: 8m 56s, 100 more iterations: 1h 29m 23s, 500 more iterations: 7h 26m 55s. +[2026-03-26 02:16:00,774][__main__][INFO] - Starting iteration 344. +[2026-03-26 02:16:01,172][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 34 and human policies 1. +[2026-03-26 02:16:01,173][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:16:17,706][__main__][INFO] - Number of regex retries in iteration 344: 0 +[2026-03-26 02:16:17,706][__main__][INFO] - agents played in iteration 344 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:16:18,461][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:16:18,485][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:16:18,508][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:16:18,532][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:16:18,532][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:16:18,533][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:16:19,210][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:16:19,662][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:16:20,164][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:16:20,659][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:16:21,156][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:16:21,651][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:16:22,147][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:16:22,642][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:16:23,135][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:16:23,630][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:16:24,126][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:16:24,621][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:16:25,123][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:16:25,623][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:16:26,119][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:16:26,614][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:16:27,111][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:16:27,606][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:16:28,101][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:16:28,597][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:16:29,092][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:16:29,588][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:16:30,095][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:16:30,597][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:16:31,094][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:16:31,592][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:16:32,093][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:16:32,590][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:16:33,088][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:16:33,586][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:16:34,084][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:16:34,580][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:16:35,077][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:16:35,575][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:16:36,070][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:16:36,570][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:16:37,070][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:16:37,568][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:16:38,068][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:16:38,569][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:16:39,067][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:16:39,565][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:16:40,066][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:16:40,563][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:16:41,059][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:16:41,577][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:16:42,079][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:16:42,578][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:16:43,076][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:16:43,571][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:16:44,072][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:16:44,570][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:16:45,069][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:16:45,565][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:16:46,061][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:16:46,557][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:16:47,053][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:16:47,558][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:16:48,055][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:16:48,552][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:16:49,049][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:16:49,546][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:16:50,046][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:16:50,545][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:16:51,040][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10098 tokens. +[2026-03-26 02:16:51,757][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 60.54%, Block Peak % of device VRAM: 62.14%, ΔTime: 00:00:32 +[2026-03-26 02:16:52,532][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:16:52,535][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:16:52,536][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:16:53,408][__main__][INFO] - Iteration 345 took 52s (31.65% Gen, 66.68% Train). Generation: 16s, Training: 34s. Estimated remaining time: 38h 20m 36s. Estimated total time: 43h 31m 49s. Time estimates for 10 more iterations: 8m 42s, 100 more iterations: 1h 27m 3s, 500 more iterations: 7h 15m 18s. +[2026-03-26 02:16:53,410][__main__][INFO] - Starting iteration 345. +[2026-03-26 02:16:53,808][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 34 and human policies 1. +[2026-03-26 02:16:53,809][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:16:59,837][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 02:17:09,145][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 20 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 02:17:10,938][__main__][INFO] - Number of regex retries in iteration 345: 2 +[2026-03-26 02:17:10,939][__main__][INFO] - agents played in iteration 345 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:17:11,699][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:17:11,723][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:17:11,747][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:17:11,770][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:17:11,771][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:17:11,772][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:17:12,452][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:17:12,916][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:17:13,422][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:17:13,925][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:17:14,427][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:17:14,928][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:17:15,426][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:17:15,927][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:17:16,428][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:17:16,939][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:17:17,439][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:17:17,937][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:17:18,434][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:17:18,933][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:17:19,434][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:17:19,934][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:17:20,433][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:17:20,932][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:17:21,433][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:17:21,940][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:17:22,441][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:17:22,942][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:17:23,443][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:17:23,943][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:17:24,442][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:17:24,942][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:17:25,441][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:17:25,939][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:17:26,439][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:17:26,937][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:17:27,436][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:17:27,943][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:17:28,444][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:17:28,940][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:17:29,438][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:17:29,935][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:17:30,432][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:17:30,929][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:17:31,427][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:17:31,924][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:17:32,420][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:17:32,917][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:17:33,417][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:17:33,917][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:17:34,415][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:17:34,911][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:17:35,408][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:17:35,908][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:17:36,418][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:17:36,919][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:17:37,420][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:17:37,920][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:17:38,421][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:17:38,925][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:17:39,424][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:17:39,927][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:17:40,428][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:17:40,927][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:17:41,424][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:17:41,922][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:17:42,418][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:17:42,914][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:17:43,411][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:17:43,910][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:17:44,410][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10265 tokens. +[2026-03-26 02:17:45,124][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 60.60%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 +[2026-03-26 02:17:45,889][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:17:45,891][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:17:45,894][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:17:46,760][__main__][INFO] - Iteration 346 took 52s (32.35% Gen, 66.01% Train). Generation: 17s, Training: 34s. Estimated remaining time: 38h 55m 32s. Estimated total time: 44h 7m 39s. Time estimates for 10 more iterations: 8m 49s, 100 more iterations: 1h 28m 15s, 500 more iterations: 7h 21m 16s. +[2026-03-26 02:17:46,763][__main__][INFO] - Starting iteration 346. +[2026-03-26 02:17:47,162][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 34 and human policies 1. +[2026-03-26 02:17:47,163][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:18:02,994][__main__][INFO] - Number of regex retries in iteration 346: 0 +[2026-03-26 02:18:02,995][__main__][INFO] - agents played in iteration 346 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:18:03,754][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:18:03,778][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:18:03,801][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:18:03,824][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:18:03,825][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:18:03,825][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:18:04,513][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:18:04,968][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:18:05,469][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:18:05,966][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:18:06,466][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:18:07,041][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:18:07,541][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:18:08,771][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:18:09,271][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:18:09,769][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:18:10,267][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:18:10,764][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:18:11,262][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:18:11,766][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:18:12,264][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:18:12,762][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:18:13,259][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:18:13,755][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:18:14,261][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:18:14,763][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:18:15,260][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:18:15,757][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:18:16,254][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:18:16,752][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:18:17,249][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:18:17,751][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:18:18,251][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:18:18,750][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:18:19,248][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:18:19,748][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:18:20,244][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:18:20,744][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:18:21,244][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:18:21,741][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:18:22,240][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:18:22,739][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:18:23,238][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:18:23,741][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:18:24,239][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:18:24,737][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:18:25,234][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:18:25,733][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:18:26,230][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:18:26,727][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:18:27,225][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:18:27,721][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:18:28,217][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:18:28,714][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:18:29,211][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:18:29,707][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:18:30,207][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:18:30,705][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:18:31,201][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:18:31,699][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:18:32,197][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:18:32,694][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:18:33,192][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:18:33,696][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:18:34,195][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:18:34,694][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:18:35,192][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:18:35,690][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:18:36,189][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:18:36,688][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:18:37,188][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10185 tokens. +[2026-03-26 02:18:37,971][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 60.54%, Block Peak % of device VRAM: 62.10%, ΔTime: 00:00:33 +[2026-03-26 02:18:38,744][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:18:38,746][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:18:38,748][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:18:39,635][__main__][INFO] - Iteration 347 took 52s (30.17% Gen, 68.14% Train). Generation: 15s, Training: 35s. Estimated remaining time: 38h 30m 42s. Estimated total time: 43h 43m 41s. Time estimates for 10 more iterations: 8m 44s, 100 more iterations: 1h 27m 27s, 500 more iterations: 7h 17m 16s. +[2026-03-26 02:18:39,653][__main__][INFO] - Starting iteration 347. +[2026-03-26 02:18:40,060][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 34 and human policies 1. +[2026-03-26 02:18:40,061][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:18:51,385][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 02:18:59,407][__main__][INFO] - Number of regex retries in iteration 347: 1 +[2026-03-26 02:18:59,408][__main__][INFO] - agents played in iteration 347 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:19:00,213][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:19:00,237][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:19:00,260][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:19:00,283][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:19:00,284][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:19:00,284][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:19:00,970][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:19:01,428][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:19:01,930][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:19:02,429][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:19:02,926][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:19:03,427][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:19:03,923][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:19:04,421][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:19:04,918][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:19:05,415][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:19:05,915][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:19:06,410][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:19:06,908][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:19:07,404][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:19:07,900][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:19:08,397][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:19:08,896][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:19:09,393][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:19:09,891][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:19:10,389][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:19:10,888][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:19:11,389][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:19:11,885][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:19:12,383][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:19:12,880][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:19:13,377][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:19:13,873][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:19:14,371][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:19:14,868][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:19:15,366][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:19:15,866][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:19:16,366][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:19:16,864][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:19:17,367][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:19:17,865][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:19:18,365][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:19:18,865][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:19:19,362][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:19:19,860][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:19:20,360][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:19:20,864][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:19:21,366][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:19:21,866][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:19:22,370][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:19:22,872][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:19:23,372][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:19:23,872][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:19:24,373][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:19:24,872][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:19:25,373][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:19:25,873][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:19:26,372][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:19:26,873][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:19:27,371][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:19:27,870][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:19:28,370][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:19:28,867][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:19:29,365][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:19:29,862][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:19:30,358][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:19:30,854][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:19:31,350][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:19:31,848][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:19:32,345][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:19:32,845][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10200 tokens. +[2026-03-26 02:19:33,569][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 60.60%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:32 +[2026-03-26 02:19:34,322][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:19:34,324][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:19:34,326][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:19:35,066][__main__][INFO] - Iteration 348 took 55s (35.17% Gen, 63.48% Train). Generation: 19s, Training: 34s. Estimated remaining time: 40h 36m 25s. Estimated total time: 45h 50m 20s. Time estimates for 10 more iterations: 9m 10s, 100 more iterations: 1h 31m 40s, 500 more iterations: 7h 38m 23s. +[2026-03-26 02:19:35,069][__main__][INFO] - Starting iteration 348. +[2026-03-26 02:19:35,468][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 34 and human policies 1. +[2026-03-26 02:19:35,469][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:19:46,161][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 02:19:55,117][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 1 y books, 1 z balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 02:19:59,150][__main__][INFO] - Number of regex retries in iteration 348: 2 +[2026-03-26 02:19:59,151][__main__][INFO] - agents played in iteration 348 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:19:59,915][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:19:59,938][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:19:59,962][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:19:59,985][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:19:59,985][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:19:59,986][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:20:00,672][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:20:01,127][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:20:01,632][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:20:02,136][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:20:02,633][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:20:03,132][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:20:03,629][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:20:04,130][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:20:04,631][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:20:05,133][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:20:05,634][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:20:06,134][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:20:06,633][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:20:07,133][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:20:07,633][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:20:08,135][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:20:08,637][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:20:09,136][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:20:09,637][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:20:10,138][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:20:10,637][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:20:11,133][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:20:11,630][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:20:12,128][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:20:12,626][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:20:13,126][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:20:13,626][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:20:14,124][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:20:14,630][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:20:15,130][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:20:15,629][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:20:16,129][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:20:16,629][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:20:17,132][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:20:17,632][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:20:18,130][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:20:18,629][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:20:19,129][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:20:19,629][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:20:20,128][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:20:20,627][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:20:21,124][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:20:21,625][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:20:22,121][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:20:22,618][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:20:23,120][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:20:23,620][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:20:24,118][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:20:24,616][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:20:25,117][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:20:25,617][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:20:26,116][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:20:26,612][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:20:27,112][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:20:27,610][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:20:28,112][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:20:28,612][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:20:29,109][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:20:29,606][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:20:30,108][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:20:30,606][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:20:31,105][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:20:31,602][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:20:32,101][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:20:32,597][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10313 tokens. +[2026-03-26 02:20:33,333][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:32 +[2026-03-26 02:20:34,105][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:20:34,107][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:20:34,109][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:20:34,828][__main__][INFO] - Iteration 349 took 59s (39.90% Gen, 58.89% Train). Generation: 23s, Training: 34s. Estimated remaining time: 44h 13m 8s. Estimated total time: 49h 28m 2s. Time estimates for 10 more iterations: 9m 53s, 100 more iterations: 1h 38m 56s, 500 more iterations: 8h 14m 40s. +[2026-03-26 02:20:34,830][__main__][INFO] - Starting iteration 349. +[2026-03-26 02:20:35,231][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 34 and human policies 1. +[2026-03-26 02:20:35,232][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:20:51,867][__main__][INFO] - Number of regex retries in iteration 349: 0 +[2026-03-26 02:20:51,868][__main__][INFO] - agents played in iteration 349 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:20:52,635][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:20:52,659][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:20:52,682][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:20:52,705][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:20:52,706][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:20:52,706][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:20:53,385][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:20:53,837][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:20:54,341][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:20:54,839][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:20:55,337][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:20:55,834][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:20:56,331][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:20:56,827][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:20:57,324][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:20:57,826][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:20:58,327][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:20:58,826][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:20:59,323][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:20:59,822][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:21:00,321][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:21:00,825][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:21:01,340][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:21:01,863][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:21:02,365][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:21:02,867][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:21:03,367][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:21:03,869][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:21:04,368][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:21:04,867][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:21:05,365][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:21:05,864][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:21:06,361][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:21:06,859][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:21:07,355][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:21:07,852][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:21:08,348][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:21:08,844][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:21:09,340][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:21:09,839][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:21:10,338][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:21:10,833][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:21:11,333][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:21:11,829][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:21:12,328][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:21:12,824][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:21:13,320][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:21:13,819][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:21:14,318][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:21:14,816][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:21:15,316][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:21:15,814][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:21:16,313][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:21:16,812][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:21:17,310][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:21:17,808][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:21:18,308][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:21:18,805][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:21:19,303][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:21:19,802][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:21:20,303][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:21:20,803][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:21:21,304][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:21:21,806][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:21:22,303][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:21:22,800][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:21:23,299][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:21:23,797][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:21:24,295][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:21:24,794][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:21:25,297][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10274 tokens. +[2026-03-26 02:21:26,025][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 60.66%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:32 +[2026-03-26 02:21:26,793][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:21:26,796][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:21:26,797][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:21:27,520][__main__][INFO] - Iteration 350 took 52s (31.81% Gen, 66.80% Train). Generation: 16s, Training: 34s. Estimated remaining time: 38h 18m 41s. Estimated total time: 43h 34m 28s. Time estimates for 10 more iterations: 8m 42s, 100 more iterations: 1h 27m 8s, 500 more iterations: 7h 15m 44s. +[2026-03-26 02:21:27,525][__main__][INFO] - Starting iteration 350. +[2026-03-26 02:21:27,922][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 34 and human policies 1. +[2026-03-26 02:21:27,923][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:21:34,955][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 02:21:47,046][__main__][INFO] - Number of regex retries in iteration 350: 1 +[2026-03-26 02:21:47,046][__main__][INFO] - agents played in iteration 350 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:21:47,854][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:21:47,878][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:21:47,901][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:21:47,924][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:21:47,925][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:21:47,926][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:21:48,614][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:21:49,067][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:21:49,571][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:21:50,083][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:21:50,751][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:21:51,249][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:21:51,750][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:21:52,253][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:21:52,753][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:21:53,251][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:21:53,798][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:21:54,298][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:21:54,796][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:21:55,294][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:21:55,794][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:21:56,294][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:21:56,794][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:21:57,294][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:21:57,791][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:21:58,289][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:21:58,787][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:21:59,287][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:21:59,786][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:22:00,285][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:22:00,783][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:22:01,282][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:22:01,780][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:22:02,277][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:22:02,773][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:22:03,270][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:22:03,766][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:22:04,263][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:22:04,756][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:22:05,255][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:22:05,760][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:22:06,259][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:22:06,758][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:22:07,256][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:22:07,755][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:22:08,272][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:22:08,771][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:22:09,266][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:22:09,765][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:22:10,260][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:22:10,754][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:22:11,249][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:22:11,744][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:22:12,242][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:22:12,743][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:22:13,240][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:22:13,753][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:22:14,251][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:22:14,749][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:22:15,243][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:22:15,746][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:22:16,244][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:22:16,741][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:22:17,253][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:22:17,754][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:22:18,251][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:22:18,752][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:22:19,251][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:22:19,754][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:22:20,253][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:22:20,755][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10191 tokens. +[2026-03-26 02:22:21,535][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 60.65%, Block Peak % of device VRAM: 62.42%, ΔTime: 00:00:32 +[2026-03-26 02:22:22,308][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:22:22,310][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:22:22,312][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:22:23,943][__main__][INFO] - Iteration 351 took 56s (34.14% Gen, 62.95% Train). Generation: 19s, Training: 35s. Estimated remaining time: 41h 24m 20s. Estimated total time: 46h 41m 4s. Time estimates for 10 more iterations: 9m 20s, 100 more iterations: 1h 33m 22s, 500 more iterations: 7h 46m 50s. +[2026-03-26 02:22:23,947][__main__][INFO] - Starting iteration 351. +[2026-03-26 02:22:24,348][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 35 and human policies 1. +[2026-03-26 02:22:24,348][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:22:40,485][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 02:22:53,852][__main__][INFO] - Number of regex retries in iteration 351: 1 +[2026-03-26 02:22:53,852][__main__][INFO] - agents played in iteration 351 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:22:54,665][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:22:54,688][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:22:54,711][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:22:54,734][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:22:54,735][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:22:54,735][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:22:55,447][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:22:55,978][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:22:56,483][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:22:56,980][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:22:57,477][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:22:57,976][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:22:58,473][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:22:58,974][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:22:59,471][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:22:59,971][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:23:00,471][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:23:00,968][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:23:01,466][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:23:01,965][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:23:02,461][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:23:02,961][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:23:03,458][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:23:03,955][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:23:04,453][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:23:04,954][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:23:05,454][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:23:05,950][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:23:06,447][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:23:06,945][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:23:07,445][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:23:07,942][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:23:08,441][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:23:08,937][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:23:09,436][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:23:09,933][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:23:10,432][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:23:10,930][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:23:11,430][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:23:11,928][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:23:12,424][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:23:12,920][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:23:13,415][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:23:13,912][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:23:14,413][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:23:14,915][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:23:15,413][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:23:15,913][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:23:16,412][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:23:16,912][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:23:17,413][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:23:17,916][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:23:18,414][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:23:18,912][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:23:19,412][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:23:19,910][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:23:20,408][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:23:20,903][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:23:21,404][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:23:21,903][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:23:22,401][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:23:22,898][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:23:23,395][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:23:23,893][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:23:24,389][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:23:24,886][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:23:25,383][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:23:25,886][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:23:26,383][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:23:26,880][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:23:27,379][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10335 tokens. +[2026-03-26 02:23:28,103][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 60.64%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:32 +[2026-03-26 02:23:28,877][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:23:28,880][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:23:28,881][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:23:29,536][__main__][INFO] - Iteration 352 took 1m 5s (45.26% Gen, 53.73% Train). Generation: 29s, Training: 35s. Estimated remaining time: 49h 1m 38s. Estimated total time: 54h 19m 27s. Time estimates for 10 more iterations: 10m 51s, 100 more iterations: 1h 48m 38s, 500 more iterations: 9h 3m 14s. +[2026-03-26 02:23:29,540][__main__][INFO] - Starting iteration 352. +[2026-03-26 02:23:29,944][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 35 and human policies 1. +[2026-03-26 02:23:29,945][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:24:11,352][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 02:24:16,200][__main__][INFO] - Number of regex retries in iteration 352: 1 +[2026-03-26 02:24:16,201][__main__][INFO] - agents played in iteration 352 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:24:16,963][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:24:16,987][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:24:17,010][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:24:17,033][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:24:17,034][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:24:17,034][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:24:17,743][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:24:18,238][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:24:18,740][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:24:19,237][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:24:19,744][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:24:20,247][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:24:20,749][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:24:21,249][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:24:21,749][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:24:22,265][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:24:22,768][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:24:23,269][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:24:23,795][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:24:24,296][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:24:24,797][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:24:25,300][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:24:25,802][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:24:26,301][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:24:26,801][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:24:27,299][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:24:27,800][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:24:28,299][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:24:28,817][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:24:29,318][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:24:29,825][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:24:30,322][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:24:30,825][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:24:31,330][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:24:31,831][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:24:32,334][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:24:32,833][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:24:33,330][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:24:33,829][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:24:34,478][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:24:35,269][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:24:35,770][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:24:36,270][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:24:36,773][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:24:37,274][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:24:37,775][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:24:38,274][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:24:38,771][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:24:39,269][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:24:39,767][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:24:40,262][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:24:40,761][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:24:41,261][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:24:41,760][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:24:42,260][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:24:42,755][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:24:43,256][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:24:43,752][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:24:44,248][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:24:44,748][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:24:45,243][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:24:45,741][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:24:46,244][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:24:46,745][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:24:47,242][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:24:47,738][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:24:48,233][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:24:48,729][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:24:49,224][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:24:49,719][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:24:50,214][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10221 tokens. +[2026-03-26 02:24:51,045][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 60.56%, Block Peak % of device VRAM: 62.38%, ΔTime: 00:00:33 +[2026-03-26 02:24:51,869][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:24:51,872][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:24:51,873][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:24:52,623][__main__][INFO] - Iteration 353 took 1m 22s (55.94% Gen, 43.15% Train). Generation: 46s, Training: 35s. Estimated remaining time: 63h 34m 44s. Estimated total time: 68h 53m 57s. Time estimates for 10 more iterations: 13m 46s, 100 more iterations: 2h 17m 47s, 500 more iterations: 11h 28m 59s. +[2026-03-26 02:24:52,631][__main__][INFO] - Starting iteration 353. +[2026-03-26 02:24:53,035][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 35 and human policies 1. +[2026-03-26 02:24:53,035][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:25:36,094][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 02:25:46,026][__main__][INFO] - Number of regex retries in iteration 353: 1 +[2026-03-26 02:25:46,027][__main__][INFO] - agents played in iteration 353 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:25:46,903][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:25:46,927][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:25:46,951][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:25:46,978][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:25:46,979][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:25:46,979][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:25:47,744][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:25:48,256][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:25:48,772][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:25:49,270][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:25:49,776][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:25:50,276][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:25:50,775][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:25:51,276][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:25:51,775][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:25:52,278][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:25:52,774][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:25:53,274][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:25:53,773][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:25:54,273][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:25:54,775][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:25:55,296][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:25:55,801][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:25:56,305][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:25:56,810][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:25:57,311][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:25:57,814][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:25:58,317][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:25:58,817][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:25:59,320][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:25:59,820][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:26:00,317][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:26:00,819][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:26:01,315][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:26:01,814][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:26:02,312][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:26:02,808][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:26:03,305][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:26:03,802][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:26:04,298][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:26:04,794][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:26:05,293][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:26:05,790][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:26:06,288][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:26:06,784][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:26:07,280][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:26:07,779][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:26:08,275][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:26:08,774][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:26:09,273][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:26:09,771][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:26:10,268][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:26:10,763][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:26:11,259][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:26:11,756][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:26:12,980][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:26:13,479][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:26:13,978][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:26:14,478][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:26:14,978][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:26:15,477][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:26:15,975][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:26:16,476][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:26:16,975][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:26:17,476][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:26:17,974][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:26:18,472][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:26:18,973][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:26:19,474][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:26:19,975][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:26:20,478][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10159 tokens. +[2026-03-26 02:26:21,318][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 60.56%, Block Peak % of device VRAM: 62.35%, ΔTime: 00:00:33 +[2026-03-26 02:26:21,992][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:26:21,994][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:26:21,995][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:26:22,673][__main__][INFO] - Iteration 354 took 1m 29s (59.12% Gen, 40.13% Train). Generation: 52s, Training: 35s. Estimated remaining time: 69h 21m 13s. Estimated total time: 74h 41m 56s. Time estimates for 10 more iterations: 14m 56s, 100 more iterations: 2h 29m 23s, 500 more iterations: 12h 26m 59s. +[2026-03-26 02:26:22,688][__main__][INFO] - Starting iteration 354. +[2026-03-26 02:26:23,093][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 35 and human policies 1. +[2026-03-26 02:26:23,094][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:27:09,133][__main__][INFO] - Number of regex retries in iteration 354: 0 +[2026-03-26 02:27:09,135][__main__][INFO] - agents played in iteration 354 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:27:10,567][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:27:10,590][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:27:10,613][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:27:10,636][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:27:10,637][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:27:10,638][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:27:11,608][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:27:12,297][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:27:12,811][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:27:13,312][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:27:13,810][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:27:14,308][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:27:14,812][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:27:15,311][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:27:15,808][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:27:16,310][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:27:16,815][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:27:17,315][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:27:17,841][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:27:18,340][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:27:18,841][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:27:19,341][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:27:19,842][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:27:20,339][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:27:20,839][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:27:21,353][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:27:21,858][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:27:22,360][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:27:22,862][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:27:23,363][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:27:23,864][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:27:24,360][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:27:24,857][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:27:25,354][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:27:25,851][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:27:26,349][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:27:26,847][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:27:27,344][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:27:27,839][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:27:28,341][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:27:28,841][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:27:29,343][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:27:29,845][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:27:30,345][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:27:30,845][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:27:31,347][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:27:31,847][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:27:32,347][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:27:32,847][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:27:33,346][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:27:33,846][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:27:34,346][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:27:34,841][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:27:35,341][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:27:35,837][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:27:36,333][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:27:36,832][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:27:37,328][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:27:37,826][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:27:38,322][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:27:38,818][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:27:39,314][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:27:39,810][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:27:40,306][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:27:40,803][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:27:41,299][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:27:41,795][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:27:42,292][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:27:42,790][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:27:43,288][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:27:43,789][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10159 tokens. +[2026-03-26 02:27:44,540][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.26%, Current % of VRAM taken: 60.68%, Block Peak % of device VRAM: 62.43%, ΔTime: 00:00:32 +[2026-03-26 02:27:45,228][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:27:45,231][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:27:45,233][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:27:45,873][__main__][INFO] - Iteration 355 took 1m 22s (55.62% Gen, 43.61% Train). Generation: 46s, Training: 36s. Estimated remaining time: 63h 36m 55s. Estimated total time: 68h 59m 1s. Time estimates for 10 more iterations: 13m 47s, 100 more iterations: 2h 17m 58s, 500 more iterations: 11h 29m 50s. +[2026-03-26 02:27:45,878][__main__][INFO] - Starting iteration 355. +[2026-03-26 02:27:46,281][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 35 and human policies 1. +[2026-03-26 02:27:46,281][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:28:20,066][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 02:28:32,107][__main__][INFO] - Number of regex retries in iteration 355: 1 +[2026-03-26 02:28:32,108][__main__][INFO] - agents played in iteration 355 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:28:32,878][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:28:32,902][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:28:32,925][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:28:32,948][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:28:32,949][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:28:32,949][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:28:33,636][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:28:34,091][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:28:34,608][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:28:35,109][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:28:35,608][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:28:36,112][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:28:36,616][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:28:37,123][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:28:37,624][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:28:38,130][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:28:38,632][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:28:39,133][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:28:39,634][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:28:40,134][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:28:40,636][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:28:41,136][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:28:41,636][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:28:42,132][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:28:42,631][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:28:43,131][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:28:43,632][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:28:44,132][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:28:44,628][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:28:45,133][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:28:45,632][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:28:46,132][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:28:46,630][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:28:47,128][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:28:47,626][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:28:48,127][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:28:48,633][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:28:49,137][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:28:49,637][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:28:50,139][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:28:50,641][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:28:51,143][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:28:51,646][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:28:52,148][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:28:52,648][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:28:53,150][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:28:53,651][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:28:54,150][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:28:54,649][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:28:55,148][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:28:55,647][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:28:56,148][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:28:56,647][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:28:57,147][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:28:57,653][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:28:58,151][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:28:58,649][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:28:59,146][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:28:59,643][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:29:00,138][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:29:00,634][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:29:01,130][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:29:01,629][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:29:02,129][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:29:02,627][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:29:03,123][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:29:03,620][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:29:04,117][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:29:04,613][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:29:05,110][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:29:05,607][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10213 tokens. +[2026-03-26 02:29:06,332][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 60.50%, Block Peak % of device VRAM: 62.41%, ΔTime: 00:00:32 +[2026-03-26 02:29:07,084][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:29:07,087][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:29:07,088][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:29:07,840][__main__][INFO] - Iteration 356 took 1m 21s (56.19% Gen, 42.89% Train). Generation: 45s, Training: 34s. Estimated remaining time: 62h 34m 30s. Estimated total time: 67h 57m 58s. Time estimates for 10 more iterations: 13m 35s, 100 more iterations: 2h 15m 55s, 500 more iterations: 11h 19m 39s. +[2026-03-26 02:29:07,842][__main__][INFO] - Starting iteration 356. +[2026-03-26 02:29:08,242][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 35 and human policies 1. +[2026-03-26 02:29:08,242][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:29:33,542][__main__][INFO] - Number of regex retries in iteration 356: 0 +[2026-03-26 02:29:33,543][__main__][INFO] - agents played in iteration 356 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:29:34,298][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:29:34,321][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:29:34,345][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:29:34,369][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:29:34,369][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:29:34,370][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:29:35,065][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:29:35,521][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:29:36,030][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:29:36,527][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:29:37,026][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:29:37,524][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:29:38,033][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:29:38,535][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:29:39,035][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:29:39,536][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:29:40,036][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:29:40,532][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:29:41,033][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:29:41,530][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:29:42,028][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:29:42,524][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:29:43,020][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:29:43,516][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:29:44,013][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:29:44,511][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:29:45,009][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:29:45,507][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:29:46,012][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:29:46,514][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:29:47,011][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:29:47,510][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:29:48,008][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:29:48,510][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:29:49,007][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:29:49,505][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:29:50,003][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:29:50,500][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:29:50,998][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:29:51,496][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:29:51,994][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:29:52,492][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:29:52,993][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:29:53,489][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:29:53,987][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:29:54,483][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:29:54,979][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:29:55,475][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:29:55,973][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:29:56,471][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:29:56,969][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:29:57,468][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:29:57,968][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:29:58,465][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:29:58,962][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:29:59,461][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:29:59,960][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:30:00,455][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:30:00,952][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:30:01,452][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:30:01,951][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:30:02,449][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:30:02,947][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:30:03,443][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:30:03,942][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:30:04,438][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:30:04,934][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:30:05,431][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:30:05,932][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:30:06,431][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:30:06,930][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10221 tokens. +[2026-03-26 02:30:07,650][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 60.54%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 +[2026-03-26 02:30:08,421][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:30:08,423][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:30:08,425][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:30:09,126][__main__][INFO] - Iteration 357 took 1m 0s (41.55% Gen, 57.29% Train). Generation: 25s, Training: 34s. Estimated remaining time: 45h 19m 45s. Estimated total time: 50h 44m 14s. Time estimates for 10 more iterations: 10m 8s, 100 more iterations: 1h 41m 28s, 500 more iterations: 8h 27m 22s. +[2026-03-26 02:30:09,128][__main__][INFO] - Starting iteration 357. +[2026-03-26 02:30:09,528][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 35 and human policies 1. +[2026-03-26 02:30:09,529][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:30:28,071][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 02:30:40,448][__main__][INFO] - Number of regex retries in iteration 357: 1 +[2026-03-26 02:30:40,448][__main__][INFO] - agents played in iteration 357 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:30:41,208][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:30:41,232][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:30:41,255][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:30:41,278][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:30:41,279][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:30:41,279][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:30:41,972][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:30:42,434][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:30:42,939][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:30:43,444][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:30:43,948][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:30:44,451][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:30:44,953][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:30:45,457][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:30:45,959][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:30:46,457][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:30:46,955][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:30:47,460][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:30:47,961][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:30:48,462][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:30:48,961][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:30:49,458][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:30:49,956][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:30:50,457][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:30:50,956][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:30:51,457][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:30:51,956][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:30:52,458][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:30:52,956][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:30:53,456][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:30:53,957][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:30:54,456][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:30:54,956][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:30:55,463][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:30:55,967][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:30:56,467][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:30:56,966][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:30:57,466][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:30:57,964][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:30:58,463][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:30:58,961][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:30:59,461][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:30:59,958][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:31:00,455][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:31:00,953][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:31:01,451][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:31:01,952][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:31:02,678][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:31:03,869][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:31:04,374][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:31:04,886][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:31:05,388][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:31:05,887][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:31:06,385][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:31:06,883][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:31:07,382][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:31:07,881][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:31:08,379][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:31:08,878][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:31:09,381][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:31:09,878][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:31:10,376][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:31:10,875][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:31:11,375][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:31:11,872][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:31:12,369][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:31:12,868][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:31:13,365][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:31:13,862][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:31:14,361][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:31:14,864][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10260 tokens. +[2026-03-26 02:31:15,678][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.25%, Current % of VRAM taken: 60.68%, Block Peak % of device VRAM: 62.41%, ΔTime: 00:00:33 +[2026-03-26 02:31:16,389][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:31:16,392][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:31:16,393][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:31:17,034][__main__][INFO] - Iteration 358 took 1m 7s (45.80% Gen, 53.25% Train). Generation: 30s, Training: 35s. Estimated remaining time: 50h 49m 40s. Estimated total time: 56h 15m 17s. Time estimates for 10 more iterations: 11m 15s, 100 more iterations: 1h 52m 30s, 500 more iterations: 9h 22m 32s. +[2026-03-26 02:31:17,048][__main__][INFO] - Starting iteration 358. +[2026-03-26 02:31:17,452][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 35 and human policies 1. +[2026-03-26 02:31:17,452][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:32:27,060][__main__][INFO] - Number of regex retries in iteration 358: 0 +[2026-03-26 02:32:27,062][__main__][INFO] - agents played in iteration 358 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:32:27,997][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:32:28,025][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:32:28,049][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:32:28,072][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:32:28,073][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:32:28,073][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:32:28,853][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:32:29,397][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:32:29,909][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:32:30,411][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:32:30,908][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:32:31,405][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:32:31,903][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:32:32,400][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:32:32,899][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:32:33,401][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:32:33,903][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:32:34,404][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:32:34,914][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:32:35,417][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:32:35,924][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:32:36,425][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:32:36,928][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:32:37,430][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:32:37,930][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:32:38,872][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:32:39,372][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:32:39,876][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:32:40,376][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:32:40,878][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:32:41,377][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:32:41,879][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:32:42,378][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:32:42,876][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:32:43,377][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:32:43,878][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:32:44,373][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:32:44,874][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:32:45,376][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:32:45,873][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:32:46,371][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:32:46,870][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:32:47,369][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:32:47,870][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:32:48,367][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:32:48,864][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:32:49,361][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:32:49,863][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:32:50,360][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:32:50,863][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:32:51,362][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:32:51,861][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:32:52,358][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:32:52,854][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:32:53,351][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:32:53,850][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:32:54,347][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:32:54,846][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:32:55,346][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:32:55,846][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:32:56,342][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:32:56,839][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:32:57,335][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:32:57,836][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:32:58,335][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:32:58,833][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:32:59,332][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:32:59,833][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:33:00,334][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:33:00,831][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:33:01,330][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10236 tokens. +[2026-03-26 02:33:02,142][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 60.50%, Block Peak % of device VRAM: 62.43%, ΔTime: 00:00:33 +[2026-03-26 02:33:02,876][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:33:02,880][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:33:02,882][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:33:03,663][__main__][INFO] - Iteration 359 took 1m 46s (65.54% Gen, 33.72% Train). Generation: 1m 9s, Training: 35s. Estimated remaining time: 83h 3m 12s. Estimated total time: 88h 30m 35s. Time estimates for 10 more iterations: 17m 42s, 100 more iterations: 2h 57m 1s, 500 more iterations: 14h 45m 5s. +[2026-03-26 02:33:03,676][__main__][INFO] - Starting iteration 359. +[2026-03-26 02:33:04,082][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 35 and human policies 1. +[2026-03-26 02:33:04,083][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:33:51,467][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 02:34:04,378][__main__][INFO] - Number of regex retries in iteration 359: 1 +[2026-03-26 02:34:04,379][__main__][INFO] - agents played in iteration 359 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:34:05,222][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:34:05,245][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:34:05,269][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:34:05,292][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:34:05,292][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:34:05,293][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:34:06,052][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:34:06,537][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:34:07,051][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:34:07,553][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:34:08,051][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:34:08,549][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:34:09,052][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:34:09,558][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:34:10,058][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:34:10,556][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:34:11,055][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:34:11,558][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:34:12,058][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:34:12,556][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:34:13,073][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:34:13,577][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:34:14,076][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:34:14,576][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:34:15,073][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:34:15,574][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:34:16,074][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:34:16,574][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:34:17,073][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:34:17,570][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:34:18,066][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:34:18,561][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:34:19,058][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:34:19,554][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:34:20,051][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:34:20,549][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:34:21,047][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:34:21,543][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:34:22,039][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:34:22,536][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:34:23,033][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:34:23,529][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:34:24,027][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:34:24,524][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:34:25,031][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:34:25,532][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:34:26,031][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:34:26,529][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:34:27,027][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:34:27,523][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:34:28,019][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:34:28,516][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:34:29,020][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:34:29,517][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:34:30,016][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:34:30,513][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:34:31,012][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:34:31,510][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:34:32,010][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:34:32,510][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:34:33,007][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:34:33,511][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:34:34,014][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:34:34,511][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:34:35,009][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:34:35,508][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:34:36,005][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:34:36,504][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:34:37,002][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:34:37,498][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:34:37,995][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10170 tokens. +[2026-03-26 02:34:38,721][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.09%, Current % of VRAM taken: 60.52%, Block Peak % of device VRAM: 62.21%, ΔTime: 00:00:32 +[2026-03-26 02:34:39,474][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:34:39,476][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:34:39,478][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:34:40,196][__main__][INFO] - Iteration 360 took 1m 36s (62.73% Gen, 36.52% Train). Generation: 1m 0s, Training: 35s. Estimated remaining time: 74h 36m 44s. Estimated total time: 80h 5m 44s. Time estimates for 10 more iterations: 16m 1s, 100 more iterations: 2h 40m 11s, 500 more iterations: 13h 20m 57s. +[2026-03-26 02:34:40,198][__main__][INFO] - Starting iteration 360. +[2026-03-26 02:34:40,596][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 35 and human policies 1. +[2026-03-26 02:34:40,596][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:35:23,868][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 20 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 02:35:24,796][__main__][INFO] - Number of regex retries in iteration 360: 1 +[2026-03-26 02:35:24,797][__main__][INFO] - agents played in iteration 360 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:35:25,557][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:35:25,580][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:35:25,604][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:35:25,628][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:35:25,628][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:35:25,629][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:35:26,326][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:35:26,782][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:35:27,287][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:35:27,789][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:35:28,290][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:35:28,792][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:35:29,292][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:35:29,802][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:35:30,303][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:35:30,802][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:35:31,300][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:35:31,801][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:35:32,302][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:35:32,800][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:35:33,298][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:35:33,795][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:35:34,294][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:35:34,795][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:35:36,106][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:35:36,605][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:35:37,103][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:35:37,603][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:35:38,113][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:35:38,615][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:35:39,115][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:35:39,613][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:35:40,112][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:35:40,609][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:35:41,107][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:35:41,605][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:35:42,102][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:35:42,598][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:35:43,095][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:35:43,593][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:35:44,091][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:35:44,590][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:35:45,089][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:35:45,587][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:35:46,083][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:35:46,579][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:35:47,074][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:35:47,570][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:35:48,069][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:35:48,568][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:35:49,068][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:35:49,569][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:35:50,065][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:35:50,560][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:35:51,055][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:35:51,555][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:35:52,051][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:35:52,548][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:35:53,045][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:35:53,542][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:35:54,041][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:35:54,539][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:35:55,035][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:35:55,532][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:35:56,029][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:35:56,527][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:35:57,026][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:35:57,522][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:35:58,023][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:35:58,520][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:35:59,017][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10171 tokens. +[2026-03-26 02:35:59,866][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.58%, Block Peak % of device VRAM: 62.22%, ΔTime: 00:00:33 +[2026-03-26 02:36:00,701][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:36:00,703][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:36:00,704][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:36:02,092][__main__][INFO] - Iteration 361 took 1m 21s (54.24% Gen, 44.06% Train). Generation: 44s, Training: 35s. Estimated remaining time: 62h 24m 29s. Estimated total time: 67h 54m 51s. Time estimates for 10 more iterations: 13m 34s, 100 more iterations: 2h 15m 49s, 500 more iterations: 11h 19m 8s. +[2026-03-26 02:36:02,099][__main__][INFO] - Starting iteration 361. +[2026-03-26 02:36:02,501][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 36 and human policies 1. +[2026-03-26 02:36:02,501][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:36:22,989][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 02:36:23,391][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 02:36:32,163][__main__][INFO] - Number of regex retries in iteration 361: 2 +[2026-03-26 02:36:32,164][__main__][INFO] - agents played in iteration 361 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:36:32,926][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:36:32,950][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:36:32,974][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:36:32,997][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:36:32,997][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:36:32,998][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:36:33,686][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:36:34,139][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:36:34,642][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:36:35,142][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:36:35,639][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:36:36,148][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:36:36,650][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:36:37,148][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:36:37,644][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:36:38,147][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:36:38,650][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:36:39,148][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:36:39,646][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:36:40,146][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:36:40,644][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:36:41,141][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:36:41,638][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:36:42,136][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:36:42,633][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:36:43,131][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:36:43,630][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:36:44,127][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:36:44,631][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:36:45,132][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:36:45,638][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:36:46,137][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:36:46,638][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:36:47,140][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:36:47,637][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:36:48,135][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:36:48,634][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:36:49,132][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:36:49,630][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:36:50,127][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:36:50,627][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:36:51,124][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:36:51,619][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:36:52,115][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:36:52,611][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:36:53,106][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:36:53,602][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:36:54,099][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:36:54,600][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:36:55,100][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:36:55,600][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:36:56,100][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:36:56,595][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:36:57,095][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:36:57,591][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:36:58,088][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:36:58,586][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:36:59,081][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:36:59,577][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:37:00,077][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:37:00,574][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:37:01,071][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:37:01,570][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:37:02,072][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:37:02,600][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:37:03,102][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:37:03,600][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:37:04,099][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:37:04,599][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:37:05,098][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:37:05,598][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10220 tokens. +[2026-03-26 02:37:06,318][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 60.64%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:32 +[2026-03-26 02:37:07,079][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:37:07,081][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:37:07,083][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:37:07,936][__main__][INFO] - Iteration 362 took 1m 5s (45.33% Gen, 53.36% Train). Generation: 29s, Training: 34s. Estimated remaining time: 49h 0m 20s. Estimated total time: 54h 31m 48s. Time estimates for 10 more iterations: 10m 54s, 100 more iterations: 1h 49m 3s, 500 more iterations: 9h 5m 18s. +[2026-03-26 02:37:07,938][__main__][INFO] - Starting iteration 362. +[2026-03-26 02:37:08,336][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 36 and human policies 1. +[2026-03-26 02:37:08,337][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:37:27,247][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 02:37:32,113][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 02:37:37,425][__main__][INFO] - Number of regex retries in iteration 362: 2 +[2026-03-26 02:37:37,426][__main__][INFO] - agents played in iteration 362 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:37:38,191][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:37:38,214][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:37:38,238][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:37:38,261][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:37:38,261][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:37:38,262][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:37:38,949][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:37:39,403][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:37:39,905][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:37:40,401][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:37:40,897][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:37:41,394][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:37:41,891][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:37:42,392][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:37:42,891][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:37:43,389][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:37:43,889][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:37:44,387][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:37:44,884][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:37:45,386][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:37:45,881][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:37:46,381][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:37:46,877][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:37:47,375][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:37:47,872][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:37:48,380][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:37:48,882][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:37:49,385][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:37:49,887][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:37:50,385][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:37:50,883][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:37:51,380][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:37:51,877][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:37:52,374][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:37:52,871][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:37:53,366][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:37:53,862][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:37:54,357][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:37:54,852][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:37:55,350][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:37:55,849][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:37:56,349][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:37:56,846][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:37:57,344][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:37:57,843][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:37:58,339][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:37:58,839][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:37:59,334][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:37:59,833][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:38:00,330][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:38:00,828][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:38:01,325][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:38:01,823][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:38:02,322][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:38:02,817][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:38:03,313][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:38:03,809][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:38:04,306][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:38:04,808][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:38:05,305][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:38:05,806][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:38:06,303][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:38:06,800][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:38:07,300][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:38:07,795][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:38:08,293][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:38:08,790][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:38:09,286][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:38:09,782][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:38:10,277][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:38:10,777][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10124 tokens. +[2026-03-26 02:38:11,497][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.08%, ΔTime: 00:00:32 +[2026-03-26 02:38:12,243][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:38:12,246][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:38:12,247][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:38:12,969][__main__][INFO] - Iteration 363 took 1m 4s (45.01% Gen, 53.87% Train). Generation: 29s, Training: 34s. Estimated remaining time: 48h 19m 8s. Estimated total time: 53h 51m 40s. Time estimates for 10 more iterations: 10m 46s, 100 more iterations: 1h 47m 43s, 500 more iterations: 8h 58m 36s. +[2026-03-26 02:38:12,971][__main__][INFO] - Starting iteration 363. +[2026-03-26 02:38:13,370][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 36 and human policies 1. +[2026-03-26 02:38:13,371][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:38:32,216][__main__][INFO] - Number of regex retries in iteration 363: 0 +[2026-03-26 02:38:32,217][__main__][INFO] - agents played in iteration 363 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:38:33,036][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:38:33,059][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:38:33,087][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:38:33,113][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:38:33,114][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:38:33,115][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:38:33,824][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:38:34,385][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:38:34,893][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:38:35,396][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:38:35,894][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:38:36,401][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:38:36,905][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:38:37,406][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:38:37,908][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:38:38,404][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:38:38,902][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:38:39,400][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:38:39,897][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:38:40,395][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:38:40,902][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:38:41,401][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:38:41,900][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:38:42,399][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:38:42,897][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:38:43,400][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:38:43,899][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:38:44,399][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:38:44,896][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:38:45,398][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:38:45,896][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:38:46,402][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:38:46,904][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:38:47,403][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:38:47,901][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:38:48,399][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:38:48,897][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:38:49,396][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:38:49,893][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:38:50,391][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:38:50,887][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:38:51,384][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:38:51,883][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:38:52,381][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:38:52,880][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:38:53,375][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:38:53,870][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:38:54,369][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:38:54,866][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:38:55,363][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:38:55,860][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:38:56,363][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:38:56,861][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:38:57,358][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:38:57,857][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:38:58,353][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:38:58,852][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:38:59,349][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:38:59,845][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:39:00,341][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:39:00,839][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:39:01,339][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:39:01,835][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:39:02,333][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:39:02,831][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:39:03,329][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:39:03,826][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:39:04,325][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:39:04,821][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:39:05,317][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:39:05,815][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10157 tokens. +[2026-03-26 02:39:06,614][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 60.51%, Block Peak % of device VRAM: 62.07%, ΔTime: 00:00:32 +[2026-03-26 02:39:07,432][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:39:07,434][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:39:07,436][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:39:08,153][__main__][INFO] - Iteration 364 took 54s (34.40% Gen, 64.29% Train). Generation: 18s, Training: 35s. Estimated remaining time: 40h 5m 41s. Estimated total time: 45h 39m 9s. Time estimates for 10 more iterations: 9m 7s, 100 more iterations: 1h 31m 18s, 500 more iterations: 7h 36m 31s. +[2026-03-26 02:39:08,159][__main__][INFO] - Starting iteration 364. +[2026-03-26 02:39:08,561][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 36 and human policies 1. +[2026-03-26 02:39:08,561][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:39:36,355][__main__][INFO] - Number of regex retries in iteration 364: 0 +[2026-03-26 02:39:36,357][__main__][INFO] - agents played in iteration 364 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:39:37,119][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:39:37,144][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:39:37,168][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:39:37,193][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:39:37,194][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:39:37,195][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:39:37,893][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:39:38,347][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:39:38,851][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:39:39,356][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:39:39,855][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:39:40,355][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:39:40,853][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:39:41,352][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:39:41,849][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:39:42,355][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:39:42,856][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:39:43,359][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:39:43,858][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:39:44,355][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:39:44,857][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:39:45,354][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:39:45,855][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:39:46,364][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:39:46,887][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:39:47,388][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:39:47,889][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:39:48,389][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:39:48,886][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:39:49,387][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:39:49,887][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:39:50,382][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:39:50,881][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:39:51,386][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:39:51,888][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:39:52,383][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:39:52,882][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:39:53,378][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:39:53,876][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:39:54,377][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:39:54,877][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:39:55,378][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:39:55,878][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:39:56,379][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:39:56,880][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:39:57,380][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:39:57,881][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:39:58,378][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:39:58,875][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:39:59,373][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:39:59,871][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:40:00,368][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:40:00,869][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:40:01,364][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:40:01,863][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:40:02,363][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:40:02,858][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:40:03,356][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:40:03,853][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:40:04,348][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:40:04,843][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:40:05,337][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:40:05,833][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:40:06,330][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:40:06,830][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:40:07,327][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:40:07,825][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:40:08,325][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:40:08,823][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:40:09,321][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:40:09,818][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10220 tokens. +[2026-03-26 02:40:10,545][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 60.57%, Block Peak % of device VRAM: 62.44%, ΔTime: 00:00:32 +[2026-03-26 02:40:11,297][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:40:11,299][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:40:11,301][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:40:12,023][__main__][INFO] - Iteration 365 took 1m 3s (43.80% Gen, 55.06% Train). Generation: 27s, Training: 34s. Estimated remaining time: 47h 18m 37s. Estimated total time: 52h 53m 9s. Time estimates for 10 more iterations: 10m 34s, 100 more iterations: 1h 45m 46s, 500 more iterations: 8h 48m 51s. +[2026-03-26 02:40:12,025][__main__][INFO] - Starting iteration 365. +[2026-03-26 02:40:12,426][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 36 and human policies 1. +[2026-03-26 02:40:12,426][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:40:34,572][__main__][INFO] - Number of regex retries in iteration 365: 0 +[2026-03-26 02:40:34,573][__main__][INFO] - agents played in iteration 365 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:40:35,337][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:40:35,362][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:40:35,385][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:40:35,409][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:40:35,409][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:40:35,410][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:40:36,108][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:40:36,564][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:40:37,068][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:40:37,568][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:40:38,074][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:40:38,573][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:40:39,071][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:40:39,569][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:40:40,708][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:40:41,475][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:40:41,981][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:40:42,485][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:40:42,984][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:40:43,482][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:40:43,986][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:40:44,484][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:40:44,982][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:40:45,482][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:40:45,984][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:40:46,484][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:40:46,988][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:40:47,492][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:40:47,992][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:40:48,491][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:40:48,991][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:40:49,488][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:40:49,989][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:40:50,487][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:40:50,988][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:40:51,486][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:40:51,983][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:40:52,480][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:40:52,977][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:40:53,478][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:40:53,975][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:40:54,473][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:40:54,971][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:40:55,471][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:40:55,968][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:40:56,466][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:40:56,964][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:40:57,461][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:40:57,960][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:40:58,460][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:40:58,956][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:40:59,454][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:40:59,953][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:41:00,451][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:41:00,950][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:41:01,447][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:41:01,945][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:41:02,444][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:41:02,942][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:41:03,439][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:41:03,936][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:41:04,433][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:41:04,929][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:41:05,425][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:41:05,922][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:41:06,419][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:41:06,916][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:41:07,414][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:41:07,914][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:41:08,413][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:41:08,914][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10201 tokens. +[2026-03-26 02:41:09,768][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 60.53%, Block Peak % of device VRAM: 62.14%, ΔTime: 00:00:33 +[2026-03-26 02:41:10,467][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:41:10,469][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:41:10,471][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:41:11,108][__main__][INFO] - Iteration 366 took 58s (37.74% Gen, 61.17% Train). Generation: 22s, Training: 35s. Estimated remaining time: 43h 18m 36s. Estimated total time: 48h 54m 7s. Time estimates for 10 more iterations: 9m 46s, 100 more iterations: 1h 37m 48s, 500 more iterations: 8h 9m 1s. +[2026-03-26 02:41:11,118][__main__][INFO] - Starting iteration 366. +[2026-03-26 02:41:11,524][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 36 and human policies 1. +[2026-03-26 02:41:11,524][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:42:21,003][__main__][INFO] - Number of regex retries in iteration 366: 0 +[2026-03-26 02:42:21,004][__main__][INFO] - agents played in iteration 366 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:42:21,915][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:42:21,938][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:42:21,962][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:42:21,986][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:42:21,986][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:42:21,987][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:42:22,713][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:42:23,977][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:42:24,483][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:42:24,987][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:42:25,487][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:42:25,991][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:42:26,491][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:42:26,999][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:42:27,504][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:42:28,007][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:42:28,509][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:42:29,011][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:42:29,511][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:42:30,011][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:42:30,509][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:42:31,010][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:42:31,512][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:42:32,014][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:42:32,515][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:42:33,015][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:42:33,514][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:42:34,015][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:42:34,513][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:42:35,013][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:42:35,512][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:42:36,009][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:42:36,507][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:42:37,005][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:42:37,505][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:42:38,001][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:42:38,497][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:42:38,995][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:42:39,493][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:42:39,992][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:42:40,492][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:42:40,996][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:42:41,495][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:42:41,996][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:42:42,496][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:42:42,997][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:42:43,497][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:42:43,994][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:42:44,491][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:42:44,992][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:42:45,493][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:42:45,991][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:42:46,493][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:42:46,991][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:42:47,491][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:42:47,991][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:42:48,513][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:42:49,015][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:42:49,515][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:42:50,016][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:42:50,517][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:42:51,018][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:42:51,518][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:42:52,017][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:42:52,518][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:42:53,017][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:42:53,520][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:42:54,020][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:42:54,518][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:42:55,017][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:42:55,516][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10364 tokens. +[2026-03-26 02:42:56,240][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 60.65%, Block Peak % of device VRAM: 62.45%, ΔTime: 00:00:33 +[2026-03-26 02:42:57,126][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:42:57,131][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:42:57,134][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:42:57,859][__main__][INFO] - Iteration 367 took 1m 46s (65.34% Gen, 33.98% Train). Generation: 1m 9s, Training: 36s. Estimated remaining time: 82h 59m 29s. Estimated total time: 88h 36m 46s. Time estimates for 10 more iterations: 17m 43s, 100 more iterations: 2h 57m 13s, 500 more iterations: 14h 46m 7s. +[2026-03-26 02:42:57,861][__main__][INFO] - Starting iteration 367. +[2026-03-26 02:42:58,260][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 36 and human policies 1. +[2026-03-26 02:42:58,260][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:43:55,161][__main__][INFO] - Number of regex retries in iteration 367: 0 +[2026-03-26 02:43:55,162][__main__][INFO] - agents played in iteration 367 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:43:55,935][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:43:55,961][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:43:55,986][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:43:56,010][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:43:56,011][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:43:56,011][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:43:56,706][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:43:57,162][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:43:57,666][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:43:58,165][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:43:58,662][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:43:59,159][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:43:59,661][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:44:00,159][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:44:00,663][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:44:01,182][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:44:01,691][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:44:02,198][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:44:02,698][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:44:03,199][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:44:03,700][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:44:04,200][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:44:04,699][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:44:05,203][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:44:05,704][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:44:06,204][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:44:06,704][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:44:07,204][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:44:07,711][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:44:08,212][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:44:08,710][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:44:09,206][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:44:09,702][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:44:10,200][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:44:10,696][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:44:11,194][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:44:11,692][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:44:12,190][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:44:12,687][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:44:13,184][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:44:13,683][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:44:14,181][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:44:14,677][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:44:15,177][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:44:15,674][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:44:16,172][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:44:16,672][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:44:17,171][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:44:17,670][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:44:18,164][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:44:18,663][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:44:19,163][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:44:19,657][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:44:20,154][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:44:20,653][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:44:21,150][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:44:21,646][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:44:22,146][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:44:22,645][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:44:23,143][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:44:23,638][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:44:24,135][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:44:24,635][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:44:25,132][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:44:25,630][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:44:26,128][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:44:26,628][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:44:27,124][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:44:27,623][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:44:28,120][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:44:28,617][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10227 tokens. +[2026-03-26 02:44:29,351][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 60.56%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:32 +[2026-03-26 02:44:30,098][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:44:30,100][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:44:30,101][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:44:30,858][__main__][INFO] - Iteration 368 took 1m 32s (61.45% Gen, 37.73% Train). Generation: 56s, Training: 34s. Estimated remaining time: 71h 31m 6s. Estimated total time: 77h 9m 56s. Time estimates for 10 more iterations: 15m 25s, 100 more iterations: 2h 34m 19s, 500 more iterations: 12h 51m 39s. +[2026-03-26 02:44:30,860][__main__][INFO] - Starting iteration 368. +[2026-03-26 02:44:31,260][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 36 and human policies 1. +[2026-03-26 02:44:31,261][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:45:06,974][__main__][INFO] - Number of regex retries in iteration 368: 0 +[2026-03-26 02:45:06,975][__main__][INFO] - agents played in iteration 368 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:45:07,865][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:45:07,889][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:45:07,912][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:45:07,936][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:45:07,936][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:45:07,937][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:45:08,657][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:45:09,253][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:45:09,771][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:45:10,271][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:45:10,773][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:45:11,272][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:45:11,772][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:45:12,269][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:45:12,772][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:45:13,275][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:45:13,780][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:45:14,279][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:45:14,776][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:45:15,274][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:45:15,770][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:45:16,266][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:45:16,762][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:45:17,261][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:45:17,757][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:45:18,263][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:45:18,764][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:45:19,263][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:45:19,761][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:45:20,259][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:45:20,756][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:45:21,257][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:45:21,757][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:45:22,257][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:45:22,756][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:45:23,256][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:45:23,755][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:45:24,255][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:45:24,755][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:45:25,259][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:45:25,758][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:45:26,255][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:45:26,752][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:45:27,249][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:45:27,746][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:45:28,248][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:45:29,214][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:45:30,104][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:45:30,602][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:45:31,101][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:45:31,600][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:45:32,100][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:45:32,600][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:45:33,101][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:45:33,600][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:45:34,097][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:45:34,594][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:45:35,090][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:45:35,587][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:45:36,081][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:45:36,577][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:45:37,081][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:45:37,584][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:45:38,085][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:45:38,582][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:45:39,080][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:45:39,578][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:45:40,076][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:45:40,574][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:45:41,073][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:45:41,576][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10073 tokens. +[2026-03-26 02:45:42,436][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 60.60%, Block Peak % of device VRAM: 62.11%, ΔTime: 00:00:33 +[2026-03-26 02:45:43,216][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:45:43,219][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:45:43,220][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:45:43,942][__main__][INFO] - Iteration 369 took 1m 12s (49.14% Gen, 49.87% Train). Generation: 35s, Training: 36s. Estimated remaining time: 54h 54m 2s. Estimated total time: 60h 34m 6s. Time estimates for 10 more iterations: 12m 6s, 100 more iterations: 2h 1m 8s, 500 more iterations: 10h 5m 41s. +[2026-03-26 02:45:43,994][__main__][INFO] - Starting iteration 369. +[2026-03-26 02:45:44,399][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 36 and human policies 1. +[2026-03-26 02:45:44,400][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:46:22,836][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 02:46:27,718][__main__][INFO] - Number of regex retries in iteration 369: 1 +[2026-03-26 02:46:27,718][__main__][INFO] - agents played in iteration 369 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:46:28,479][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:46:28,504][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:46:28,528][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:46:28,552][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:46:28,553][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:46:28,553][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:46:29,247][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:46:29,699][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:46:30,210][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:46:30,708][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:46:31,207][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:46:31,710][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:46:32,209][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:46:32,706][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:46:33,204][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:46:33,704][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:46:34,200][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:46:34,699][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:46:35,197][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:46:35,695][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:46:36,193][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:46:36,693][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:46:37,191][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:46:37,688][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:46:38,194][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:46:38,694][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:46:39,201][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:46:39,699][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:46:40,199][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:46:40,697][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:46:41,200][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:46:41,697][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:46:42,196][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:46:42,697][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:46:43,195][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:46:43,692][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:46:44,193][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:46:44,693][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:46:45,191][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:46:45,692][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:46:46,193][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:46:46,697][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:46:47,197][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:46:47,697][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:46:48,198][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:46:48,698][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:46:49,199][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:46:49,699][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:46:50,197][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:46:50,693][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:46:51,195][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:46:51,695][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:46:52,195][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:46:52,693][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:46:53,195][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:46:53,694][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:46:54,193][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:46:54,693][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:46:55,193][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:46:55,693][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:46:56,193][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:46:56,692][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:46:57,191][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:46:57,685][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:46:58,181][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:46:58,676][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:46:59,175][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:46:59,672][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:47:00,170][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:47:00,666][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:47:01,161][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10195 tokens. +[2026-03-26 02:47:01,900][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 60.50%, Block Peak % of device VRAM: 62.33%, ΔTime: 00:00:32 +[2026-03-26 02:47:02,654][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:47:02,656][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:47:02,658][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:47:03,384][__main__][INFO] - Iteration 370 took 1m 18s (54.84% Gen, 44.23% Train). Generation: 43s, Training: 34s. Estimated remaining time: 60h 7m 53s. Estimated total time: 65h 49m 16s. Time estimates for 10 more iterations: 13m 9s, 100 more iterations: 2h 11m 38s, 500 more iterations: 10h 58m 12s. +[2026-03-26 02:47:03,386][__main__][INFO] - Starting iteration 370. +[2026-03-26 02:47:03,786][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 36 and human policies 1. +[2026-03-26 02:47:03,787][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:47:24,551][__main__][INFO] - Number of regex retries in iteration 370: 0 +[2026-03-26 02:47:24,552][__main__][INFO] - agents played in iteration 370 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:47:25,309][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:47:25,333][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:47:25,356][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:47:25,379][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:47:25,379][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:47:25,380][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:47:26,081][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:47:26,532][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:47:27,046][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:47:27,557][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:47:28,059][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:47:28,558][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:47:29,058][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:47:29,560][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:47:30,068][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:47:30,573][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:47:31,073][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:47:31,574][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:47:32,076][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:47:32,575][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:47:33,080][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:47:33,584][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:47:34,083][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:47:34,581][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:47:35,077][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:47:35,576][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:47:36,072][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:47:36,571][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:47:37,069][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:47:37,574][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:47:38,078][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:47:38,579][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:47:39,081][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:47:39,581][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:47:40,081][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:47:40,587][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:47:41,092][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:47:41,594][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:47:42,094][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:47:42,601][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:47:43,103][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:47:43,607][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:47:44,109][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:47:44,606][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:47:45,111][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:47:45,611][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:47:46,115][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:47:46,613][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:47:47,111][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:47:47,612][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:47:48,118][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:47:48,621][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:47:49,120][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:47:49,620][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:47:50,118][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:47:50,619][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:47:51,119][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:47:51,620][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:47:52,120][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:47:52,620][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:47:53,121][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:47:53,623][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:47:54,124][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:47:54,624][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:47:55,125][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:47:55,630][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:47:56,138][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:47:56,640][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:47:57,141][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:47:57,643][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:47:58,146][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10261 tokens. +[2026-03-26 02:47:58,904][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 60.63%, Block Peak % of device VRAM: 62.36%, ΔTime: 00:00:32 +[2026-03-26 02:47:59,661][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:47:59,663][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:47:59,665][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:48:00,996][__main__][INFO] - Iteration 371 took 57s (36.30% Gen, 61.38% Train). Generation: 20s, Training: 35s. Estimated remaining time: 41h 58m 9s. Estimated total time: 47h 40m 29s. Time estimates for 10 more iterations: 9m 32s, 100 more iterations: 1h 35m 20s, 500 more iterations: 7h 56m 44s. +[2026-03-26 02:48:00,998][__main__][INFO] - Starting iteration 371. +[2026-03-26 02:48:01,398][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 37 and human policies 1. +[2026-03-26 02:48:01,398][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:48:07,863][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 02:48:15,317][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 02:48:19,789][__main__][INFO] - Number of regex retries in iteration 371: 2 +[2026-03-26 02:48:19,790][__main__][INFO] - agents played in iteration 371 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:48:20,549][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:48:20,572][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:48:20,596][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:48:20,619][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:48:20,619][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:48:20,620][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:48:21,320][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:48:21,862][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:48:22,434][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:48:23,489][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:48:23,988][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:48:24,486][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:48:24,992][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:48:25,494][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:48:26,004][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:48:26,505][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:48:27,006][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:48:27,510][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:48:28,012][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:48:28,513][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:48:29,017][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:48:29,520][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:48:30,019][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:48:30,517][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:48:31,021][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:48:31,523][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:48:32,026][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:48:32,526][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:48:33,025][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:48:33,522][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:48:34,022][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:48:34,521][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:48:35,021][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:48:35,519][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:48:36,019][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:48:36,519][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:48:37,019][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:48:37,525][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:48:38,027][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:48:38,524][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:48:39,025][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:48:39,523][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:48:40,024][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:48:40,525][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:48:41,025][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:48:41,528][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:48:42,029][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:48:42,526][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:48:43,024][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:48:43,522][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:48:44,023][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:48:44,520][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:48:45,019][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:48:45,516][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:48:46,014][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:48:46,514][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:48:47,016][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:48:47,517][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:48:48,018][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:48:48,519][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:48:49,021][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:48:49,522][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:48:50,025][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:48:50,523][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:48:51,020][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:48:51,518][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:48:52,017][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:48:52,514][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:48:53,013][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:48:53,511][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:48:54,011][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10323 tokens. +[2026-03-26 02:48:55,034][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 60.54%, Block Peak % of device VRAM: 62.33%, ΔTime: 00:00:33 +[2026-03-26 02:48:55,686][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:48:55,688][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:48:55,690][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:48:56,339][__main__][INFO] - Iteration 372 took 54s (33.47% Gen, 65.34% Train). Generation: 18s, Training: 35s. Estimated remaining time: 40h 3m 48s. Estimated total time: 45h 47m 4s. Time estimates for 10 more iterations: 9m 9s, 100 more iterations: 1h 31m 34s, 500 more iterations: 7h 37m 50s. +[2026-03-26 02:48:56,360][__main__][INFO] - Starting iteration 372. +[2026-03-26 02:48:56,766][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 37 and human policies 1. +[2026-03-26 02:48:56,767][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:49:59,815][__main__][INFO] - Number of regex retries in iteration 372: 0 +[2026-03-26 02:49:59,816][__main__][INFO] - agents played in iteration 372 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:50:00,681][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:50:00,705][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:50:00,728][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:50:00,751][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:50:00,752][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:50:00,752][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:50:01,493][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:50:01,964][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:50:02,479][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:50:02,981][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:50:03,482][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:50:03,984][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:50:04,484][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:50:04,993][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:50:05,495][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:50:05,994][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:50:06,496][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:50:06,995][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:50:07,496][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:50:08,004][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:50:08,504][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:50:09,006][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:50:09,509][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:50:10,008][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:50:10,510][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:50:11,014][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:50:11,518][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:50:12,025][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:50:12,523][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:50:13,021][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:50:13,520][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:50:14,025][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:50:14,524][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:50:15,033][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:50:15,537][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:50:16,037][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:50:16,541][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:50:17,040][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:50:17,539][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:50:18,036][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:50:18,534][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:50:19,031][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:50:19,527][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:50:20,028][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:50:20,534][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:50:21,033][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:50:21,531][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:50:22,030][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:50:22,527][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:50:23,028][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:50:23,525][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:50:24,023][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:50:24,520][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:50:25,018][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:50:25,520][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:50:26,020][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:50:26,518][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:50:27,019][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:50:27,517][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:50:28,017][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:50:28,516][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:50:29,016][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:50:29,515][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:50:30,016][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:50:30,513][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:50:31,012][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:50:31,512][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:50:32,011][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:50:32,512][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:50:33,028][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:50:33,525][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10228 tokens. +[2026-03-26 02:50:34,260][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.23%, ΔTime: 00:00:32 +[2026-03-26 02:50:35,004][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:50:35,006][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:50:35,008][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:50:35,731][__main__][INFO] - Iteration 373 took 1m 38s (63.71% Gen, 35.56% Train). Generation: 1m 3s, Training: 35s. Estimated remaining time: 76h 43m 20s. Estimated total time: 82h 28m 15s. Time estimates for 10 more iterations: 16m 29s, 100 more iterations: 2h 44m 56s, 500 more iterations: 13h 44m 42s. +[2026-03-26 02:50:35,732][__main__][INFO] - Starting iteration 373. +[2026-03-26 02:50:36,128][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 37 and human policies 1. +[2026-03-26 02:50:36,129][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:51:15,692][__main__][INFO] - Number of regex retries in iteration 373: 0 +[2026-03-26 02:51:15,693][__main__][INFO] - agents played in iteration 373 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:51:16,453][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:51:16,477][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:51:16,501][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:51:16,526][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:51:16,526][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:51:16,527][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:51:17,224][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:51:17,687][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:51:18,187][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:51:18,683][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:51:19,180][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:51:19,678][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:51:20,178][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:51:20,675][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:51:21,172][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:51:21,668][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:51:22,164][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:51:22,660][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:51:23,158][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:51:23,656][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:51:24,153][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:51:24,649][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:51:25,155][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:51:25,654][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:51:26,160][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:51:26,658][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:51:27,160][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:51:27,657][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:51:28,158][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:51:28,658][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:51:29,154][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:51:29,651][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:51:30,153][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:51:30,650][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:51:31,147][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:51:31,644][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:51:32,142][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:51:32,642][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:51:33,138][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:51:33,636][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:51:34,140][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:51:34,640][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:51:35,138][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:51:35,635][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:51:36,133][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:51:36,632][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:51:37,129][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:51:37,626][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:51:38,122][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:51:38,621][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:51:39,118][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:51:39,615][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:51:40,115][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:51:40,614][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:51:41,112][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:51:41,608][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:51:42,103][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:51:42,599][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:51:43,096][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:51:43,594][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:51:44,093][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:51:44,590][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:51:45,088][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:51:45,586][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:51:46,082][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:51:46,577][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:51:47,073][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:51:47,568][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:51:48,065][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:51:48,560][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:51:49,057][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10129 tokens. +[2026-03-26 02:51:49,798][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 60.51%, Block Peak % of device VRAM: 62.08%, ΔTime: 00:00:32 +[2026-03-26 02:51:50,557][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:51:50,559][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:51:50,560][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:51:51,287][__main__][INFO] - Iteration 374 took 1m 15s (52.64% Gen, 46.39% Train). Generation: 39s, Training: 34s. Estimated remaining time: 56h 51m 47s. Estimated total time: 62h 37m 58s. Time estimates for 10 more iterations: 12m 31s, 100 more iterations: 2h 5m 15s, 500 more iterations: 10h 26m 19s. +[2026-03-26 02:51:51,289][__main__][INFO] - Starting iteration 374. +[2026-03-26 02:51:51,689][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 37 and human policies 1. +[2026-03-26 02:51:51,690][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:52:43,870][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 1 z ball did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 02:52:47,905][__main__][INFO] - Number of regex retries in iteration 374: 1 +[2026-03-26 02:52:47,906][__main__][INFO] - agents played in iteration 374 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:52:48,769][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:52:48,793][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:52:48,816][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:52:48,839][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:52:48,840][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:52:48,840][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:52:49,602][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:52:50,297][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:52:50,805][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:52:51,305][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:52:51,805][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:52:52,305][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:52:52,820][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:52:53,324][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:52:53,821][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:52:54,319][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:52:54,819][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:52:55,324][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:52:55,828][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:52:56,326][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:52:56,824][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:52:57,321][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:52:57,819][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:52:58,320][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:52:58,830][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:52:59,329][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:52:59,831][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:53:00,330][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:53:00,831][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:53:01,329][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:53:01,829][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:53:02,330][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:53:02,828][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:53:03,327][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:53:03,848][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:53:04,353][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:53:04,852][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:53:05,354][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:53:05,853][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:53:06,351][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:53:06,848][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:53:07,345][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:53:07,841][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:53:08,338][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:53:08,835][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:53:09,334][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:53:09,830][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:53:10,333][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:53:10,833][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:53:11,331][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:53:11,831][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:53:12,330][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:53:12,828][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:53:13,327][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:53:13,825][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:53:14,326][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:53:14,821][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:53:15,321][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:53:15,820][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:53:16,319][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:53:16,817][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:53:17,317][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:53:17,819][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:53:18,318][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:53:18,818][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:53:19,317][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:53:19,814][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:53:20,313][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:53:20,809][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:53:21,309][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:53:21,811][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10353 tokens. +[2026-03-26 02:53:22,594][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 60.65%, Block Peak % of device VRAM: 62.27%, ΔTime: 00:00:33 +[2026-03-26 02:53:23,299][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:53:23,303][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:53:23,305][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:53:23,971][__main__][INFO] - Iteration 375 took 1m 32s (60.92% Gen, 38.36% Train). Generation: 56s, Training: 35s. Estimated remaining time: 71h 6m 24s. Estimated total time: 76h 54m 8s. Time estimates for 10 more iterations: 15m 22s, 100 more iterations: 2h 33m 48s, 500 more iterations: 12h 49m 1s. +[2026-03-26 02:53:23,977][__main__][INFO] - Starting iteration 375. +[2026-03-26 02:53:24,378][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 37 and human policies 1. +[2026-03-26 02:53:24,378][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:54:23,080][__main__][INFO] - Number of regex retries in iteration 375: 0 +[2026-03-26 02:54:23,081][__main__][INFO] - agents played in iteration 375 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:54:23,842][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:54:23,866][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:54:23,889][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:54:23,913][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:54:23,913][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:54:23,914][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:54:24,619][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:54:25,074][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:54:25,573][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:54:26,085][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:54:26,584][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:54:27,083][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:54:27,581][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:54:28,079][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:54:28,577][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:54:29,075][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:54:29,573][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:54:30,073][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:54:30,575][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:54:31,074][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:54:31,571][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:54:32,074][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:54:32,582][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:54:33,081][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:54:33,580][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:54:34,077][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:54:34,582][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:54:35,080][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:54:35,579][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:54:36,080][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:54:36,580][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:54:37,081][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:54:37,580][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:54:38,076][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:54:38,577][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:54:39,077][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:54:39,575][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:54:40,075][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:54:40,575][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:54:41,072][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:54:41,569][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:54:42,065][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:54:42,562][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:54:43,057][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:54:43,554][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:54:44,050][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:54:44,546][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:54:45,046][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:54:45,544][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:54:46,043][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:54:46,543][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:54:47,042][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:54:47,540][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:54:48,039][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:54:48,538][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:54:49,037][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:54:49,534][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:54:50,031][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:54:50,527][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:54:51,026][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:54:51,525][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:54:52,025][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:54:52,520][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:54:53,018][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:54:53,516][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:54:54,016][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:54:54,515][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:54:55,014][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:54:55,512][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:54:56,010][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:54:56,509][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10183 tokens. +[2026-03-26 02:54:57,258][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 60.53%, Block Peak % of device VRAM: 62.24%, ΔTime: 00:00:32 +[2026-03-26 02:54:57,965][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:54:57,967][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:54:57,969][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:54:58,739][__main__][INFO] - Iteration 376 took 1m 34s (62.21% Gen, 36.97% Train). Generation: 58s, Training: 34s. Estimated remaining time: 72h 48m 46s. Estimated total time: 78h 38m 4s. Time estimates for 10 more iterations: 15m 43s, 100 more iterations: 2h 37m 16s, 500 more iterations: 13h 6m 20s. +[2026-03-26 02:54:58,741][__main__][INFO] - Starting iteration 376. +[2026-03-26 02:54:59,142][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 37 and human policies 1. +[2026-03-26 02:54:59,143][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:55:32,575][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 02:55:41,133][__main__][INFO] - Number of regex retries in iteration 376: 1 +[2026-03-26 02:55:41,133][__main__][INFO] - agents played in iteration 376 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:55:41,893][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:55:41,917][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:55:41,941][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:55:41,965][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:55:41,965][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:55:41,966][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:55:42,689][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:55:43,150][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:55:43,657][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:55:44,154][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:55:44,656][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:55:45,168][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:55:45,667][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:55:46,166][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:55:46,664][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:55:47,167][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:55:47,668][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:55:48,183][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:55:48,689][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:55:49,193][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:55:49,695][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:55:50,196][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:55:50,697][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:55:51,204][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:55:51,711][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:55:52,209][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:55:52,708][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:55:53,207][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:55:53,705][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:55:54,203][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:55:54,705][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:55:55,204][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:55:55,702][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:55:56,213][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:55:56,711][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:55:57,216][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:55:57,715][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:55:58,214][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:55:58,712][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:55:59,209][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:55:59,707][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:56:00,203][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:56:00,701][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:56:01,201][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:56:01,706][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:56:02,209][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:56:02,707][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:56:03,206][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:56:03,705][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:56:04,204][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:56:04,704][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:56:05,201][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:56:05,702][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:56:06,199][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:56:06,698][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:56:07,196][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:56:07,698][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:56:08,198][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:56:08,695][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:56:09,196][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:56:09,693][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:56:10,193][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:56:10,689][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:56:11,190][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:56:11,697][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:56:12,204][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:56:12,703][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:56:13,200][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:56:13,702][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:56:14,200][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:56:14,698][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10217 tokens. +[2026-03-26 02:56:15,448][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.01%, Current % of VRAM taken: 60.44%, Block Peak % of device VRAM: 62.33%, ΔTime: 00:00:32 +[2026-03-26 02:56:16,175][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:56:16,178][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:56:16,179][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:56:16,884][__main__][INFO] - Iteration 377 took 1m 17s (54.01% Gen, 45.08% Train). Generation: 41s, Training: 35s. Estimated remaining time: 58h 56m 29s. Estimated total time: 64h 47m 6s. Time estimates for 10 more iterations: 12m 57s, 100 more iterations: 2h 9m 34s, 500 more iterations: 10h 47m 51s. +[2026-03-26 02:56:16,886][__main__][INFO] - Starting iteration 377. +[2026-03-26 02:56:17,286][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 37 and human policies 1. +[2026-03-26 02:56:17,287][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:56:42,164][__main__][INFO] - Number of regex retries in iteration 377: 0 +[2026-03-26 02:56:42,165][__main__][INFO] - agents played in iteration 377 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:56:42,923][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.44%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:56:42,948][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.44%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:56:42,973][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.44%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:56:42,997][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.44%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:56:42,998][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:56:42,998][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:56:43,705][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:56:44,160][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:56:44,666][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:56:45,164][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:56:45,663][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:56:46,161][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:56:46,660][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:56:47,157][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:56:47,655][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:56:48,153][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:56:48,651][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:56:49,149][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:56:49,647][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:56:50,144][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:56:50,642][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:56:51,140][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:56:51,640][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:56:52,137][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:56:52,648][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:56:53,154][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:56:53,659][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:56:54,160][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:56:54,658][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:56:55,157][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:56:55,659][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:56:56,159][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:56:56,682][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:56:57,186][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:56:57,689][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:56:58,192][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:56:58,690][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:56:59,190][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:56:59,691][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:57:00,186][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:57:00,682][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:57:01,178][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:57:01,675][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:57:02,176][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:57:02,674][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:57:03,171][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:57:03,667][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:57:04,162][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:57:04,662][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:57:05,158][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:57:05,657][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:57:06,158][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:57:06,656][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:57:07,154][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:57:07,650][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:57:08,146][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:57:08,641][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:57:09,137][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:57:09,634][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:57:10,129][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:57:10,625][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:57:11,121][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:57:11,619][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:57:12,119][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:57:12,616][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:57:13,113][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:57:13,610][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:57:14,106][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:57:14,603][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:57:15,100][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:57:15,597][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10192 tokens. +[2026-03-26 02:57:16,346][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 60.57%, Block Peak % of device VRAM: 62.18%, ΔTime: 00:00:32 +[2026-03-26 02:57:17,073][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:57:17,075][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:57:17,077][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:57:17,774][__main__][INFO] - Iteration 378 took 1m 0s (41.13% Gen, 57.72% Train). Generation: 24s, Training: 34s. Estimated remaining time: 44h 32m 48s. Estimated total time: 50h 24m 25s. Time estimates for 10 more iterations: 10m 4s, 100 more iterations: 1h 40m 48s, 500 more iterations: 8h 24m 4s. +[2026-03-26 02:57:17,776][__main__][INFO] - Starting iteration 378. +[2026-03-26 02:57:18,178][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 37 and human policies 1. +[2026-03-26 02:57:18,178][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:57:31,307][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 02:57:42,702][__main__][INFO] - Number of regex retries in iteration 378: 1 +[2026-03-26 02:57:42,702][__main__][INFO] - agents played in iteration 378 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:57:43,464][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:57:43,489][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:57:43,513][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:57:43,538][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:57:43,538][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:57:43,539][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:57:44,235][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:57:44,691][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:57:45,192][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:57:45,689][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:57:46,186][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:57:46,686][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:57:47,183][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:57:47,679][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:57:48,176][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:57:48,680][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:57:49,199][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:57:49,700][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:57:50,203][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:57:50,704][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:57:51,204][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:57:51,703][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:57:52,202][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:57:52,699][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:57:53,197][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:57:53,696][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:57:54,198][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:57:54,695][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:57:55,193][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:57:55,690][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:57:56,192][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:57:56,695][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:57:57,197][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:57:57,699][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:57:58,201][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:57:58,708][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:57:59,210][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:57:59,720][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:58:00,223][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:58:00,720][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:58:01,221][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:58:01,719][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:58:02,219][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:58:02,716][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:58:03,216][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:58:03,713][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:58:04,211][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:58:04,711][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:58:05,210][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:58:05,706][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:58:06,209][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:58:06,711][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:58:07,212][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:58:07,712][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:58:08,211][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:58:08,706][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:58:09,202][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:58:09,698][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:58:10,197][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:58:10,694][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:58:11,190][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:58:11,687][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:58:12,182][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:58:12,679][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:58:13,177][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:58:13,673][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:58:14,173][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:58:14,669][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:58:15,170][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:58:15,667][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:58:16,164][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10183 tokens. +[2026-03-26 02:58:16,909][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 60.54%, Block Peak % of device VRAM: 62.44%, ΔTime: 00:00:32 +[2026-03-26 02:58:17,642][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:58:17,644][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:58:17,645][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:58:18,356][__main__][INFO] - Iteration 379 took 1m 0s (40.75% Gen, 58.06% Train). Generation: 24s, Training: 34s. Estimated remaining time: 44h 16m 19s. Estimated total time: 50h 8m 57s. Time estimates for 10 more iterations: 10m 1s, 100 more iterations: 1h 40m 17s, 500 more iterations: 8h 21m 29s. +[2026-03-26 02:58:18,364][__main__][INFO] - Starting iteration 379. +[2026-03-26 02:58:18,765][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 37 and human policies 1. +[2026-03-26 02:58:18,766][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:58:35,371][__main__][INFO] - Number of regex retries in iteration 379: 0 +[2026-03-26 02:58:35,372][__main__][INFO] - agents played in iteration 379 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:58:36,138][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:58:36,161][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:58:36,185][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:58:36,208][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:58:36,209][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:58:36,209][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:58:36,913][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:58:37,366][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:58:37,868][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:58:38,365][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:58:38,864][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:58:39,360][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:58:39,856][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:58:40,358][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:58:40,852][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:58:41,349][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:58:41,844][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:58:42,346][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:58:42,845][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:58:43,351][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:58:43,849][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:58:44,347][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:58:44,845][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:58:45,351][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:58:45,875][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:58:46,378][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:58:46,879][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:58:47,379][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:58:47,880][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:58:48,378][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:58:48,879][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:58:49,376][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:58:49,872][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:58:50,369][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:58:50,865][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:58:51,360][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:58:51,858][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:58:52,355][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:58:52,853][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:58:53,351][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:58:53,846][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:58:54,339][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:58:54,836][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:58:55,331][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:58:55,826][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:58:56,322][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:58:56,818][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:58:57,325][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:58:57,825][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:58:58,328][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:58:58,828][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:58:59,330][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:58:59,827][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:59:00,327][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:59:00,829][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:59:01,330][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:59:01,831][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:59:02,331][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:59:02,833][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:59:03,333][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:59:03,834][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:59:04,337][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:59:04,838][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:59:05,337][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:59:05,836][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:59:06,335][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:59:06,835][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 02:59:07,335][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 02:59:07,833][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 02:59:08,332][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 02:59:08,832][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10203 tokens. +[2026-03-26 02:59:09,587][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 60.61%, Block Peak % of device VRAM: 62.40%, ΔTime: 00:00:32 +[2026-03-26 02:59:10,312][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 02:59:10,315][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 02:59:10,316][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 02:59:11,026][__main__][INFO] - Iteration 380 took 52s (31.78% Gen, 66.86% Train). Generation: 16s, Training: 34s. Estimated remaining time: 37h 39m 33s. Estimated total time: 43h 33m 4s. Time estimates for 10 more iterations: 8m 42s, 100 more iterations: 1h 27m 6s, 500 more iterations: 7h 15m 30s. +[2026-03-26 02:59:11,028][__main__][INFO] - Starting iteration 380. +[2026-03-26 02:59:11,428][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 37 and human policies 1. +[2026-03-26 02:59:11,428][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 02:59:28,233][__main__][INFO] - Number of regex retries in iteration 380: 0 +[2026-03-26 02:59:28,234][__main__][INFO] - agents played in iteration 380 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 02:59:29,000][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:59:29,023][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:59:29,047][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:59:29,070][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 02:59:29,070][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 02:59:29,071][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 02:59:29,767][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 02:59:30,231][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 02:59:30,739][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 02:59:31,239][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 02:59:31,743][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 02:59:32,243][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 02:59:32,753][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 02:59:33,254][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 02:59:33,754][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 02:59:34,252][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 02:59:34,751][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 02:59:35,253][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 02:59:35,755][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 02:59:36,255][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 02:59:36,754][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 02:59:37,252][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 02:59:37,750][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 02:59:38,248][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 02:59:38,749][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 02:59:39,246][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 02:59:39,743][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 02:59:40,241][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 02:59:40,739][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 02:59:41,237][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 02:59:41,739][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 02:59:42,238][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 02:59:42,739][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 02:59:43,242][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 02:59:43,742][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 02:59:44,242][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 02:59:44,740][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 02:59:45,238][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 02:59:45,736][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 02:59:46,232][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 02:59:46,727][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 02:59:47,224][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 02:59:47,733][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 02:59:48,234][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 02:59:48,734][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 02:59:49,232][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 02:59:49,731][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 02:59:50,230][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 02:59:50,729][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 02:59:51,232][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 02:59:51,738][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 02:59:52,235][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 02:59:52,733][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 02:59:53,234][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 02:59:53,730][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 02:59:54,227][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 02:59:54,725][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 02:59:55,223][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 02:59:55,727][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 02:59:56,230][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 02:59:56,728][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 02:59:57,228][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 02:59:57,726][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 02:59:58,224][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 02:59:58,722][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 02:59:59,219][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 02:59:59,718][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:00:00,217][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:00:00,719][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:00:01,227][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:00:01,733][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10087 tokens. +[2026-03-26 03:00:02,492][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.02%, Current % of VRAM taken: 60.45%, Block Peak % of device VRAM: 62.10%, ΔTime: 00:00:32 +[2026-03-26 03:00:03,230][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:00:03,232][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:00:03,234][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:00:04,518][__main__][INFO] - Iteration 381 took 53s (31.65% Gen, 65.92% Train). Generation: 16s, Training: 34s. Estimated remaining time: 38h 20m 7s. Estimated total time: 44h 14m 31s. Time estimates for 10 more iterations: 8m 50s, 100 more iterations: 1h 28m 29s, 500 more iterations: 7h 22m 25s. +[2026-03-26 03:00:04,520][__main__][INFO] - Starting iteration 381. +[2026-03-26 03:00:04,920][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 38 and human policies 1. +[2026-03-26 03:00:04,921][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:00:25,016][__main__][INFO] - Number of regex retries in iteration 381: 0 +[2026-03-26 03:00:25,017][__main__][INFO] - agents played in iteration 381 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:00:25,775][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:00:25,799][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:00:25,822][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:00:25,846][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:00:25,847][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:00:25,847][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:00:26,568][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:00:27,020][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:00:27,535][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:00:28,035][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:00:28,540][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:00:29,038][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:00:29,539][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:00:30,038][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:00:30,537][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:00:31,038][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:00:31,537][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:00:32,039][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:00:32,544][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:00:33,043][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:00:33,541][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:00:34,052][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:00:34,553][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:00:35,057][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:00:35,555][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:00:36,054][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:00:36,552][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:00:37,053][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:00:37,564][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:00:38,066][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:00:38,567][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:00:39,071][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:00:39,574][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:00:40,077][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:00:40,579][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:00:41,081][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:00:41,587][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:00:42,094][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:00:42,595][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:00:43,097][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:00:43,596][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:00:44,096][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:00:44,595][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:00:45,095][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:00:45,595][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:00:46,095][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:00:46,596][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:00:47,096][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:00:47,595][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:00:48,099][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:00:48,600][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:00:49,101][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:00:49,602][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:00:50,102][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:00:50,603][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:00:51,100][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:00:51,600][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:00:52,098][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:00:52,596][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:00:53,094][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:00:53,592][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:00:54,093][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:00:54,591][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:00:55,090][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:00:55,585][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:00:56,080][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:00:56,585][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:00:57,091][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:00:57,590][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:00:58,086][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:00:58,585][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10132 tokens. +[2026-03-26 03:00:59,353][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.01%, Current % of VRAM taken: 60.44%, Block Peak % of device VRAM: 62.06%, ΔTime: 00:00:32 +[2026-03-26 03:01:00,088][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:01:00,090][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:01:00,092][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:01:00,806][__main__][INFO] - Iteration 382 took 55s (35.96% Gen, 62.76% Train). Generation: 20s, Training: 35s. Estimated remaining time: 40h 38m 58s. Estimated total time: 46h 34m 19s. Time estimates for 10 more iterations: 9m 18s, 100 more iterations: 1h 33m 8s, 500 more iterations: 7h 45m 43s. +[2026-03-26 03:01:00,808][__main__][INFO] - Starting iteration 382. +[2026-03-26 03:01:01,208][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 38 and human policies 1. +[2026-03-26 03:01:01,209][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:01:12,533][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 03:01:17,543][__main__][INFO] - Number of regex retries in iteration 382: 1 +[2026-03-26 03:01:17,544][__main__][INFO] - agents played in iteration 382 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:01:18,301][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:01:18,324][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:01:18,348][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:01:18,371][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:01:18,371][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:01:18,372][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:01:19,099][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:01:19,555][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:01:20,066][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:01:20,578][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:01:21,082][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:01:21,583][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:01:22,086][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:01:22,589][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:01:23,105][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:01:23,608][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:01:24,109][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:01:24,612][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:01:25,114][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:01:25,615][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:01:26,128][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:01:26,631][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:01:27,135][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:01:27,634][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:01:28,132][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:01:28,630][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:01:29,127][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:01:29,624][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:01:30,135][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:01:30,635][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:01:31,138][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:01:31,636][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:01:32,136][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:01:32,636][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:01:33,139][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:01:33,638][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:01:34,137][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:01:34,639][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:01:35,142][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:01:35,645][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:01:36,144][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:01:36,647][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:01:37,146][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:01:37,645][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:01:38,143][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:01:38,640][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:01:39,138][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:01:39,638][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:01:40,138][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:01:40,638][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:01:41,139][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:01:41,645][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:01:42,153][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:01:42,656][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:01:43,158][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:01:43,660][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:01:44,159][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:01:44,657][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:01:45,156][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:01:45,655][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:01:46,156][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:01:46,654][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:01:47,152][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:01:47,651][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:01:48,149][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:01:48,650][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:01:49,147][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:01:49,652][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:01:50,160][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:01:50,659][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:01:51,158][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10149 tokens. +[2026-03-26 03:01:51,935][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.28%, ΔTime: 00:00:32 +[2026-03-26 03:01:52,667][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:01:52,670][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:01:52,671][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:01:53,427][__main__][INFO] - Iteration 383 took 52s (31.28% Gen, 67.27% Train). Generation: 16s, Training: 35s. Estimated remaining time: 37h 34m 45s. Estimated total time: 43h 30m 58s. Time estimates for 10 more iterations: 8m 42s, 100 more iterations: 1h 27m 1s, 500 more iterations: 7h 15m 9s. +[2026-03-26 03:01:53,429][__main__][INFO] - Starting iteration 383. +[2026-03-26 03:01:53,828][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 38 and human policies 1. +[2026-03-26 03:01:53,828][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:02:09,812][__main__][INFO] - Number of regex retries in iteration 383: 0 +[2026-03-26 03:02:09,813][__main__][INFO] - agents played in iteration 383 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:02:10,569][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:02:10,593][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:02:10,617][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:02:10,640][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:02:10,641][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:02:10,641][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:02:11,342][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:02:11,808][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:02:12,314][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:02:12,814][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:02:13,314][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:02:13,815][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:02:14,326][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:02:14,831][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:02:15,332][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:02:15,834][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:02:16,332][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:02:16,830][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:02:17,331][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:02:17,832][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:02:18,332][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:02:18,829][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:02:19,326][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:02:19,824][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:02:20,320][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:02:20,817][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:02:21,315][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:02:21,814][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:02:22,309][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:02:22,806][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:02:23,305][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:02:23,802][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:02:24,300][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:02:24,798][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:02:25,295][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:02:25,794][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:02:26,291][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:02:26,788][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:02:27,285][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:02:27,780][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:02:28,278][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:02:28,775][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:02:29,271][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:02:29,784][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:02:30,284][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:02:30,783][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:02:31,280][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:02:31,780][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:02:32,279][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:02:32,782][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:02:33,287][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:02:33,785][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:02:34,283][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:02:34,780][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:02:35,280][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:02:35,778][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:02:36,276][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:02:36,783][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:02:37,281][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:02:37,784][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:02:38,282][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:02:38,781][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:02:39,280][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:02:39,777][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:02:40,278][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:02:40,776][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:02:41,281][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:02:41,781][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:02:42,281][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:02:42,779][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:02:43,281][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10178 tokens. +[2026-03-26 03:02:44,019][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 60.65%, Block Peak % of device VRAM: 62.20%, ΔTime: 00:00:32 +[2026-03-26 03:02:44,793][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:02:44,796][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:02:44,798][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:02:45,497][__main__][INFO] - Iteration 384 took 51s (30.94% Gen, 67.71% Train). Generation: 15s, Training: 34s. Estimated remaining time: 37h 6m 23s. Estimated total time: 43h 3m 28s. Time estimates for 10 more iterations: 8m 36s, 100 more iterations: 1h 26m 6s, 500 more iterations: 7h 10m 34s. +[2026-03-26 03:02:45,499][__main__][INFO] - Starting iteration 384. +[2026-03-26 03:02:45,896][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 38 and human policies 1. +[2026-03-26 03:02:45,897][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:03:00,718][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 03:03:03,307][__main__][INFO] - Number of regex retries in iteration 384: 1 +[2026-03-26 03:03:03,308][__main__][INFO] - agents played in iteration 384 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:03:04,073][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:03:04,100][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:03:04,125][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:03:04,150][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:03:04,150][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:03:04,151][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:03:04,871][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:03:05,333][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:03:05,837][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:03:06,340][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:03:06,842][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:03:07,341][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:03:07,839][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:03:08,338][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:03:08,837][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:03:09,335][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:03:09,833][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:03:10,334][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:03:10,838][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:03:11,335][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:03:11,834][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:03:12,331][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:03:12,833][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:03:13,330][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:03:13,831][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:03:14,328][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:03:14,825][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:03:15,322][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:03:15,819][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:03:16,318][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:03:16,815][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:03:17,312][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:03:17,808][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:03:18,305][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:03:18,802][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:03:19,299][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:03:19,796][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:03:20,292][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:03:20,786][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:03:21,285][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:03:21,780][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:03:22,276][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:03:22,773][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:03:23,272][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:03:23,768][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:03:24,268][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:03:24,764][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:03:25,259][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:03:25,756][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:03:26,255][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:03:26,750][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:03:27,245][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:03:27,741][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:03:28,236][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:03:28,731][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:03:29,226][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:03:29,721][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:03:30,217][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:03:30,713][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:03:31,218][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:03:31,716][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:03:32,219][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:03:32,718][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:03:33,216][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:03:33,713][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:03:34,210][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:03:34,710][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:03:35,206][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:03:35,705][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:03:36,201][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:03:36,698][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10086 tokens. +[2026-03-26 03:03:37,435][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.09%, Current % of VRAM taken: 60.52%, Block Peak % of device VRAM: 61.94%, ΔTime: 00:00:32 +[2026-03-26 03:03:38,167][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:03:38,170][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:03:38,174][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:03:38,980][__main__][INFO] - Iteration 385 took 53s (32.80% Gen, 65.68% Train). Generation: 17s, Training: 34s. Estimated remaining time: 38h 16m 15s. Estimated total time: 44h 14m 14s. Time estimates for 10 more iterations: 8m 50s, 100 more iterations: 1h 28m 28s, 500 more iterations: 7h 22m 22s. +[2026-03-26 03:03:38,982][__main__][INFO] - Starting iteration 385. +[2026-03-26 03:03:39,380][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 38 and human policies 1. +[2026-03-26 03:03:39,380][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:03:55,742][__main__][INFO] - Number of regex retries in iteration 385: 0 +[2026-03-26 03:03:55,743][__main__][INFO] - agents played in iteration 385 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:03:56,512][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:03:56,536][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:03:56,559][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:03:56,582][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:03:56,583][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:03:56,583][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:03:57,292][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:03:57,745][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:03:58,256][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:03:58,762][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:03:59,263][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:03:59,762][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:04:00,260][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:04:00,760][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:04:01,258][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:04:01,758][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:04:02,258][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:04:02,763][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:04:03,261][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:04:03,759][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:04:04,259][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:04:04,760][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:04:05,262][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:04:05,764][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:04:06,265][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:04:06,760][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:04:07,257][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:04:07,754][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:04:08,250][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:04:08,755][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:04:09,259][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:04:09,758][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:04:10,258][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:04:10,757][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:04:11,255][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:04:11,753][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:04:12,253][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:04:12,752][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:04:13,249][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:04:13,750][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:04:14,247][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:04:14,745][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:04:15,245][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:04:15,742][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:04:16,243][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:04:16,740][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:04:17,241][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:04:17,736][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:04:18,230][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:04:18,727][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:04:19,222][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:04:19,717][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:04:20,210][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:04:20,704][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:04:21,205][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:04:21,710][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:04:22,207][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:04:22,704][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:04:23,202][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:04:23,700][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:04:24,200][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:04:24,698][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:04:25,201][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:04:25,701][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:04:26,200][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:04:26,699][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:04:27,205][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:04:27,705][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:04:28,207][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:04:28,706][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:04:29,209][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10135 tokens. +[2026-03-26 03:04:29,965][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 60.65%, Block Peak % of device VRAM: 62.15%, ΔTime: 00:00:32 +[2026-03-26 03:04:30,710][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:04:30,713][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:04:30,714][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:04:31,431][__main__][INFO] - Iteration 386 took 52s (31.43% Gen, 67.18% Train). Generation: 16s, Training: 34s. Estimated remaining time: 37h 23m 45s. Estimated total time: 43h 22m 36s. Time estimates for 10 more iterations: 8m 40s, 100 more iterations: 1h 26m 45s, 500 more iterations: 7h 13m 46s. +[2026-03-26 03:04:31,433][__main__][INFO] - Starting iteration 386. +[2026-03-26 03:04:31,838][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 38 and human policies 1. +[2026-03-26 03:04:31,839][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:04:48,639][__main__][INFO] - Number of regex retries in iteration 386: 0 +[2026-03-26 03:04:48,640][__main__][INFO] - agents played in iteration 386 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:04:49,499][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:04:49,523][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:04:49,546][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:04:49,569][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:04:49,570][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:04:49,570][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:04:50,293][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:04:50,754][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:04:51,260][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:04:51,760][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:04:52,265][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:04:52,777][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:04:53,278][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:04:53,780][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:04:54,281][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:04:54,782][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:04:55,283][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:04:55,787][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:04:56,291][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:04:56,790][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:04:57,288][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:04:57,787][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:04:58,288][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:04:58,786][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:04:59,290][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:04:59,797][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:05:00,298][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:05:00,798][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:05:01,297][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:05:01,800][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:05:02,301][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:05:02,801][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:05:03,299][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:05:03,798][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:05:04,302][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:05:04,804][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:05:05,306][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:05:05,807][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:05:06,305][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:05:06,806][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:05:07,309][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:05:07,808][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:05:08,308][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:05:08,815][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:05:09,316][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:05:09,819][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:05:10,325][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:05:10,826][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:05:11,324][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:05:11,821][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:05:12,320][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:05:12,818][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:05:13,316][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:05:13,817][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:05:14,313][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:05:14,814][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:05:15,311][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:05:15,814][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:05:16,312][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:05:16,810][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:05:17,309][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:05:17,816][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:05:18,314][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:05:18,816][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:05:19,316][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:05:19,818][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:05:20,318][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:05:20,816][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:05:21,313][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:05:21,809][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:05:22,305][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10230 tokens. +[2026-03-26 03:05:23,059][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 60.49%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 +[2026-03-26 03:05:23,813][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:05:23,815][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:05:23,817][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:05:24,650][__main__][INFO] - Iteration 387 took 52s (31.81% Gen, 66.61% Train). Generation: 16s, Training: 35s. Estimated remaining time: 38h 0m 53s. Estimated total time: 44h 0m 37s. Time estimates for 10 more iterations: 8m 48s, 100 more iterations: 1h 28m 1s, 500 more iterations: 7h 20m 6s. +[2026-03-26 03:05:24,652][__main__][INFO] - Starting iteration 387. +[2026-03-26 03:05:25,053][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 38 and human policies 1. +[2026-03-26 03:05:25,054][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:05:31,021][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 03:05:43,863][__main__][INFO] - Number of regex retries in iteration 387: 1 +[2026-03-26 03:05:43,863][__main__][INFO] - agents played in iteration 387 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:05:44,650][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:05:44,674][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:05:44,697][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:05:44,720][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:05:44,721][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:05:44,722][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:05:45,427][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:05:45,878][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:05:46,383][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:05:46,885][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:05:47,383][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:05:47,884][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:05:48,382][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:05:48,889][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:05:49,389][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:05:49,896][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:05:50,397][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:05:50,897][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:05:51,397][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:05:51,896][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:05:52,395][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:05:52,891][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:05:53,388][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:05:53,889][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:05:54,387][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:05:54,888][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:05:55,390][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:05:55,887][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:05:56,387][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:05:56,884][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:05:57,384][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:05:57,882][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:05:58,382][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:05:58,883][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:05:59,380][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:05:59,881][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:06:00,379][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:06:00,878][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:06:01,379][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:06:01,877][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:06:02,376][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:06:02,877][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:06:03,373][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:06:03,868][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:06:04,374][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:06:04,876][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:06:05,384][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:06:05,883][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:06:06,388][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:06:06,888][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:06:07,393][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:06:07,892][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:06:08,390][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:06:08,888][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:06:09,391][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:06:09,887][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:06:10,387][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:06:10,884][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:06:11,381][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:06:11,882][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:06:12,380][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:06:12,878][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:06:13,377][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:06:13,872][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:06:14,368][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:06:14,866][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:06:15,362][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:06:15,859][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:06:16,356][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:06:16,856][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:06:17,353][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10185 tokens. +[2026-03-26 03:06:18,091][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 60.49%, Block Peak % of device VRAM: 62.15%, ΔTime: 00:00:32 +[2026-03-26 03:06:18,833][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:06:18,835][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:06:18,837][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:06:19,672][__main__][INFO] - Iteration 388 took 54s (34.44% Gen, 64.03% Train). Generation: 18s, Training: 34s. Estimated remaining time: 39h 30m 19s. Estimated total time: 45h 30m 59s. Time estimates for 10 more iterations: 9m 6s, 100 more iterations: 1h 31m 1s, 500 more iterations: 7h 35m 9s. +[2026-03-26 03:06:19,674][__main__][INFO] - Starting iteration 388. +[2026-03-26 03:06:20,073][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 38 and human policies 1. +[2026-03-26 03:06:20,073][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:06:38,209][__main__][INFO] - Number of regex retries in iteration 388: 0 +[2026-03-26 03:06:38,210][__main__][INFO] - agents played in iteration 388 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:06:38,997][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:06:39,021][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:06:39,045][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:06:39,068][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:06:39,069][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:06:39,069][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:06:39,770][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:06:40,223][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:06:40,724][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:06:41,221][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:06:41,718][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:06:42,215][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:06:42,715][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:06:43,215][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:06:43,709][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:06:44,208][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:06:44,707][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:06:45,206][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:06:45,709][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:06:46,207][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:06:46,705][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:06:47,201][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:06:47,699][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:06:48,197][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:06:48,696][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:06:49,191][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:06:49,692][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:06:50,188][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:06:50,685][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:06:51,185][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:06:51,685][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:06:52,194][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:06:52,695][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:06:53,193][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:06:53,695][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:06:54,196][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:06:54,715][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:06:55,215][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:06:55,715][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:06:56,213][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:06:56,714][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:06:57,214][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:06:57,711][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:06:58,211][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:06:58,714][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:06:59,210][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:06:59,706][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:07:00,203][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:07:00,700][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:07:01,197][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:07:01,694][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:07:02,191][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:07:02,688][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:07:03,183][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:07:03,679][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:07:04,175][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:07:04,671][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:07:05,165][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:07:05,659][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:07:06,155][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:07:06,649][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:07:07,144][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:07:07,644][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:07:08,140][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:07:08,639][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:07:09,136][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:07:09,634][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:07:10,132][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:07:10,629][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:07:11,128][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:07:11,627][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10178 tokens. +[2026-03-26 03:07:12,363][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 60.63%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:32 +[2026-03-26 03:07:13,103][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:07:13,105][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:07:13,107][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:07:13,929][__main__][INFO] - Iteration 389 took 53s (33.67% Gen, 64.80% Train). Generation: 18s, Training: 34s. Estimated remaining time: 38h 51m 16s. Estimated total time: 44h 52m 50s. Time estimates for 10 more iterations: 8m 58s, 100 more iterations: 1h 29m 45s, 500 more iterations: 7h 28m 48s. +[2026-03-26 03:07:13,932][__main__][INFO] - Starting iteration 389. +[2026-03-26 03:07:14,329][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 38 and human policies 1. +[2026-03-26 03:07:14,330][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:07:30,568][__main__][INFO] - Number of regex retries in iteration 389: 0 +[2026-03-26 03:07:30,569][__main__][INFO] - agents played in iteration 389 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:07:31,336][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:07:31,359][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:07:31,383][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:07:31,407][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:07:31,407][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:07:31,408][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:07:32,115][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:07:32,568][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:07:33,070][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:07:33,580][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:07:34,080][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:07:34,580][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:07:35,083][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:07:35,582][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:07:36,081][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:07:36,585][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:07:37,092][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:07:37,601][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:07:38,106][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:07:38,608][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:07:39,109][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:07:39,606][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:07:40,108][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:07:40,607][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:07:41,105][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:07:41,604][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:07:42,103][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:07:42,603][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:07:43,102][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:07:43,604][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:07:44,102][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:07:44,599][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:07:45,097][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:07:45,596][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:07:46,095][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:07:46,593][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:07:47,091][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:07:47,588][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:07:48,087][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:07:48,583][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:07:49,082][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:07:49,579][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:07:50,084][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:07:50,589][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:07:51,087][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:07:51,590][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:07:52,088][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:07:52,588][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:07:53,086][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:07:53,587][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:07:54,085][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:07:54,585][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:07:55,105][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:07:55,608][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:07:56,106][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:07:56,604][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:07:57,104][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:07:57,607][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:07:58,104][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:07:58,605][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:07:59,102][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:07:59,608][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:08:00,107][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:08:00,608][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:08:01,108][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:08:01,607][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:08:02,108][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:08:02,605][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:08:03,103][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:08:03,603][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:08:04,100][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10115 tokens. +[2026-03-26 03:08:04,839][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.09%, Current % of VRAM taken: 60.52%, Block Peak % of device VRAM: 62.12%, ΔTime: 00:00:32 +[2026-03-26 03:08:05,575][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:08:05,577][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:08:05,579][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:08:06,296][__main__][INFO] - Iteration 390 took 51s (31.25% Gen, 67.37% Train). Generation: 16s, Training: 35s. Estimated remaining time: 37h 15m 56s. Estimated total time: 43h 18m 22s. Time estimates for 10 more iterations: 8m 39s, 100 more iterations: 1h 26m 36s, 500 more iterations: 7h 13m 3s. +[2026-03-26 03:08:06,298][__main__][INFO] - Starting iteration 390. +[2026-03-26 03:08:06,696][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 38 and human policies 1. +[2026-03-26 03:08:06,696][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:08:12,852][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 03:08:24,274][__main__][INFO] - Number of regex retries in iteration 390: 1 +[2026-03-26 03:08:24,275][__main__][INFO] - agents played in iteration 390 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:08:25,033][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:08:25,057][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:08:25,081][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:08:25,104][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:08:25,104][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:08:25,105][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:08:25,823][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:08:26,274][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:08:26,777][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:08:27,274][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:08:27,772][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:08:28,269][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:08:28,766][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:08:29,266][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:08:29,769][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:08:30,265][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:08:30,793][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:08:31,305][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:08:31,806][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:08:32,310][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:08:32,811][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:08:33,310][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:08:33,810][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:08:34,309][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:08:34,811][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:08:35,313][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:08:35,811][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:08:36,313][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:08:36,814][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:08:37,313][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:08:37,810][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:08:38,307][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:08:38,804][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:08:39,302][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:08:39,800][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:08:40,299][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:08:40,800][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:08:41,298][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:08:41,796][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:08:42,295][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:08:42,792][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:08:43,290][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:08:43,787][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:08:44,283][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:08:44,780][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:08:45,285][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:08:45,784][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:08:46,289][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:08:46,787][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:08:47,287][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:08:47,785][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:08:48,285][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:08:48,787][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:08:49,285][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:08:49,783][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:08:50,282][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:08:50,781][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:08:51,280][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:08:51,781][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:08:52,285][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:08:52,785][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:08:53,283][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:08:53,782][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:08:54,282][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:08:54,783][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:08:55,285][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:08:55,784][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:08:56,283][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:08:56,783][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:08:57,283][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:08:57,784][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10188 tokens. +[2026-03-26 03:08:58,541][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 60.60%, Block Peak % of device VRAM: 62.21%, ΔTime: 00:00:32 +[2026-03-26 03:08:59,299][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:08:59,302][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:08:59,303][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:09:00,693][__main__][INFO] - Iteration 391 took 53s (32.55% Gen, 64.87% Train). Generation: 17s, Training: 35s. Estimated remaining time: 38h 56m 32s. Estimated total time: 44h 59m 53s. Time estimates for 10 more iterations: 8m 59s, 100 more iterations: 1h 29m 59s, 500 more iterations: 7h 29m 58s. +[2026-03-26 03:09:00,696][__main__][INFO] - Starting iteration 391. +[2026-03-26 03:09:01,096][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 39 and human policies 1. +[2026-03-26 03:09:01,097][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:09:06,211][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 03:09:16,636][__main__][INFO] - Number of regex retries in iteration 391: 1 +[2026-03-26 03:09:16,637][__main__][INFO] - agents played in iteration 391 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:09:17,389][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:09:17,412][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:09:17,435][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:09:17,458][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:09:17,459][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:09:17,460][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:09:18,173][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:09:18,625][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:09:19,126][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:09:19,622][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:09:20,123][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:09:20,619][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:09:21,116][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:09:21,612][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:09:22,110][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:09:22,605][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:09:23,101][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:09:23,598][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:09:24,095][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:09:24,591][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:09:25,090][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:09:25,585][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:09:26,085][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:09:26,581][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:09:27,078][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:09:27,576][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:09:28,073][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:09:28,570][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:09:29,069][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:09:29,574][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:09:30,074][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:09:30,571][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:09:31,078][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:09:31,578][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:09:32,077][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:09:32,577][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:09:33,075][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:09:33,573][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:09:34,069][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:09:34,564][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:09:35,059][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:09:35,554][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:09:36,051][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:09:36,547][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:09:37,043][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:09:37,540][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:09:38,038][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:09:38,536][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:09:39,034][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:09:39,531][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:09:40,029][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:09:40,526][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:09:41,024][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:09:41,522][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:09:42,020][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:09:42,516][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:09:43,014][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:09:43,509][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:09:44,006][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:09:44,502][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:09:44,998][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:09:45,492][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:09:45,987][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:09:46,483][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:09:46,986][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:09:47,484][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:09:47,980][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:09:48,476][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:09:48,973][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:09:49,468][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:09:49,964][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10033 tokens. +[2026-03-26 03:09:50,703][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 60.50%, Block Peak % of device VRAM: 62.00%, ΔTime: 00:00:32 +[2026-03-26 03:09:51,434][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:09:51,437][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:09:51,438][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:09:52,185][__main__][INFO] - Iteration 392 took 51s (30.42% Gen, 68.12% Train). Generation: 15s, Training: 34s. Estimated remaining time: 36h 30m 16s. Estimated total time: 42h 34m 28s. Time estimates for 10 more iterations: 8m 30s, 100 more iterations: 1h 25m 8s, 500 more iterations: 7h 5m 44s. +[2026-03-26 03:09:52,187][__main__][INFO] - Starting iteration 392. +[2026-03-26 03:09:52,588][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 39 and human policies 1. +[2026-03-26 03:09:52,589][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:09:56,304][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 03:10:07,585][__main__][INFO] - Number of regex retries in iteration 392: 1 +[2026-03-26 03:10:07,586][__main__][INFO] - agents played in iteration 392 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:10:08,337][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:10:08,360][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:10:08,384][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:10:08,407][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:10:08,407][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:10:08,408][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:10:09,111][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:10:09,564][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:10:10,065][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:10:10,560][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:10:11,055][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:10:11,550][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:10:12,044][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:10:12,542][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:10:13,037][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:10:13,534][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:10:14,029][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:10:14,526][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:10:15,025][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:10:15,523][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:10:16,021][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:10:16,519][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:10:17,018][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:10:17,517][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:10:18,015][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:10:18,534][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:10:19,037][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:10:19,536][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:10:20,033][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:10:20,529][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:10:21,026][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:10:21,522][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:10:22,018][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:10:22,515][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:10:23,011][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:10:23,506][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:10:24,002][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:10:24,499][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:10:25,000][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:10:25,501][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:10:26,001][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:10:26,499][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:10:27,007][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:10:27,506][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:10:28,007][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:10:28,512][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:10:29,011][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:10:29,510][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:10:30,008][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:10:30,507][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:10:31,004][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:10:31,501][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:10:32,003][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:10:32,500][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:10:32,996][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:10:33,496][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:10:33,997][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:10:34,493][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:10:34,991][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:10:35,490][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:10:35,987][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:10:36,487][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:10:36,984][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:10:37,483][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:10:37,980][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:10:38,477][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:10:38,978][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:10:39,478][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:10:39,975][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:10:40,475][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:10:40,976][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10157 tokens. +[2026-03-26 03:10:41,711][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 60.66%, Block Peak % of device VRAM: 62.21%, ΔTime: 00:00:32 +[2026-03-26 03:10:42,440][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:10:42,442][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:10:42,444][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:10:43,159][__main__][INFO] - Iteration 393 took 50s (29.66% Gen, 68.93% Train). Generation: 14s, Training: 34s. Estimated remaining time: 36h 3m 33s. Estimated total time: 42h 8m 35s. Time estimates for 10 more iterations: 8m 25s, 100 more iterations: 1h 24m 17s, 500 more iterations: 7h 1m 25s. +[2026-03-26 03:10:43,162][__main__][INFO] - Starting iteration 393. +[2026-03-26 03:10:43,560][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 39 and human policies 1. +[2026-03-26 03:10:43,561][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:10:59,934][__main__][INFO] - Number of regex retries in iteration 393: 0 +[2026-03-26 03:10:59,935][__main__][INFO] - agents played in iteration 393 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:11:00,696][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:11:00,720][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:11:00,743][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:11:00,766][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:11:00,767][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:11:00,768][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:11:01,474][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:11:01,939][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:11:02,441][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:11:02,942][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:11:03,439][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:11:03,936][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:11:04,434][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:11:04,932][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:11:05,430][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:11:05,926][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:11:06,423][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:11:06,919][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:11:07,417][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:11:07,914][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:11:08,411][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:11:08,907][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:11:09,404][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:11:09,901][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:11:10,398][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:11:10,895][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:11:11,391][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:11:11,888][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:11:12,384][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:11:12,882][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:11:13,379][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:11:13,880][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:11:14,379][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:11:14,878][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:11:15,377][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:11:15,877][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:11:16,372][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:11:16,872][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:11:17,367][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:11:17,862][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:11:18,360][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:11:18,856][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:11:19,353][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:11:19,852][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:11:20,346][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:11:20,851][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:11:21,351][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:11:21,858][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:11:22,358][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:11:22,859][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:11:23,361][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:11:23,860][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:11:24,360][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:11:24,860][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:11:25,356][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:11:25,852][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:11:26,350][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:11:26,847][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:11:27,344][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:11:27,841][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:11:28,339][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:11:28,835][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:11:29,337][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:11:29,835][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:11:30,335][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:11:30,834][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:11:31,330][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:11:31,827][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:11:32,323][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:11:32,820][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:11:33,317][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10174 tokens. +[2026-03-26 03:11:34,069][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.09%, Current % of VRAM taken: 60.52%, Block Peak % of device VRAM: 62.18%, ΔTime: 00:00:32 +[2026-03-26 03:11:34,794][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:11:34,796][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:11:34,798][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:11:35,624][__main__][INFO] - Iteration 394 took 52s (31.45% Gen, 66.96% Train). Generation: 16s, Training: 34s. Estimated remaining time: 37h 17m 19s. Estimated total time: 43h 23m 14s. Time estimates for 10 more iterations: 8m 40s, 100 more iterations: 1h 26m 46s, 500 more iterations: 7h 13m 52s. +[2026-03-26 03:11:35,627][__main__][INFO] - Starting iteration 394. +[2026-03-26 03:11:36,027][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 39 and human policies 1. +[2026-03-26 03:11:36,028][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:11:51,899][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 03:11:53,542][__main__][INFO] - Number of regex retries in iteration 394: 1 +[2026-03-26 03:11:53,543][__main__][INFO] - agents played in iteration 394 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:11:54,302][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:11:54,326][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:11:54,349][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:11:54,373][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:11:54,373][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:11:54,374][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:11:55,078][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:11:55,532][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:11:56,037][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:11:56,535][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:11:57,036][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:11:57,532][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:11:58,029][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:11:58,529][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:11:59,025][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:11:59,525][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:12:00,022][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:12:00,518][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:12:01,018][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:12:01,514][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:12:02,017][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:12:02,522][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:12:03,021][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:12:03,523][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:12:04,024][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:12:04,523][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:12:05,024][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:12:05,531][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:12:06,029][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:12:06,535][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:12:07,037][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:12:07,537][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:12:08,039][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:12:08,540][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:12:09,038][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:12:09,537][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:12:10,037][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:12:10,537][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:12:11,038][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:12:11,541][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:12:12,040][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:12:12,537][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:12:13,037][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:12:13,533][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:12:14,030][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:12:14,531][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:12:15,032][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:12:15,528][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:12:16,026][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:12:16,523][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:12:17,021][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:12:17,518][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:12:18,016][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:12:18,514][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:12:19,012][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:12:19,507][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:12:20,003][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:12:20,501][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:12:20,998][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:12:21,497][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:12:21,996][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:12:22,493][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:12:22,993][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:12:23,488][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:12:23,988][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:12:24,513][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:12:25,011][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:12:25,510][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:12:26,011][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:12:26,509][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:12:27,008][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10283 tokens. +[2026-03-26 03:12:27,755][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 60.61%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 +[2026-03-26 03:12:28,497][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:12:28,499][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:12:28,501][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:12:29,226][__main__][INFO] - Iteration 395 took 53s (32.92% Gen, 65.71% Train). Generation: 17s, Training: 34s. Estimated remaining time: 38h 13m 9s. Estimated total time: 44h 19m 58s. Time estimates for 10 more iterations: 8m 51s, 100 more iterations: 1h 28m 39s, 500 more iterations: 7h 23m 19s. +[2026-03-26 03:12:29,228][__main__][INFO] - Starting iteration 395. +[2026-03-26 03:12:29,626][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 39 and human policies 1. +[2026-03-26 03:12:29,627][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:12:45,088][__main__][INFO] - Number of regex retries in iteration 395: 0 +[2026-03-26 03:12:45,089][__main__][INFO] - agents played in iteration 395 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:12:45,846][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:12:45,871][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:12:45,895][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:12:45,919][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:12:45,920][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:12:45,920][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:12:46,634][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:12:47,088][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:12:47,594][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:12:48,093][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:12:48,596][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:12:49,106][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:12:49,606][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:12:50,106][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:12:50,606][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:12:51,105][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:12:51,609][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:12:52,106][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:12:52,605][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:12:53,104][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:12:53,605][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:12:54,107][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:12:54,617][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:12:55,119][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:12:55,617][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:12:56,114][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:12:56,611][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:12:57,109][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:12:57,606][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:12:58,104][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:12:58,603][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:12:59,104][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:12:59,605][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:13:00,103][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:13:00,601][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:13:01,100][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:13:01,602][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:13:02,100][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:13:02,598][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:13:03,096][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:13:03,592][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:13:04,093][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:13:04,589][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:13:05,085][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:13:05,583][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:13:06,081][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:13:06,578][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:13:07,075][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:13:07,571][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:13:08,067][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:13:08,566][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:13:09,061][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:13:09,558][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:13:10,053][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:13:10,550][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:13:11,046][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:13:11,544][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:13:12,041][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:13:12,538][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:13:13,036][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:13:13,540][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:13:14,037][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:13:14,534][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:13:15,029][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:13:15,524][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:13:16,021][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:13:16,517][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:13:17,013][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:13:17,507][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:13:18,002][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:13:18,498][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10076 tokens. +[2026-03-26 03:13:19,248][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.04%, Current % of VRAM taken: 60.46%, Block Peak % of device VRAM: 62.16%, ΔTime: 00:00:32 +[2026-03-26 03:13:19,976][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:13:19,978][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:13:19,980][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:13:20,797][__main__][INFO] - Iteration 396 took 51s (30.22% Gen, 68.18% Train). Generation: 15s, Training: 34s. Estimated remaining time: 36h 30m 53s. Estimated total time: 42h 38m 34s. Time estimates for 10 more iterations: 8m 31s, 100 more iterations: 1h 25m 17s, 500 more iterations: 7h 6m 25s. +[2026-03-26 03:13:20,799][__main__][INFO] - Starting iteration 396. +[2026-03-26 03:13:21,200][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 39 and human policies 1. +[2026-03-26 03:13:21,201][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:13:38,150][__main__][INFO] - Number of regex retries in iteration 396: 0 +[2026-03-26 03:13:38,151][__main__][INFO] - agents played in iteration 396 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:13:38,912][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:13:38,935][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:13:38,959][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:13:38,981][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:13:38,982][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:13:38,983][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:13:39,711][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:13:40,171][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:13:40,676][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:13:41,179][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:13:41,685][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:13:42,188][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:13:42,686][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:13:43,187][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:13:43,686][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:13:44,184][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:13:44,681][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:13:45,181][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:13:45,679][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:13:46,185][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:13:46,684][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:13:47,182][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:13:47,683][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:13:48,181][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:13:48,679][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:13:49,177][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:13:49,680][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:13:50,187][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:13:50,686][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:13:51,187][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:13:51,688][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:13:52,184][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:13:52,681][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:13:53,182][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:13:53,684][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:13:54,192][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:13:54,689][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:13:55,186][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:13:55,687][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:13:56,184][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:13:56,682][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:13:57,180][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:13:57,678][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:13:58,177][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:13:58,674][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:13:59,171][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:13:59,673][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:14:00,174][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:14:00,670][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:14:01,170][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:14:01,667][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:14:02,166][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:14:02,667][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:14:03,176][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:14:03,674][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:14:04,177][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:14:04,677][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:14:05,180][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:14:05,677][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:14:06,180][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:14:06,679][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:14:07,178][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:14:07,679][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:14:08,175][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:14:08,675][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:14:09,177][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:14:09,683][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:14:10,183][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:14:10,682][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:14:11,180][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:14:11,680][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10203 tokens. +[2026-03-26 03:14:12,431][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 60.49%, Block Peak % of device VRAM: 62.18%, ΔTime: 00:00:32 +[2026-03-26 03:14:13,168][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:14:13,170][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:14:13,172][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:14:13,885][__main__][INFO] - Iteration 397 took 52s (32.17% Gen, 66.47% Train). Generation: 16s, Training: 35s. Estimated remaining time: 37h 45m 44s. Estimated total time: 43h 54m 18s. Time estimates for 10 more iterations: 8m 46s, 100 more iterations: 1h 27m 48s, 500 more iterations: 7h 19m 3s. +[2026-03-26 03:14:13,887][__main__][INFO] - Starting iteration 397. +[2026-03-26 03:14:14,288][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 39 and human policies 1. +[2026-03-26 03:14:14,289][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:14:30,475][__main__][INFO] - Number of regex retries in iteration 397: 0 +[2026-03-26 03:14:30,476][__main__][INFO] - agents played in iteration 397 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:14:31,238][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:14:31,262][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:14:31,286][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:14:31,310][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:14:31,311][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:14:31,311][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:14:32,026][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:14:32,485][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:14:32,991][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:14:33,494][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:14:33,996][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:14:34,499][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:14:35,001][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:14:35,498][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:14:35,998][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:14:36,498][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:14:36,998][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:14:37,496][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:14:37,997][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:14:38,502][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:14:39,002][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:14:39,502][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:14:40,010][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:14:40,513][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:14:41,018][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:14:41,521][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:14:42,022][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:14:42,521][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:14:43,021][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:14:43,520][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:14:44,021][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:14:44,520][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:14:45,020][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:14:45,518][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:14:46,018][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:14:46,517][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:14:47,017][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:14:47,516][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:14:48,015][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:14:48,517][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:14:49,014][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:14:49,511][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:14:50,008][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:14:50,509][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:14:51,009][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:14:51,506][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:14:52,004][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:14:52,502][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:14:53,000][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:14:53,499][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:14:53,999][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:14:54,498][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:14:54,999][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:14:55,497][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:14:55,996][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:14:56,494][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:14:56,995][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:14:57,491][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:14:57,986][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:14:58,483][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:14:58,980][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:14:59,477][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:14:59,978][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:15:00,476][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:15:00,971][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:15:01,466][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:15:01,963][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:15:02,460][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:15:02,958][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:15:03,455][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:15:03,953][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10207 tokens. +[2026-03-26 03:15:04,695][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 60.56%, Block Peak % of device VRAM: 62.27%, ΔTime: 00:00:32 +[2026-03-26 03:15:05,398][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:15:05,403][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:15:05,405][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:15:06,129][__main__][INFO] - Iteration 398 took 51s (31.22% Gen, 67.38% Train). Generation: 16s, Training: 34s. Estimated remaining time: 37h 2m 39s. Estimated total time: 43h 12m 5s. Time estimates for 10 more iterations: 8m 38s, 100 more iterations: 1h 26m 24s, 500 more iterations: 7h 12m 0s. +[2026-03-26 03:15:06,132][__main__][INFO] - Starting iteration 398. +[2026-03-26 03:15:06,533][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 39 and human policies 1. +[2026-03-26 03:15:06,534][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:15:10,792][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 03:15:22,281][__main__][INFO] - Number of regex retries in iteration 398: 1 +[2026-03-26 03:15:22,282][__main__][INFO] - agents played in iteration 398 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:15:23,042][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:15:23,067][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:15:23,091][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:15:23,115][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:15:23,116][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:15:23,116][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:15:23,839][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:15:24,296][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:15:24,799][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:15:25,298][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:15:25,801][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:15:26,300][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:15:26,801][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:15:27,302][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:15:27,800][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:15:28,299][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:15:28,799][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:15:29,298][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:15:29,797][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:15:30,298][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:15:30,796][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:15:31,294][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:15:31,793][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:15:32,292][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:15:32,791][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:15:33,287][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:15:33,784][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:15:34,282][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:15:34,785][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:15:35,284][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:15:35,788][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:15:36,287][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:15:36,789][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:15:37,287][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:15:37,786][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:15:38,287][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:15:38,789][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:15:39,289][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:15:39,787][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:15:40,287][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:15:40,789][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:15:41,290][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:15:41,787][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:15:42,287][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:15:42,786][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:15:43,285][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:15:43,790][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:15:44,288][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:15:44,786][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:15:45,284][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:15:45,782][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:15:46,281][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:15:46,780][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:15:47,281][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:15:47,779][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:15:48,277][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:15:48,774][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:15:49,270][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:15:49,770][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:15:50,267][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:15:50,764][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:15:51,262][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:15:51,759][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:15:52,259][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:15:52,756][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:15:53,252][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:15:53,747][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:15:54,246][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:15:54,744][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:15:55,241][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:15:55,738][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10220 tokens. +[2026-03-26 03:15:56,491][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 60.50%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 +[2026-03-26 03:15:57,239][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:15:57,241][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:15:57,245][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:15:57,965][__main__][INFO] - Iteration 399 took 51s (30.62% Gen, 67.98% Train). Generation: 15s, Training: 34s. Estimated remaining time: 36h 41m 20s. Estimated total time: 42h 51m 38s. Time estimates for 10 more iterations: 8m 34s, 100 more iterations: 1h 25m 43s, 500 more iterations: 7h 8m 36s. +[2026-03-26 03:15:57,967][__main__][INFO] - Starting iteration 399. +[2026-03-26 03:15:58,368][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 39 and human policies 1. +[2026-03-26 03:15:58,368][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:16:16,176][__main__][INFO] - Number of regex retries in iteration 399: 0 +[2026-03-26 03:16:16,177][__main__][INFO] - agents played in iteration 399 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:16:16,939][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:16:16,963][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:16:16,986][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:16:17,009][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:16:17,010][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:16:17,011][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:16:17,713][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:16:18,165][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:16:18,669][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:16:19,166][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:16:19,662][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:16:20,159][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:16:20,657][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:16:21,158][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:16:21,655][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:16:22,150][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:16:22,647][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:16:23,143][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:16:23,637][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:16:24,132][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:16:24,629][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:16:25,127][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:16:25,624][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:16:26,122][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:16:26,625][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:16:27,128][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:16:27,634][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:16:28,139][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:16:28,638][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:16:29,138][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:16:29,637][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:16:30,161][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:16:30,662][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:16:31,162][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:16:31,664][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:16:32,166][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:16:32,666][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:16:33,165][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:16:33,665][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:16:34,165][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:16:34,664][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:16:35,162][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:16:35,660][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:16:36,160][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:16:36,661][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:16:37,161][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:16:37,660][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:16:38,157][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:16:38,652][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:16:39,149][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:16:39,645][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:16:40,142][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:16:40,641][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:16:41,137][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:16:41,637][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:16:42,136][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:16:42,636][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:16:43,135][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:16:43,636][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:16:44,143][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:16:44,644][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:16:45,144][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:16:45,643][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:16:46,146][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:16:46,643][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:16:47,143][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:16:47,644][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:16:48,141][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:16:48,638][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:16:49,135][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:16:49,631][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10144 tokens. +[2026-03-26 03:16:50,363][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.04%, Current % of VRAM taken: 60.46%, Block Peak % of device VRAM: 62.20%, ΔTime: 00:00:32 +[2026-03-26 03:16:51,095][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:16:51,097][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:16:51,099][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:16:51,857][__main__][INFO] - Iteration 400 took 53s (33.29% Gen, 65.29% Train). Generation: 17s, Training: 34s. Estimated remaining time: 38h 23m 18s. Estimated total time: 44h 34m 29s. Time estimates for 10 more iterations: 8m 54s, 100 more iterations: 1h 29m 8s, 500 more iterations: 7h 25m 44s. +[2026-03-26 03:16:51,859][__main__][INFO] - Starting iteration 400. +[2026-03-26 03:16:52,258][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 39 and human policies 1. +[2026-03-26 03:16:52,259][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:17:08,664][__main__][INFO] - Number of regex retries in iteration 400: 0 +[2026-03-26 03:17:08,665][__main__][INFO] - agents played in iteration 400 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:17:09,424][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:17:09,448][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:17:09,471][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:17:09,495][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:17:09,495][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:17:09,496][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:17:10,209][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:17:10,662][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:17:11,165][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:17:11,663][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:17:12,162][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:17:12,660][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:17:13,159][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:17:13,658][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:17:14,157][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:17:14,656][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:17:15,155][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:17:15,654][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:17:16,153][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:17:16,653][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:17:17,156][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:17:17,655][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:17:18,157][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:17:18,657][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:17:19,157][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:17:19,660][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:17:20,156][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:17:20,656][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:17:21,153][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:17:21,656][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:17:22,154][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:17:22,654][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:17:23,149][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:17:23,644][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:17:24,140][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:17:24,638][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:17:25,133][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:17:25,629][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:17:26,125][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:17:26,621][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:17:27,122][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:17:27,619][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:17:28,116][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:17:28,613][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:17:29,109][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:17:29,605][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:17:30,102][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:17:30,600][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:17:31,100][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:17:31,600][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:17:32,101][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:17:32,599][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:17:33,097][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:17:33,599][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:17:34,098][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:17:34,598][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:17:35,097][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:17:35,605][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:17:36,104][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:17:36,606][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:17:37,117][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:17:37,622][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:17:38,139][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:17:38,641][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:17:39,141][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:17:39,645][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:17:40,143][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:17:40,642][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:17:41,140][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:17:41,641][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:17:42,139][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10182 tokens. +[2026-03-26 03:17:42,890][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.18%, ΔTime: 00:00:32 +[2026-03-26 03:17:43,626][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:17:43,628][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:17:43,630][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:17:45,072][__main__][INFO] - Iteration 401 took 52s (31.06% Gen, 66.20% Train). Generation: 16s, Training: 34s. Estimated remaining time: 37h 48m 40s. Estimated total time: 44h 0m 45s. Time estimates for 10 more iterations: 8m 48s, 100 more iterations: 1h 28m 1s, 500 more iterations: 7h 20m 7s. +[2026-03-26 03:17:45,075][__main__][INFO] - Starting iteration 401. +[2026-03-26 03:17:45,472][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 40 and human policies 1. +[2026-03-26 03:17:45,473][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:18:01,804][__main__][INFO] - Number of regex retries in iteration 401: 0 +[2026-03-26 03:18:01,805][__main__][INFO] - agents played in iteration 401 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:18:02,564][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:18:02,587][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:18:02,611][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:18:02,634][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:18:02,634][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:18:02,635][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:18:03,341][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:18:03,795][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:18:04,300][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:18:04,802][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:18:05,300][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:18:05,799][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:18:06,301][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:18:06,800][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:18:07,299][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:18:07,797][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:18:08,295][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:18:08,792][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:18:09,289][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:18:09,785][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:18:10,282][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:18:10,780][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:18:11,279][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:18:11,776][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:18:12,271][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:18:12,770][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:18:13,264][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:18:13,761][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:18:14,257][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:18:14,751][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:18:15,246][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:18:15,744][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:18:16,239][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:18:16,740][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:18:17,239][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:18:17,739][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:18:18,241][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:18:18,741][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:18:19,241][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:18:19,740][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:18:20,236][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:18:20,736][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:18:21,230][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:18:21,726][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:18:22,225][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:18:22,722][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:18:23,221][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:18:23,716][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:18:24,211][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:18:24,706][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:18:25,203][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:18:25,700][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:18:26,198][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:18:26,698][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:18:27,201][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:18:27,699][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:18:28,196][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:18:28,691][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:18:29,187][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:18:29,687][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:18:30,183][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:18:30,679][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:18:31,177][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:18:31,674][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:18:32,172][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:18:32,668][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:18:33,164][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:18:33,660][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:18:34,158][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:18:34,654][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:18:35,155][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10099 tokens. +[2026-03-26 03:18:35,903][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 +[2026-03-26 03:18:36,641][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:18:36,644][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:18:36,645][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:18:37,415][__main__][INFO] - Iteration 402 took 51s (31.44% Gen, 67.07% Train). Generation: 16s, Training: 34s. Estimated remaining time: 37h 4m 14s. Estimated total time: 43h 17m 12s. Time estimates for 10 more iterations: 8m 39s, 100 more iterations: 1h 26m 34s, 500 more iterations: 7h 12m 52s. +[2026-03-26 03:18:37,418][__main__][INFO] - Starting iteration 402. +[2026-03-26 03:18:37,818][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 40 and human policies 1. +[2026-03-26 03:18:37,819][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:18:47,174][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 03:18:53,420][__main__][INFO] - Number of regex retries in iteration 402: 1 +[2026-03-26 03:18:53,421][__main__][INFO] - agents played in iteration 402 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:18:54,173][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:18:54,197][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:18:54,220][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:18:54,243][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:18:54,244][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:18:54,244][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:18:54,945][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:18:55,401][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:18:55,905][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:18:56,406][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:18:56,904][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:18:57,403][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:18:57,904][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:18:58,404][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:18:58,903][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:18:59,398][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:18:59,894][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:19:00,388][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:19:00,882][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:19:01,379][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:19:01,877][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:19:02,372][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:19:02,871][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:19:03,366][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:19:03,863][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:19:04,360][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:19:04,857][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:19:05,354][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:19:05,851][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:19:06,350][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:19:06,846][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:19:07,343][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:19:07,839][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:19:08,337][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:19:08,833][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:19:09,329][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:19:09,826][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:19:10,325][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:19:10,821][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:19:11,316][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:19:11,810][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:19:12,307][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:19:12,802][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:19:13,298][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:19:13,794][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:19:14,289][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:19:14,784][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:19:15,279][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:19:15,773][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:19:16,267][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:19:16,761][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:19:17,255][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:19:17,751][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:19:18,244][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:19:18,741][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:19:19,236][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:19:19,732][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:19:20,227][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:19:20,722][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:19:21,218][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:19:21,714][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:19:22,210][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:19:22,706][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:19:23,201][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:19:23,697][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:19:24,192][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:19:24,691][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:19:25,190][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:19:25,688][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:19:26,182][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:19:26,678][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10050 tokens. +[2026-03-26 03:19:27,415][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 60.53%, Block Peak % of device VRAM: 62.21%, ΔTime: 00:00:32 +[2026-03-26 03:19:28,186][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:19:28,188][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:19:28,190][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:19:28,841][__main__][INFO] - Iteration 403 took 51s (30.58% Gen, 68.14% Train). Generation: 15s, Training: 34s. Estimated remaining time: 36h 17m 22s. Estimated total time: 42h 31m 10s. Time estimates for 10 more iterations: 8m 30s, 100 more iterations: 1h 25m 2s, 500 more iterations: 7h 5m 11s. +[2026-03-26 03:19:28,843][__main__][INFO] - Starting iteration 403. +[2026-03-26 03:19:29,241][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 40 and human policies 1. +[2026-03-26 03:19:29,242][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:19:44,823][__main__][INFO] - Number of regex retries in iteration 403: 0 +[2026-03-26 03:19:44,824][__main__][INFO] - agents played in iteration 403 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:19:45,573][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:19:45,597][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:19:45,620][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:19:45,644][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:19:45,644][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:19:45,645][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:19:46,348][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:19:46,800][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:19:47,302][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:19:47,800][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:19:48,299][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:19:48,796][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:19:49,310][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:19:49,810][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:19:50,308][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:19:50,810][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:19:51,312][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:19:51,816][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:19:52,328][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:19:52,828][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:19:53,329][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:19:53,829][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:19:54,326][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:19:54,826][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:19:55,323][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:19:55,821][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:19:56,318][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:19:56,814][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:19:57,311][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:19:57,810][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:19:58,306][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:19:58,806][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:19:59,306][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:19:59,802][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:20:00,299][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:20:00,795][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:20:01,291][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:20:01,788][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:20:02,287][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:20:02,784][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:20:03,280][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:20:03,776][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:20:04,271][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:20:04,768][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:20:05,265][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:20:05,763][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:20:06,259][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:20:06,755][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:20:07,250][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:20:07,745][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:20:08,240][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:20:08,738][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:20:09,237][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:20:09,733][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:20:10,230][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:20:10,727][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:20:11,227][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:20:11,722][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:20:12,220][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:20:12,734][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:20:13,233][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:20:13,732][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:20:14,234][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:20:14,731][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:20:15,229][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:20:15,726][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:20:16,227][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:20:16,730][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:20:17,233][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:20:17,734][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:20:18,235][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10122 tokens. +[2026-03-26 03:20:18,987][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 60.61%, Block Peak % of device VRAM: 62.14%, ΔTime: 00:00:32 +[2026-03-26 03:20:19,717][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:20:19,719][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:20:19,721][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:20:20,445][__main__][INFO] - Iteration 404 took 51s (30.43% Gen, 68.15% Train). Generation: 15s, Training: 34s. Estimated remaining time: 36h 25m 34s. Estimated total time: 42h 40m 15s. Time estimates for 10 more iterations: 8m 32s, 100 more iterations: 1h 25m 20s, 500 more iterations: 7h 6m 42s. +[2026-03-26 03:20:20,448][__main__][INFO] - Starting iteration 404. +[2026-03-26 03:20:20,848][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 40 and human policies 1. +[2026-03-26 03:20:20,849][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:20:38,204][__main__][INFO] - Number of regex retries in iteration 404: 0 +[2026-03-26 03:20:38,205][__main__][INFO] - agents played in iteration 404 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:20:38,960][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:20:38,985][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:20:39,009][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:20:39,033][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:20:39,034][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:20:39,035][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:20:39,745][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:20:40,199][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:20:40,709][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:20:41,217][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:20:41,719][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:20:42,217][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:20:42,716][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:20:43,213][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:20:43,715][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:20:44,219][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:20:44,734][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:20:45,235][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:20:45,734][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:20:46,263][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:20:46,763][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:20:47,266][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:20:47,763][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:20:48,264][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:20:48,768][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:20:49,265][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:20:49,763][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:20:50,260][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:20:50,757][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:20:51,254][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:20:51,749][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:20:52,251][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:20:52,750][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:20:53,252][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:20:53,761][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:20:54,260][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:20:54,761][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:20:55,259][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:20:55,755][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:20:56,253][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:20:56,750][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:20:57,247][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:20:57,744][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:20:58,241][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:20:58,742][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:20:59,239][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:20:59,742][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:21:00,250][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:21:00,749][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:21:01,249][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:21:01,751][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:21:02,251][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:21:02,750][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:21:03,250][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:21:03,748][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:21:04,247][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:21:04,746][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:21:05,244][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:21:05,745][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:21:06,246][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:21:06,749][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:21:07,246][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:21:07,744][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:21:08,242][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:21:08,740][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:21:09,238][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:21:09,736][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:21:10,234][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:21:10,731][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:21:11,229][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:21:11,726][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10107 tokens. +[2026-03-26 03:21:12,467][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 60.54%, Block Peak % of device VRAM: 62.10%, ΔTime: 00:00:32 +[2026-03-26 03:21:13,204][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:21:13,207][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:21:13,208][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:21:14,692][__main__][INFO] - Iteration 405 took 53s (32.23% Gen, 65.01% Train). Generation: 17s, Training: 35s. Estimated remaining time: 38h 36m 39s. Estimated total time: 44h 52m 13s. Time estimates for 10 more iterations: 8m 58s, 100 more iterations: 1h 29m 44s, 500 more iterations: 7h 28m 42s. +[2026-03-26 03:21:14,694][__main__][INFO] - Starting iteration 405. +[2026-03-26 03:21:15,091][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 40 and human policies 1. +[2026-03-26 03:21:15,092][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:21:30,202][__main__][INFO] - Number of regex retries in iteration 405: 0 +[2026-03-26 03:21:30,203][__main__][INFO] - agents played in iteration 405 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:21:30,954][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:21:30,977][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:21:31,001][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:21:31,025][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:21:31,025][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:21:31,026][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:21:31,743][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:21:32,203][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:21:32,706][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:21:33,206][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:21:33,705][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:21:34,204][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:21:34,702][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:21:35,203][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:21:35,707][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:21:36,210][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:21:36,710][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:21:37,215][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:21:37,713][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:21:38,211][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:21:38,710][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:21:39,214][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:21:39,718][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:21:40,215][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:21:40,715][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:21:41,216][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:21:41,714][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:21:42,214][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:21:42,712][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:21:43,210][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:21:43,707][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:21:44,209][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:21:44,714][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:21:45,213][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:21:45,714][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:21:46,213][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:21:46,712][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:21:47,210][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:21:47,708][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:21:48,206][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:21:48,706][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:21:49,205][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:21:49,703][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:21:50,207][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:21:50,714][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:21:51,213][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:21:51,715][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:21:52,212][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:21:52,710][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:21:53,207][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:21:53,705][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:21:54,202][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:21:54,701][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:21:55,198][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:21:55,695][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:21:56,195][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:21:56,691][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:21:57,190][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:21:57,687][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:21:58,185][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:21:58,683][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:21:59,181][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:21:59,679][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:22:00,175][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:22:00,672][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:22:01,170][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:22:01,666][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:22:02,164][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:22:02,661][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:22:03,156][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:22:03,650][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10033 tokens. +[2026-03-26 03:22:04,408][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.00%, Current % of VRAM taken: 60.43%, Block Peak % of device VRAM: 62.07%, ΔTime: 00:00:32 +[2026-03-26 03:22:05,142][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:22:05,144][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:22:05,146][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:22:05,863][__main__][INFO] - Iteration 406 took 50s (29.76% Gen, 68.82% Train). Generation: 15s, Training: 34s. Estimated remaining time: 36h 2m 12s. Estimated total time: 42h 18m 37s. Time estimates for 10 more iterations: 8m 27s, 100 more iterations: 1h 24m 37s, 500 more iterations: 7h 3m 6s. +[2026-03-26 03:22:05,872][__main__][INFO] - Starting iteration 406. +[2026-03-26 03:22:06,272][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 40 and human policies 1. +[2026-03-26 03:22:06,273][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:22:22,024][__main__][INFO] - Number of regex retries in iteration 406: 0 +[2026-03-26 03:22:22,025][__main__][INFO] - agents played in iteration 406 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:22:22,777][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.45%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:22:22,800][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.45%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:22:22,824][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.45%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:22:22,848][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.45%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:22:22,848][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:22:22,849][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:22:23,553][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:22:24,005][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:22:24,523][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:22:25,025][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:22:25,525][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:22:26,024][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:22:26,523][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:22:27,022][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:22:27,521][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:22:28,018][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:22:28,515][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:22:29,013][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:22:29,511][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:22:30,008][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:22:30,506][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:22:31,017][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:22:31,517][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:22:32,024][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:22:32,522][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:22:33,024][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:22:33,522][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:22:34,022][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:22:34,525][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:22:35,032][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:22:35,535][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:22:36,034][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:22:36,536][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:22:37,039][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:22:37,537][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:22:38,036][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:22:38,539][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:22:39,040][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:22:39,538][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:22:40,044][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:22:40,547][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:22:41,045][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:22:41,545][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:22:42,047][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:22:42,546][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:22:43,048][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:22:43,545][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:22:44,043][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:22:44,541][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:22:45,042][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:22:45,543][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:22:46,046][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:22:46,554][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:22:47,054][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:22:47,556][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:22:48,053][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:22:48,549][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:22:49,047][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:22:49,545][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:22:50,042][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:22:50,539][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:22:51,036][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:22:51,532][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:22:52,032][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:22:52,532][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:22:53,032][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:22:53,538][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:22:54,040][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:22:54,541][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:22:55,040][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:22:55,539][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10124 tokens. +[2026-03-26 03:22:56,291][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.27%, ΔTime: 00:00:32 +[2026-03-26 03:22:57,076][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:22:57,079][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:22:57,081][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:22:57,845][__main__][INFO] - Iteration 407 took 51s (30.54% Gen, 67.97% Train). Generation: 15s, Training: 35s. Estimated remaining time: 36h 41m 22s. Estimated total time: 42h 58m 40s. Time estimates for 10 more iterations: 8m 35s, 100 more iterations: 1h 25m 57s, 500 more iterations: 7h 9m 46s. +[2026-03-26 03:22:57,847][__main__][INFO] - Starting iteration 407. +[2026-03-26 03:22:58,248][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 40 and human policies 1. +[2026-03-26 03:22:58,249][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:23:14,285][__main__][INFO] - Number of regex retries in iteration 407: 0 +[2026-03-26 03:23:14,286][__main__][INFO] - agents played in iteration 407 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:23:15,043][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:23:15,067][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:23:15,090][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:23:15,113][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:23:15,114][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:23:15,114][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:23:15,829][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:23:16,286][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:23:16,790][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:23:17,288][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:23:17,788][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:23:18,290][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:23:18,789][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:23:19,304][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:23:19,807][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:23:20,308][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:23:20,806][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:23:21,306][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:23:21,807][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:23:22,307][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:23:22,808][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:23:23,310][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:23:23,810][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:23:24,308][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:23:24,809][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:23:25,307][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:23:25,806][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:23:26,305][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:23:26,803][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:23:27,312][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:23:27,810][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:23:28,309][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:23:28,807][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:23:29,306][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:23:29,807][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:23:30,305][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:23:30,803][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:23:31,302][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:23:31,807][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:23:32,316][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:23:32,816][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:23:33,313][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:23:33,811][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:23:34,310][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:23:34,807][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:23:35,306][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:23:35,805][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:23:36,303][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:23:36,799][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:23:37,297][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:23:37,794][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:23:38,290][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:23:38,792][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:23:39,298][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:23:39,799][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:23:40,301][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:23:40,800][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:23:41,298][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:23:41,797][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:23:42,295][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:23:42,795][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:23:43,293][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:23:43,795][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:23:44,295][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:23:44,792][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:23:45,290][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:23:45,791][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:23:46,294][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:23:46,793][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:23:47,292][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:23:47,789][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10128 tokens. +[2026-03-26 03:23:48,542][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.05%, Current % of VRAM taken: 60.48%, Block Peak % of device VRAM: 62.34%, ΔTime: 00:00:32 +[2026-03-26 03:23:49,315][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:23:49,317][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:23:49,319][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:23:50,166][__main__][INFO] - Iteration 408 took 51s (30.89% Gen, 67.48% Train). Generation: 16s, Training: 35s. Estimated remaining time: 36h 57m 45s. Estimated total time: 43h 15m 55s. Time estimates for 10 more iterations: 8m 39s, 100 more iterations: 1h 26m 31s, 500 more iterations: 7h 12m 39s. +[2026-03-26 03:23:50,170][__main__][INFO] - Starting iteration 408. +[2026-03-26 03:23:50,570][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 40 and human policies 1. +[2026-03-26 03:23:50,570][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:24:05,399][__main__][INFO] - Number of regex retries in iteration 408: 0 +[2026-03-26 03:24:05,400][__main__][INFO] - agents played in iteration 408 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:24:06,155][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:24:06,179][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:24:06,203][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:24:06,226][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:24:06,227][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:24:06,228][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:24:06,953][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:24:07,405][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:24:07,913][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:24:08,409][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:24:08,906][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:24:09,403][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:24:09,901][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:24:10,404][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:24:10,909][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:24:11,409][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:24:11,907][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:24:12,405][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:24:12,904][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:24:13,402][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:24:13,899][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:24:14,403][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:24:14,911][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:24:15,414][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:24:15,914][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:24:16,413][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:24:16,913][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:24:17,413][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:24:17,911][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:24:18,411][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:24:18,909][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:24:19,412][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:24:19,920][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:24:20,419][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:24:20,924][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:24:21,424][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:24:21,922][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:24:22,421][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:24:22,919][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:24:23,416][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:24:23,913][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:24:24,410][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:24:24,908][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:24:25,404][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:24:25,903][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:24:26,407][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:24:26,906][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:24:27,408][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:24:27,906][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:24:28,407][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:24:28,904][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:24:29,402][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:24:29,899][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:24:30,397][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:24:30,893][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:24:31,390][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:24:31,889][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:24:32,387][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:24:32,884][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:24:33,382][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:24:33,881][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:24:34,379][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:24:34,883][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:24:35,383][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:24:35,890][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:24:36,389][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:24:36,895][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:24:37,394][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:24:37,894][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:24:38,393][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:24:38,891][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10000 tokens. +[2026-03-26 03:24:39,644][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.15%, ΔTime: 00:00:32 +[2026-03-26 03:24:40,396][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:24:40,398][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:24:40,400][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:24:41,260][__main__][INFO] - Iteration 409 took 50s (29.26% Gen, 69.05% Train). Generation: 14s, Training: 34s. Estimated remaining time: 35h 55m 30s. Estimated total time: 42h 14m 31s. Time estimates for 10 more iterations: 8m 26s, 100 more iterations: 1h 24m 29s, 500 more iterations: 7h 2m 25s. +[2026-03-26 03:24:41,262][__main__][INFO] - Starting iteration 409. +[2026-03-26 03:24:41,662][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 40 and human policies 1. +[2026-03-26 03:24:41,663][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:24:58,900][__main__][INFO] - Number of regex retries in iteration 409: 0 +[2026-03-26 03:24:58,901][__main__][INFO] - agents played in iteration 409 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:24:59,665][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:24:59,690][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:24:59,715][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:24:59,740][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:24:59,740][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:24:59,741][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:25:00,462][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:25:00,924][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:25:01,430][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:25:01,932][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:25:02,435][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:25:02,936][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:25:03,451][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:25:03,953][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:25:04,454][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:25:04,955][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:25:05,457][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:25:05,955][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:25:06,454][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:25:06,958][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:25:07,463][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:25:07,968][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:25:08,467][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:25:08,965][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:25:09,463][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:25:09,961][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:25:10,459][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:25:10,965][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:25:11,472][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:25:11,969][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:25:12,469][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:25:12,970][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:25:13,469][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:25:13,967][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:25:14,465][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:25:14,963][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:25:15,461][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:25:15,961][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:25:16,463][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:25:16,969][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:25:17,469][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:25:17,968][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:25:18,468][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:25:18,970][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:25:19,470][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:25:19,968][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:25:20,468][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:25:20,965][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:25:21,465][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:25:21,962][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:25:22,458][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:25:22,958][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:25:23,467][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:25:23,966][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:25:24,464][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:25:24,962][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:25:25,460][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:25:25,958][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:25:26,461][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:25:26,959][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:25:27,458][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:25:27,958][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:25:28,459][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:25:28,955][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:25:29,456][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:25:29,952][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:25:30,449][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:25:30,947][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:25:31,444][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:25:31,944][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:25:32,442][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10100 tokens. +[2026-03-26 03:25:33,187][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 60.54%, Block Peak % of device VRAM: 62.17%, ΔTime: 00:00:32 +[2026-03-26 03:25:33,912][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:25:33,914][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:25:33,916][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:25:34,783][__main__][INFO] - Iteration 410 took 53s (32.45% Gen, 65.91% Train). Generation: 17s, Training: 35s. Estimated remaining time: 37h 56m 10s. Estimated total time: 44h 16m 4s. Time estimates for 10 more iterations: 8m 51s, 100 more iterations: 1h 28m 32s, 500 more iterations: 7h 22m 40s. +[2026-03-26 03:25:34,786][__main__][INFO] - Starting iteration 410. +[2026-03-26 03:25:35,185][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 40 and human policies 1. +[2026-03-26 03:25:35,185][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:25:51,279][__main__][INFO] - Number of regex retries in iteration 410: 0 +[2026-03-26 03:25:51,280][__main__][INFO] - agents played in iteration 410 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:25:52,030][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:25:52,053][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:25:52,077][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:25:52,101][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:25:52,101][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:25:52,102][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:25:52,818][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:25:53,271][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:25:53,772][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:25:54,271][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:25:54,768][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:25:55,264][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:25:55,765][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:25:56,265][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:25:56,761][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:25:57,260][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:25:57,757][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:25:58,252][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:25:58,748][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:25:59,244][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:25:59,739][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:26:00,235][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:26:00,731][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:26:01,227][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:26:01,723][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:26:02,219][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:26:02,713][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:26:03,209][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:26:03,705][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:26:04,207][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:26:04,706][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:26:05,214][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:26:05,717][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:26:06,217][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:26:06,717][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:26:07,217][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:26:07,718][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:26:08,217][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:26:08,720][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:26:09,216][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:26:09,711][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:26:10,206][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:26:10,704][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:26:11,200][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:26:11,696][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:26:12,193][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:26:12,689][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:26:13,186][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:26:13,683][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:26:14,181][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:26:14,678][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:26:15,173][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:26:15,667][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:26:16,162][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:26:16,660][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:26:17,158][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:26:17,678][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:26:18,179][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:26:18,678][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:26:19,180][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:26:19,679][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:26:20,177][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:26:20,679][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:26:21,175][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:26:21,671][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:26:22,168][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:26:22,663][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:26:23,160][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:26:23,656][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:26:24,152][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:26:24,649][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10000 tokens. +[2026-03-26 03:26:25,405][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 60.49%, Block Peak % of device VRAM: 62.16%, ΔTime: 00:00:32 +[2026-03-26 03:26:26,158][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:26:26,160][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:26:26,162][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:26:27,497][__main__][INFO] - Iteration 411 took 52s (30.77% Gen, 66.68% Train). Generation: 16s, Training: 34s. Estimated remaining time: 37h 14m 53s. Estimated total time: 43h 35m 40s. Time estimates for 10 more iterations: 8m 43s, 100 more iterations: 1h 27m 11s, 500 more iterations: 7h 15m 56s. +[2026-03-26 03:26:27,500][__main__][INFO] - Starting iteration 411. +[2026-03-26 03:26:27,899][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 41 and human policies 1. +[2026-03-26 03:26:27,899][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:26:43,611][__main__][INFO] - Number of regex retries in iteration 411: 0 +[2026-03-26 03:26:43,611][__main__][INFO] - agents played in iteration 411 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:26:44,365][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:26:44,389][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:26:44,412][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:26:44,435][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:26:44,436][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:26:44,436][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:26:45,141][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:26:45,595][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:26:46,097][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:26:46,595][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:26:47,092][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:26:47,589][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:26:48,086][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:26:48,585][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:26:49,083][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:26:49,579][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:26:50,074][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:26:50,570][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:26:51,067][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:26:51,563][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:26:52,058][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:26:52,553][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:26:53,054][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:26:53,552][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:26:54,068][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:26:54,566][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:26:55,064][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:26:55,563][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:26:56,067][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:26:56,567][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:26:57,064][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:26:57,562][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:26:58,061][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:26:58,584][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:26:59,082][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:26:59,581][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:27:00,080][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:27:00,583][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:27:01,080][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:27:01,579][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:27:02,078][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:27:02,576][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:27:03,072][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:27:03,571][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:27:04,068][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:27:04,564][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:27:05,059][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:27:05,559][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:27:06,061][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:27:06,562][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:27:07,060][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:27:07,563][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:27:08,060][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:27:08,557][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:27:09,053][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:27:09,550][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:27:10,048][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:27:10,546][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:27:11,043][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:27:11,540][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:27:12,037][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:27:12,533][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:27:13,030][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:27:13,528][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:27:14,025][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:27:14,523][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:27:15,021][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:27:15,518][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:27:16,016][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:27:16,514][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:27:17,015][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10011 tokens. +[2026-03-26 03:27:17,759][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 60.64%, Block Peak % of device VRAM: 62.16%, ΔTime: 00:00:32 +[2026-03-26 03:27:18,494][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:27:18,497][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:27:18,498][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:27:19,216][__main__][INFO] - Iteration 412 took 51s (30.62% Gen, 67.98% Train). Generation: 15s, Training: 34s. Estimated remaining time: 36h 24m 13s. Estimated total time: 42h 45m 52s. Time estimates for 10 more iterations: 8m 33s, 100 more iterations: 1h 25m 31s, 500 more iterations: 7h 7m 38s. +[2026-03-26 03:27:19,218][__main__][INFO] - Starting iteration 412. +[2026-03-26 03:27:19,617][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 41 and human policies 1. +[2026-03-26 03:27:19,618][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:27:24,212][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 03:27:35,776][__main__][INFO] - Number of regex retries in iteration 412: 1 +[2026-03-26 03:27:35,777][__main__][INFO] - agents played in iteration 412 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:27:36,528][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:27:36,552][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:27:36,575][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:27:36,598][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:27:36,599][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:27:36,600][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:27:37,316][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:27:37,772][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:27:38,272][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:27:38,769][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:27:39,266][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:27:39,767][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:27:40,264][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:27:40,760][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:27:41,256][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:27:41,751][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:27:42,245][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:27:42,741][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:27:43,236][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:27:43,731][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:27:44,227][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:27:44,721][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:27:45,219][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:27:45,716][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:27:46,218][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:27:46,717][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:27:47,216][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:27:47,716][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:27:48,214][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:27:48,712][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:27:49,211][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:27:49,708][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:27:50,205][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:27:50,701][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:27:51,203][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:27:51,701][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:27:52,198][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:27:52,695][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:27:53,192][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:27:53,692][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:27:54,191][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:27:54,697][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:27:55,203][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:27:55,714][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:27:56,216][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:27:56,718][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:27:57,220][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:27:57,718][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:27:58,217][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:27:58,714][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:27:59,214][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:27:59,713][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:28:00,212][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:28:00,710][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:28:01,208][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:28:01,708][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:28:02,209][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:28:02,705][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:28:03,201][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:28:03,698][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:28:04,196][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:28:04,691][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:28:05,187][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:28:05,687][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:28:06,184][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:28:06,682][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:28:07,181][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:28:07,678][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:28:08,174][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:28:08,669][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:28:09,165][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10139 tokens. +[2026-03-26 03:28:09,927][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.09%, Current % of VRAM taken: 60.52%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:32 +[2026-03-26 03:28:10,699][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:28:10,702][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:28:10,704][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:28:11,405][__main__][INFO] - Iteration 413 took 51s (31.20% Gen, 67.44% Train). Generation: 16s, Training: 34s. Estimated remaining time: 36h 46m 56s. Estimated total time: 43h 9m 27s. Time estimates for 10 more iterations: 8m 37s, 100 more iterations: 1h 26m 18s, 500 more iterations: 7h 11m 34s. +[2026-03-26 03:28:11,407][__main__][INFO] - Starting iteration 413. +[2026-03-26 03:28:11,807][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 41 and human policies 1. +[2026-03-26 03:28:11,808][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:28:26,882][__main__][INFO] - Number of regex retries in iteration 413: 0 +[2026-03-26 03:28:26,883][__main__][INFO] - agents played in iteration 413 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:28:27,634][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:28:27,657][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:28:27,681][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:28:27,705][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:28:27,705][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:28:27,706][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:28:28,429][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:28:28,881][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:28:29,382][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:28:29,880][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:28:30,377][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:28:30,874][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:28:31,373][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:28:31,873][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:28:32,370][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:28:32,866][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:28:33,364][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:28:33,861][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:28:34,363][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:28:34,859][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:28:35,357][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:28:35,853][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:28:36,353][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:28:36,851][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:28:37,347][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:28:37,844][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:28:38,339][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:28:38,836][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:28:39,332][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:28:39,828][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:28:40,325][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:28:40,822][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:28:41,320][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:28:41,819][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:28:42,321][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:28:42,820][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:28:43,318][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:28:43,817][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:28:44,315][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:28:44,813][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:28:45,311][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:28:45,809][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:28:46,307][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:28:46,804][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:28:47,303][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:28:47,803][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:28:48,302][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:28:48,802][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:28:49,301][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:28:49,799][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:28:50,299][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:28:50,796][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:28:51,293][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:28:51,790][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:28:52,287][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:28:52,784][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:28:53,280][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:28:53,776][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:28:54,274][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:28:54,779][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:28:55,277][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:28:55,777][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:28:56,276][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:28:56,777][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:28:57,275][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:28:57,775][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:28:58,273][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:28:58,772][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:28:59,273][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:28:59,772][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:29:00,271][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10057 tokens. +[2026-03-26 03:29:01,023][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.16%, ΔTime: 00:00:32 +[2026-03-26 03:29:01,776][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:29:01,779][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:29:01,780][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:29:02,502][__main__][INFO] - Iteration 414 took 50s (29.74% Gen, 68.84% Train). Generation: 15s, Training: 34s. Estimated remaining time: 35h 51m 23s. Estimated total time: 42h 14m 45s. Time estimates for 10 more iterations: 8m 26s, 100 more iterations: 1h 24m 29s, 500 more iterations: 7h 2m 27s. +[2026-03-26 03:29:02,504][__main__][INFO] - Starting iteration 414. +[2026-03-26 03:29:02,905][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 41 and human policies 1. +[2026-03-26 03:29:02,905][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:29:15,513][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 03:29:18,722][__main__][INFO] - Number of regex retries in iteration 414: 1 +[2026-03-26 03:29:18,722][__main__][INFO] - agents played in iteration 414 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:29:19,483][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:29:19,506][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:29:19,530][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:29:19,553][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:29:19,553][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:29:19,554][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:29:20,260][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:29:20,714][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:29:21,215][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:29:21,714][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:29:22,210][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:29:22,707][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:29:23,205][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:29:23,704][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:29:24,203][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:29:24,705][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:29:25,206][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:29:25,704][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:29:26,204][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:29:26,704][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:29:27,202][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:29:27,709][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:29:28,210][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:29:28,710][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:29:29,207][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:29:29,705][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:29:30,203][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:29:30,702][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:29:31,202][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:29:31,701][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:29:32,205][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:29:32,703][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:29:33,202][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:29:33,700][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:29:34,198][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:29:34,696][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:29:35,194][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:29:35,691][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:29:36,188][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:29:36,683][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:29:37,179][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:29:37,675][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:29:38,170][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:29:38,667][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:29:39,163][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:29:39,660][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:29:40,157][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:29:40,654][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:29:41,151][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:29:41,648][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:29:42,145][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:29:42,645][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:29:43,142][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:29:43,639][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:29:44,140][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:29:44,637][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:29:45,134][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:29:45,630][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:29:46,127][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:29:46,624][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:29:47,121][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:29:47,617][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:29:48,113][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:29:48,611][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:29:49,111][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:29:49,610][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:29:50,112][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:29:50,609][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:29:51,107][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:29:51,605][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:29:52,104][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10093 tokens. +[2026-03-26 03:29:52,845][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.22%, ΔTime: 00:00:32 +[2026-03-26 03:29:53,565][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:29:53,567][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:29:53,569][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:29:54,230][__main__][INFO] - Iteration 415 took 51s (30.82% Gen, 67.89% Train). Generation: 15s, Training: 34s. Estimated remaining time: 36h 22m 2s. Estimated total time: 42h 46m 16s. Time estimates for 10 more iterations: 8m 33s, 100 more iterations: 1h 25m 32s, 500 more iterations: 7h 7m 42s. +[2026-03-26 03:29:54,233][__main__][INFO] - Starting iteration 415. +[2026-03-26 03:29:54,634][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 41 and human policies 1. +[2026-03-26 03:29:54,634][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:30:10,150][__main__][INFO] - Number of regex retries in iteration 415: 0 +[2026-03-26 03:30:10,151][__main__][INFO] - agents played in iteration 415 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:30:10,903][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:30:10,927][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:30:10,951][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:30:10,974][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:30:10,975][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:30:10,975][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:30:11,697][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:30:12,162][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:30:12,673][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:30:13,175][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:30:13,677][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:30:14,179][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:30:14,685][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:30:15,192][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:30:15,694][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:30:16,198][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:30:16,697][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:30:17,195][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:30:17,694][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:30:18,197][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:30:18,703][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:30:19,204][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:30:19,705][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:30:20,204][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:30:20,703][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:30:21,202][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:30:21,701][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:30:22,201][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:30:22,701][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:30:23,201][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:30:23,702][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:30:24,200][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:30:24,699][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:30:25,198][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:30:25,702][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:30:26,201][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:30:26,715][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:30:27,216][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:30:27,719][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:30:28,218][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:30:28,716][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:30:29,220][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:30:29,718][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:30:30,216][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:30:30,713][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:30:31,214][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:30:31,711][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:30:32,218][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:30:32,735][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:30:33,236][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:30:33,736][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:30:34,236][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:30:34,740][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:30:35,242][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:30:35,743][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:30:36,240][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:30:36,742][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:30:37,242][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:30:37,738][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:30:38,236][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:30:38,734][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:30:39,230][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:30:39,731][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:30:40,228][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:30:40,725][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:30:41,225][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:30:41,731][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:30:42,239][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:30:42,737][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:30:43,235][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:30:43,732][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10146 tokens. +[2026-03-26 03:30:44,521][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 60.49%, Block Peak % of device VRAM: 62.16%, ΔTime: 00:00:32 +[2026-03-26 03:30:45,261][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:30:45,264][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:30:45,265][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:30:45,992][__main__][INFO] - Iteration 416 took 51s (30.21% Gen, 68.37% Train). Generation: 15s, Training: 35s. Estimated remaining time: 36h 22m 50s. Estimated total time: 42h 47m 56s. Time estimates for 10 more iterations: 8m 33s, 100 more iterations: 1h 25m 35s, 500 more iterations: 7h 7m 59s. +[2026-03-26 03:30:45,994][__main__][INFO] - Starting iteration 416. +[2026-03-26 03:30:46,394][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 41 and human policies 1. +[2026-03-26 03:30:46,395][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:31:03,068][__main__][INFO] - Number of regex retries in iteration 416: 0 +[2026-03-26 03:31:03,069][__main__][INFO] - agents played in iteration 416 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:31:03,824][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:31:03,849][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:31:03,873][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:31:03,898][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:31:03,899][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:31:03,899][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:31:04,626][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:31:05,080][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:31:05,583][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:31:06,082][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:31:06,585][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:31:07,086][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:31:07,584][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:31:08,081][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:31:08,580][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:31:09,091][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:31:09,590][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:31:10,090][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:31:10,588][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:31:11,087][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:31:11,584][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:31:12,088][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:31:12,593][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:31:13,091][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:31:13,594][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:31:14,091][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:31:14,592][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:31:15,090][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:31:15,588][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:31:16,086][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:31:16,590][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:31:17,093][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:31:17,593][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:31:18,093][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:31:18,598][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:31:19,096][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:31:19,595][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:31:20,093][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:31:20,591][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:31:21,090][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:31:21,588][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:31:22,085][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:31:22,582][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:31:23,079][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:31:23,576][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:31:24,073][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:31:24,568][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:31:25,066][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:31:25,568][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:31:26,074][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:31:26,575][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:31:27,072][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:31:27,572][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:31:28,071][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:31:28,569][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:31:29,066][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:31:29,568][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:31:30,066][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:31:30,566][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:31:31,061][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:31:31,557][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:31:32,059][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:31:32,555][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:31:33,050][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:31:33,549][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:31:34,052][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:31:34,558][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:31:35,055][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:31:35,553][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:31:36,050][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:31:36,555][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10093 tokens. +[2026-03-26 03:31:37,317][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.09%, ΔTime: 00:00:32 +[2026-03-26 03:31:38,049][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:31:38,052][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:31:38,054][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:31:38,773][__main__][INFO] - Iteration 417 took 52s (31.83% Gen, 66.79% Train). Generation: 16s, Training: 34s. Estimated remaining time: 37h 12m 59s. Estimated total time: 43h 38m 58s. Time estimates for 10 more iterations: 8m 43s, 100 more iterations: 1h 27m 17s, 500 more iterations: 7h 16m 29s. +[2026-03-26 03:31:38,776][__main__][INFO] - Starting iteration 417. +[2026-03-26 03:31:39,176][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 41 and human policies 1. +[2026-03-26 03:31:39,176][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:31:56,631][__main__][INFO] - Number of regex retries in iteration 417: 0 +[2026-03-26 03:31:56,632][__main__][INFO] - agents played in iteration 417 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:31:57,431][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:31:57,455][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:31:57,479][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:31:57,503][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:31:57,504][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:31:57,504][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:31:58,220][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:31:58,680][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:31:59,187][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:31:59,690][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:32:00,195][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:32:00,693][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:32:01,193][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:32:01,702][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:32:02,213][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:32:02,715][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:32:03,221][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:32:03,722][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:32:04,223][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:32:04,729][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:32:05,231][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:32:05,732][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:32:06,236][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:32:06,743][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:32:07,246][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:32:07,748][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:32:08,249][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:32:08,751][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:32:09,254][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:32:09,755][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:32:10,255][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:32:10,752][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:32:11,255][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:32:11,768][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:32:12,267][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:32:12,767][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:32:13,270][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:32:13,769][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:32:14,267][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:32:14,768][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:32:15,266][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:32:15,765][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:32:16,263][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:32:16,761][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:32:17,260][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:32:17,757][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:32:18,259][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:32:18,762][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:32:19,264][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:32:19,765][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:32:20,264][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:32:20,762][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:32:21,264][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:32:21,761][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:32:22,259][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:32:22,761][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:32:23,259][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:32:23,757][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:32:24,255][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:32:24,752][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:32:25,249][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:32:25,746][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:32:26,244][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:32:26,749][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:32:27,260][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:32:27,762][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:32:28,260][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:32:28,760][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:32:29,260][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:32:29,761][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:32:30,263][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10234 tokens. +[2026-03-26 03:32:31,026][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 60.63%, Block Peak % of device VRAM: 62.20%, ΔTime: 00:00:32 +[2026-03-26 03:32:31,758][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:32:31,761][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:32:31,762][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:32:32,482][__main__][INFO] - Iteration 418 took 53s (32.75% Gen, 65.90% Train). Generation: 17s, Training: 35s. Estimated remaining time: 37h 58m 29s. Estimated total time: 44h 25m 22s. Time estimates for 10 more iterations: 8m 53s, 100 more iterations: 1h 28m 50s, 500 more iterations: 7h 24m 13s. +[2026-03-26 03:32:32,485][__main__][INFO] - Starting iteration 418. +[2026-03-26 03:32:32,885][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 41 and human policies 1. +[2026-03-26 03:32:32,886][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:32:48,613][__main__][INFO] - Number of regex retries in iteration 418: 0 +[2026-03-26 03:32:48,614][__main__][INFO] - agents played in iteration 418 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:32:49,369][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:32:49,392][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:32:49,416][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:32:49,439][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:32:49,439][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:32:49,440][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:32:50,149][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:32:50,600][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:32:51,103][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:32:51,602][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:32:52,101][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:32:52,603][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:32:53,103][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:32:53,609][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:32:54,112][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:32:54,609][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:32:55,107][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:32:55,611][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:32:56,109][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:32:56,606][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:32:57,104][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:32:57,601][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:32:58,099][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:32:58,595][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:32:59,091][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:32:59,586][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:33:00,083][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:33:00,579][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:33:01,075][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:33:01,570][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:33:02,068][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:33:02,570][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:33:03,066][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:33:03,562][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:33:04,062][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:33:04,579][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:33:05,079][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:33:05,583][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:33:06,090][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:33:06,595][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:33:07,096][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:33:07,600][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:33:08,102][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:33:08,602][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:33:09,103][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:33:09,602][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:33:10,104][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:33:10,605][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:33:11,103][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:33:11,602][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:33:12,098][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:33:12,595][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:33:13,097][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:33:13,594][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:33:14,091][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:33:14,591][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:33:15,090][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:33:15,590][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:33:16,088][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:33:16,588][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:33:17,087][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:33:17,586][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:33:18,087][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:33:18,583][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:33:19,080][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:33:19,578][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:33:20,075][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:33:20,570][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:33:21,068][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:33:21,566][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:33:22,063][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10092 tokens. +[2026-03-26 03:33:22,812][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.57%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:32 +[2026-03-26 03:33:23,539][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:33:23,541][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:33:23,543][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:33:24,262][__main__][INFO] - Iteration 419 took 51s (30.61% Gen, 67.99% Train). Generation: 15s, Training: 34s. Estimated remaining time: 36h 21m 8s. Estimated total time: 42h 48m 52s. Time estimates for 10 more iterations: 8m 33s, 100 more iterations: 1h 25m 37s, 500 more iterations: 7h 8m 8s. +[2026-03-26 03:33:24,264][__main__][INFO] - Starting iteration 419. +[2026-03-26 03:33:24,663][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 41 and human policies 1. +[2026-03-26 03:33:24,663][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:33:35,376][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 03:33:41,721][__main__][INFO] - Number of regex retries in iteration 419: 1 +[2026-03-26 03:33:41,722][__main__][INFO] - agents played in iteration 419 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:33:42,481][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:33:42,505][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:33:42,528][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:33:42,551][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:33:42,552][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:33:42,553][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:33:43,263][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:33:43,718][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:33:44,219][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:33:44,719][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:33:45,213][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:33:45,714][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:33:46,208][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:33:46,704][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:33:47,200][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:33:47,697][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:33:48,199][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:33:48,695][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:33:49,193][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:33:49,691][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:33:50,188][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:33:50,686][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:33:51,186][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:33:51,687][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:33:52,184][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:33:52,691][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:33:53,190][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:33:53,690][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:33:54,194][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:33:54,693][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:33:55,192][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:33:55,689][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:33:56,187][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:33:56,685][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:33:57,186][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:33:57,685][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:33:58,186][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:33:58,685][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:33:59,182][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:33:59,688][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:34:00,189][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:34:00,687][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:34:01,185][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:34:01,683][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:34:02,184][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:34:02,681][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:34:03,183][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:34:03,682][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:34:04,179][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:34:04,680][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:34:05,179][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:34:05,675][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:34:06,172][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:34:06,672][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:34:07,167][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:34:07,664][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:34:08,160][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:34:08,656][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:34:09,158][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:34:09,655][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:34:10,152][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:34:10,650][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:34:11,146][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:34:11,648][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:34:12,148][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:34:12,656][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:34:13,155][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:34:13,655][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:34:14,155][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:34:14,653][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:34:15,151][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10171 tokens. +[2026-03-26 03:34:15,917][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 60.56%, Block Peak % of device VRAM: 62.23%, ΔTime: 00:00:32 +[2026-03-26 03:34:16,652][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:34:16,654][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:34:16,655][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:34:17,378][__main__][INFO] - Iteration 420 took 52s (32.36% Gen, 66.27% Train). Generation: 17s, Training: 34s. Estimated remaining time: 37h 27m 11s. Estimated total time: 43h 55m 48s. Time estimates for 10 more iterations: 8m 47s, 100 more iterations: 1h 27m 51s, 500 more iterations: 7h 19m 18s. +[2026-03-26 03:34:17,380][__main__][INFO] - Starting iteration 420. +[2026-03-26 03:34:17,781][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 41 and human policies 1. +[2026-03-26 03:34:17,781][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:34:24,110][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 03:34:30,742][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 03:34:34,052][__main__][INFO] - Number of regex retries in iteration 420: 2 +[2026-03-26 03:34:34,053][__main__][INFO] - agents played in iteration 420 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:34:34,810][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:34:34,833][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:34:34,857][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:34:34,880][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:34:34,881][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:34:34,882][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:34:35,605][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:34:36,066][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:34:36,581][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:34:37,086][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:34:37,591][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:34:38,094][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:34:38,596][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:34:39,113][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:34:39,618][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:34:40,122][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:34:40,623][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:34:41,121][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:34:41,618][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:34:42,116][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:34:42,613][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:34:43,109][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:34:43,607][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:34:44,106][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:34:44,604][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:34:45,102][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:34:45,600][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:34:46,097][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:34:46,597][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:34:47,093][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:34:47,595][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:34:48,091][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:34:48,588][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:34:49,084][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:34:49,581][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:34:50,076][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:34:50,573][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:34:51,069][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:34:51,568][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:34:52,063][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:34:52,561][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:34:53,059][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:34:53,556][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:34:54,056][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:34:54,554][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:34:55,051][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:34:55,549][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:34:56,046][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:34:56,543][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:34:57,043][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:34:57,541][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:34:58,040][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:34:58,537][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:34:59,035][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:34:59,531][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:35:00,029][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:35:00,528][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:35:01,024][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:35:01,522][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:35:02,023][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:35:02,521][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:35:03,018][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:35:03,523][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:35:04,029][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:35:04,529][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:35:05,031][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:35:05,534][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:35:06,032][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:35:06,530][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:35:07,033][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:35:07,530][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10160 tokens. +[2026-03-26 03:35:08,293][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.47%, ΔTime: 00:00:32 +[2026-03-26 03:35:09,025][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:35:09,030][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:35:09,031][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:35:10,319][__main__][INFO] - Iteration 421 took 52s (30.97% Gen, 66.58% Train). Generation: 16s, Training: 34s. Estimated remaining time: 37h 17m 27s. Estimated total time: 43h 46m 57s. Time estimates for 10 more iterations: 8m 45s, 100 more iterations: 1h 27m 33s, 500 more iterations: 7h 17m 49s. +[2026-03-26 03:35:10,322][__main__][INFO] - Starting iteration 421. +[2026-03-26 03:35:10,720][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 42 and human policies 1. +[2026-03-26 03:35:10,721][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:35:27,321][__main__][INFO] - Number of regex retries in iteration 421: 0 +[2026-03-26 03:35:27,322][__main__][INFO] - agents played in iteration 421 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:35:28,082][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:35:28,106][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:35:28,129][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:35:28,153][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:35:28,153][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:35:28,154][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:35:28,884][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:35:29,336][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:35:29,842][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:35:30,346][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:35:30,860][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:35:31,363][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:35:31,862][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:35:32,360][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:35:32,862][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:35:33,362][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:35:33,860][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:35:34,358][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:35:34,862][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:35:35,361][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:35:35,858][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:35:36,355][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:35:36,852][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:35:37,356][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:35:37,866][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:35:38,366][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:35:38,864][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:35:39,364][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:35:39,861][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:35:40,360][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:35:40,860][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:35:41,359][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:35:41,858][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:35:42,356][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:35:42,857][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:35:43,354][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:35:43,852][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:35:44,355][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:35:44,852][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:35:45,350][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:35:45,846][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:35:46,342][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:35:46,836][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:35:47,331][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:35:47,826][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:35:48,320][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:35:48,815][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:35:49,311][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:35:49,807][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:35:50,308][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:35:50,806][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:35:51,304][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:35:51,801][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:35:52,299][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:35:52,798][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:35:53,293][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:35:53,789][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:35:54,288][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:35:54,786][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:35:55,284][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:35:55,782][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:35:56,279][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:35:56,779][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:35:57,275][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:35:57,772][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:35:58,271][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:35:58,769][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:35:59,265][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:35:59,762][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:36:00,260][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:36:00,755][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10067 tokens. +[2026-03-26 03:36:01,506][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.02%, Current % of VRAM taken: 60.45%, Block Peak % of device VRAM: 62.03%, ΔTime: 00:00:32 +[2026-03-26 03:36:02,319][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:36:02,321][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:36:02,323][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:36:03,063][__main__][INFO] - Iteration 422 took 52s (31.72% Gen, 66.87% Train). Generation: 16s, Training: 35s. Estimated remaining time: 37h 6m 46s. Estimated total time: 43h 37m 9s. Time estimates for 10 more iterations: 8m 43s, 100 more iterations: 1h 27m 14s, 500 more iterations: 7h 16m 11s. +[2026-03-26 03:36:03,065][__main__][INFO] - Starting iteration 422. +[2026-03-26 03:36:03,464][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 42 and human policies 1. +[2026-03-26 03:36:03,465][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:36:19,254][__main__][INFO] - Number of regex retries in iteration 422: 0 +[2026-03-26 03:36:19,255][__main__][INFO] - agents played in iteration 422 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:36:20,013][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:36:20,036][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:36:20,060][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:36:20,083][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:36:20,084][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:36:20,084][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:36:20,801][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:36:21,254][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:36:21,757][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:36:22,254][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:36:22,756][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:36:23,253][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:36:23,749][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:36:24,252][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:36:24,751][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:36:25,258][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:36:25,762][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:36:26,261][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:36:26,760][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:36:27,259][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:36:27,757][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:36:28,256][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:36:28,754][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:36:29,255][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:36:29,754][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:36:30,256][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:36:30,755][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:36:31,253][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:36:31,754][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:36:32,258][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:36:32,758][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:36:33,254][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:36:33,756][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:36:34,256][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:36:34,754][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:36:35,254][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:36:35,751][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:36:36,250][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:36:36,747][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:36:37,250][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:36:37,748][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:36:38,258][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:36:38,757][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:36:39,255][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:36:39,752][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:36:40,250][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:36:40,748][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:36:41,246][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:36:41,745][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:36:42,244][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:36:42,745][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:36:43,243][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:36:43,741][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:36:44,239][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:36:44,737][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:36:45,237][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:36:45,734][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:36:46,232][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:36:46,730][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:36:47,230][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:36:47,727][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:36:48,226][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:36:48,725][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:36:49,220][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:36:49,716][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:36:50,212][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:36:50,709][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:36:51,208][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:36:51,707][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:36:52,203][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:36:52,699][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10173 tokens. +[2026-03-26 03:36:53,457][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.09%, Current % of VRAM taken: 60.51%, Block Peak % of device VRAM: 62.23%, ΔTime: 00:00:32 +[2026-03-26 03:36:54,190][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:36:54,192][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:36:54,194][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:36:54,879][__main__][INFO] - Iteration 423 took 51s (30.71% Gen, 67.95% Train). Generation: 15s, Training: 34s. Estimated remaining time: 36h 19m 32s. Estimated total time: 42h 50m 46s. Time estimates for 10 more iterations: 8m 34s, 100 more iterations: 1h 25m 41s, 500 more iterations: 7h 8m 27s. +[2026-03-26 03:36:54,882][__main__][INFO] - Starting iteration 423. +[2026-03-26 03:36:55,283][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 42 and human policies 1. +[2026-03-26 03:36:55,284][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:37:11,389][__main__][INFO] - Number of regex retries in iteration 423: 0 +[2026-03-26 03:37:11,390][__main__][INFO] - agents played in iteration 423 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:37:12,147][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:37:12,170][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:37:12,194][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:37:12,218][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:37:12,218][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:37:12,219][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:37:12,942][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:37:13,394][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:37:13,898][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:37:14,396][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:37:14,894][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:37:15,392][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:37:15,889][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:37:16,391][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:37:16,889][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:37:17,386][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:37:17,883][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:37:18,381][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:37:18,882][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:37:19,377][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:37:19,873][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:37:20,367][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:37:20,863][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:37:21,363][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:37:21,858][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:37:22,355][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:37:22,852][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:37:23,348][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:37:23,845][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:37:24,343][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:37:24,841][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:37:25,338][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:37:25,835][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:37:26,332][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:37:26,828][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:37:27,325][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:37:27,822][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:37:28,321][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:37:28,817][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:37:29,314][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:37:29,810][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:37:30,310][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:37:30,806][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:37:31,308][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:37:31,812][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:37:32,311][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:37:32,812][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:37:33,312][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:37:33,811][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:37:34,311][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:37:34,810][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:37:35,307][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:37:41,863][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:37:43,689][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:37:44,190][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:37:44,688][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:37:45,186][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:37:45,688][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:37:46,185][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:37:46,684][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:37:47,190][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:37:47,689][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:37:48,187][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:37:48,687][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:37:49,186][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:37:49,683][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:37:50,179][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:37:50,675][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:37:51,171][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:37:51,667][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:37:52,177][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10074 tokens. +[2026-03-26 03:37:53,396][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.02%, Current % of VRAM taken: 60.45%, Block Peak % of device VRAM: 62.09%, ΔTime: 00:00:40 +[2026-03-26 03:37:54,445][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:37:54,447][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:37:54,449][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:37:55,468][__main__][INFO] - Iteration 424 took 1m 0s (26.76% Gen, 71.54% Train). Generation: 16s, Training: 43s. Estimated remaining time: 43h 37m 1s. Estimated total time: 50h 9m 16s. Time estimates for 10 more iterations: 10m 1s, 100 more iterations: 1h 40m 18s, 500 more iterations: 8h 21m 32s. +[2026-03-26 03:37:55,568][__main__][INFO] - Starting iteration 424. +[2026-03-26 03:37:55,974][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 42 and human policies 1. +[2026-03-26 03:37:55,975][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:39:24,531][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 03:39:32,739][__main__][INFO] - Number of regex retries in iteration 424: 1 +[2026-03-26 03:39:32,740][__main__][INFO] - agents played in iteration 424 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:39:33,768][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:39:33,792][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:39:33,815][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:39:33,838][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:39:33,839][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:39:33,839][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:39:34,654][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:39:35,111][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:39:35,615][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:39:36,112][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:39:36,609][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:39:37,107][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:39:37,607][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:39:38,104][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:39:38,602][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:39:39,100][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:39:39,597][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:39:40,093][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:39:40,589][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:39:41,090][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:39:41,587][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:39:42,084][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:39:42,582][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:39:43,080][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:39:43,577][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:39:44,073][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:39:44,569][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:39:45,069][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:39:45,565][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:39:46,060][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:39:46,558][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:39:47,055][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:39:47,551][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:39:48,049][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:39:48,546][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:39:49,042][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:39:49,539][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:39:50,035][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:39:50,531][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:39:51,028][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:39:51,526][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:39:52,024][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:39:52,522][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:39:53,021][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:39:53,521][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:39:54,017][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:39:54,517][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:39:55,013][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:39:55,509][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:39:56,005][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:39:56,507][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:39:57,006][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:39:57,507][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:39:58,013][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:39:58,512][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:39:59,016][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:39:59,519][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:40:00,024][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:40:00,523][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:40:01,024][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:40:01,524][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:40:02,023][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:40:02,525][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:40:03,027][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:40:03,531][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:40:04,030][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:40:04,528][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:40:05,027][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:40:05,525][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:40:06,023][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:40:06,524][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10121 tokens. +[2026-03-26 03:40:07,289][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.25%, Current % of VRAM taken: 60.68%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:32 +[2026-03-26 03:40:08,029][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:40:08,031][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:40:08,033][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:40:08,781][__main__][INFO] - Iteration 425 took 2m 12s (72.86% Gen, 26.57% Train). Generation: 1m 36s, Training: 35s. Estimated remaining time: 104h 5m 53s. Estimated total time: 110h 40m 21s. Time estimates for 10 more iterations: 22m 8s, 100 more iterations: 3h 41m 20s, 500 more iterations: 18h 26m 43s. +[2026-03-26 03:40:08,783][__main__][INFO] - Starting iteration 425. +[2026-03-26 03:40:09,184][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 42 and human policies 1. +[2026-03-26 03:40:09,184][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:41:04,207][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 03:41:08,305][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls + +Given the per-item values, you have a very high value for hats compared to Alice, who values hats very little. Therefore, it's strategically advantageous to keep all the hats for yourself, as you can easily meet or exceed the item quantity threshold (10 hats) with your own value, maximizing your points for this round. did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 03:41:16,503][__main__][INFO] - Number of regex retries in iteration 425: 2 +[2026-03-26 03:41:16,504][__main__][INFO] - agents played in iteration 425 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:41:17,264][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:41:17,287][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:41:17,311][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:41:17,334][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:41:17,335][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:41:17,335][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:41:18,057][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:41:18,512][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:41:19,015][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:41:19,512][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:41:20,013][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:41:20,510][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:41:21,011][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:41:21,507][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:41:22,005][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:41:22,503][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:41:23,002][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:41:23,501][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:41:24,000][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:41:24,498][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:41:24,996][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:41:25,494][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:41:25,991][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:41:26,487][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:41:26,984][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:41:27,483][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:41:27,985][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:41:28,486][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:41:28,983][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:41:29,484][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:41:29,983][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:41:30,480][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:41:30,983][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:41:31,480][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:41:31,980][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:41:32,479][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:41:32,984][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:41:33,482][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:41:33,983][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:41:34,479][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:41:34,975][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:41:35,471][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:41:35,971][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:41:36,468][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:41:36,964][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:41:37,462][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:41:37,962][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:41:38,458][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:41:38,963][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:41:39,462][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:41:39,965][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:41:40,474][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:41:40,976][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:41:41,478][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:41:41,978][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:41:42,484][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:41:42,985][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:41:43,487][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:41:43,989][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:41:44,490][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:41:44,991][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:41:45,492][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:41:45,990][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:41:46,491][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:41:46,991][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:41:47,488][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:41:47,988][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:41:48,488][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:41:49,015][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:41:49,515][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:41:50,015][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10261 tokens. +[2026-03-26 03:41:50,783][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.33%, ΔTime: 00:00:32 +[2026-03-26 03:41:51,536][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:41:51,540][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:41:51,544][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:41:52,274][__main__][INFO] - Iteration 426 took 1m 43s (65.30% Gen, 33.99% Train). Generation: 1m 7s, Training: 35s. Estimated remaining time: 79h 18m 20s. Estimated total time: 85h 54m 32s. Time estimates for 10 more iterations: 17m 10s, 100 more iterations: 2h 51m 49s, 500 more iterations: 14h 19m 5s. +[2026-03-26 03:41:52,276][__main__][INFO] - Starting iteration 426. +[2026-03-26 03:41:52,673][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 42 and human policies 1. +[2026-03-26 03:41:52,674][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:42:27,049][__main__][INFO] - Number of regex retries in iteration 426: 0 +[2026-03-26 03:42:27,050][__main__][INFO] - agents played in iteration 426 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:42:27,806][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:42:27,830][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:42:27,853][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:42:27,876][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:42:27,877][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:42:27,878][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:42:28,596][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:42:29,051][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:42:29,554][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:42:30,053][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:42:30,552][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:42:31,051][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:42:31,550][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:42:32,050][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:42:32,550][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:42:33,047][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:42:33,548][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:42:34,045][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:42:34,544][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:42:35,042][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:42:35,541][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:42:36,040][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:42:36,537][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:42:37,038][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:42:37,539][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:42:38,039][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:42:38,538][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:42:39,037][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:42:39,536][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:42:40,036][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:42:40,535][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:42:41,032][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:42:41,527][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:42:42,024][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:42:42,521][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:42:43,017][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:42:43,513][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:42:44,014][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:42:44,515][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:42:45,014][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:42:45,511][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:42:46,012][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:42:46,515][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:42:47,011][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:42:47,509][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:42:48,006][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:42:48,503][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:42:49,002][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:42:49,500][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:42:49,997][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:42:50,494][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:42:50,991][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:42:51,488][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:42:51,985][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:42:52,483][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:42:52,981][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:42:53,476][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:42:53,970][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:42:54,467][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:42:54,962][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:42:55,456][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:42:55,950][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:42:56,444][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:42:56,942][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:42:57,441][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:42:57,939][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:42:58,433][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:42:58,930][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:42:59,427][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:42:59,926][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:43:00,427][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10049 tokens. +[2026-03-26 03:43:01,187][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.25%, Current % of VRAM taken: 60.67%, Block Peak % of device VRAM: 62.21%, ΔTime: 00:00:32 +[2026-03-26 03:43:01,929][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:43:01,932][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:43:01,933][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:43:02,667][__main__][INFO] - Iteration 427 took 1m 9s (49.11% Gen, 49.84% Train). Generation: 34s, Training: 34s. Estimated remaining time: 51h 42m 22s. Estimated total time: 58h 19m 45s. Time estimates for 10 more iterations: 11m 39s, 100 more iterations: 1h 56m 39s, 500 more iterations: 9h 43m 17s. +[2026-03-26 03:43:02,669][__main__][INFO] - Starting iteration 427. +[2026-03-26 03:43:03,069][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 42 and human policies 1. +[2026-03-26 03:43:03,070][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:43:34,948][__main__][INFO] - Number of regex retries in iteration 427: 0 +[2026-03-26 03:43:34,949][__main__][INFO] - agents played in iteration 427 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:43:35,708][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:43:35,732][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:43:35,756][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:43:35,779][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:43:35,780][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:43:35,781][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:43:36,522][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:43:36,983][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:43:37,494][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:43:37,997][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:43:38,499][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:43:39,001][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:43:39,505][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:43:40,009][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:43:40,507][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:43:41,006][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:43:41,504][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:43:42,004][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:43:42,516][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:43:43,015][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:43:43,516][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:43:44,019][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:43:44,519][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:43:45,017][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:43:45,515][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:43:46,017][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:43:46,526][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:43:47,025][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:43:47,529][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:43:48,028][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:43:48,526][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:43:49,026][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:43:49,525][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:43:50,025][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:43:50,525][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:43:51,023][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:43:51,522][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:43:52,021][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:43:52,519][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:43:53,016][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:43:53,512][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:43:54,018][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:43:54,516][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:43:55,019][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:43:55,531][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:43:56,032][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:43:56,533][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:43:57,035][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:43:57,564][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:43:58,066][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:43:58,566][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:43:59,068][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:43:59,568][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:44:00,067][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:44:00,567][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:44:01,062][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:44:01,566][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:44:02,075][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:44:02,574][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:44:03,073][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:44:03,571][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:44:04,074][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:44:04,570][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:44:05,069][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:44:05,567][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:44:06,065][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:44:06,564][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:44:07,062][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:44:07,560][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:44:08,061][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:44:08,560][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10121 tokens. +[2026-03-26 03:44:09,316][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.23%, ΔTime: 00:00:32 +[2026-03-26 03:44:10,089][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:44:10,091][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:44:10,093][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:44:10,854][__main__][INFO] - Iteration 428 took 1m 7s (47.03% Gen, 51.85% Train). Generation: 31s, Training: 35s. Estimated remaining time: 49h 50m 48s. Estimated total time: 56h 29m 18s. Time estimates for 10 more iterations: 11m 17s, 100 more iterations: 1h 52m 58s, 500 more iterations: 9h 24m 53s. +[2026-03-26 03:44:10,857][__main__][INFO] - Starting iteration 428. +[2026-03-26 03:44:11,254][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 42 and human policies 1. +[2026-03-26 03:44:11,255][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:44:21,143][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 03:44:21,226][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 03:44:32,114][__main__][INFO] - Number of regex retries in iteration 428: 2 +[2026-03-26 03:44:32,115][__main__][INFO] - agents played in iteration 428 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:44:32,871][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:44:32,895][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:44:32,918][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:44:32,941][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:44:32,942][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:44:32,942][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:44:33,679][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:44:34,135][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:44:34,644][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:44:35,152][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:44:35,651][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:44:36,154][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:44:36,651][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:44:37,165][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:44:37,666][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:44:38,166][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:44:38,665][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:44:39,163][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:44:39,661][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:44:40,158][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:44:40,655][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:44:41,151][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:44:41,648][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:44:42,147][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:44:42,647][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:44:43,149][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:44:43,648][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:44:44,153][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:44:44,659][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:44:45,160][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:44:45,659][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:44:46,157][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:44:46,654][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:44:47,153][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:44:47,664][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:44:48,163][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:44:48,663][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:44:49,165][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:44:49,664][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:44:50,171][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:44:50,673][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:44:51,179][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:44:51,679][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:44:52,180][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:44:52,684][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:44:53,184][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:44:53,684][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:44:54,185][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:44:54,683][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:44:55,182][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:44:55,681][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:44:56,179][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:44:56,676][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:44:57,174][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:44:57,673][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:44:58,170][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:44:58,670][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:44:59,167][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:44:59,665][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:45:00,164][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:45:00,663][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:45:01,164][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:45:01,660][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:45:02,159][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:45:02,659][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:45:03,156][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:45:03,654][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:45:04,155][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:45:04,660][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:45:05,170][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:45:05,672][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10093 tokens. +[2026-03-26 03:45:06,487][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:32 +[2026-03-26 03:45:07,295][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:45:07,297][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:45:07,299][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:45:08,050][__main__][INFO] - Iteration 429 took 56s (36.73% Gen, 61.95% Train). Generation: 20s, Training: 35s. Estimated remaining time: 40h 40m 20s. Estimated total time: 47h 19m 47s. Time estimates for 10 more iterations: 9m 27s, 100 more iterations: 1h 34m 39s, 500 more iterations: 7h 53m 17s. +[2026-03-26 03:45:08,052][__main__][INFO] - Starting iteration 429. +[2026-03-26 03:45:08,453][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 42 and human policies 1. +[2026-03-26 03:45:08,454][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:45:47,770][__main__][INFO] - Number of regex retries in iteration 429: 0 +[2026-03-26 03:45:47,771][__main__][INFO] - agents played in iteration 429 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:45:48,534][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:45:48,560][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:45:48,585][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:45:48,610][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:45:48,611][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:45:48,611][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:45:49,362][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:45:49,820][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:45:50,326][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:45:50,826][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:45:51,323][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:45:51,820][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:45:52,318][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:45:52,816][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:45:53,320][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:45:53,826][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:45:54,326][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:45:54,833][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:45:55,334][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:45:55,835][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:45:56,333][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:45:56,832][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:45:57,334][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:45:57,832][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:45:58,330][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:45:58,830][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:45:59,327][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:45:59,831][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:46:00,342][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:46:00,841][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:46:01,340][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:46:01,842][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:46:02,342][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:46:02,840][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:46:03,341][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:46:03,838][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:46:04,334][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:46:04,831][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:46:05,331][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:46:05,830][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:46:06,330][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:46:06,828][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:46:07,324][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:46:07,824][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:46:08,320][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:46:08,816][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:46:09,313][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:46:09,834][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:46:10,337][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:46:10,837][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:46:11,335][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:46:11,833][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:46:12,331][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:46:12,829][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:46:13,328][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:46:13,827][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:46:14,323][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:46:14,819][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:46:15,315][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:46:15,810][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:46:16,307][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:46:16,804][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:46:17,302][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:46:17,803][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:46:18,298][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:46:18,794][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:46:19,290][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:46:19,785][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:46:20,286][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:46:20,794][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:46:21,292][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10074 tokens. +[2026-03-26 03:46:22,053][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.05%, Current % of VRAM taken: 60.48%, Block Peak % of device VRAM: 62.35%, ΔTime: 00:00:32 +[2026-03-26 03:46:22,858][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:46:22,861][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:46:22,862][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:46:23,628][__main__][INFO] - Iteration 430 took 1m 15s (52.30% Gen, 46.68% Train). Generation: 39s, Training: 35s. Estimated remaining time: 55h 58m 3s. Estimated total time: 62h 38m 46s. Time estimates for 10 more iterations: 12m 31s, 100 more iterations: 2h 5m 17s, 500 more iterations: 10h 26m 27s. +[2026-03-26 03:46:23,630][__main__][INFO] - Starting iteration 430. +[2026-03-26 03:46:24,028][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 42 and human policies 1. +[2026-03-26 03:46:24,028][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:46:39,270][__main__][INFO] - Number of regex retries in iteration 430: 0 +[2026-03-26 03:46:39,271][__main__][INFO] - agents played in iteration 430 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:46:40,027][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:46:40,051][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:46:40,074][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:46:40,098][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:46:40,098][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:46:40,099][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:46:40,825][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:46:41,280][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:46:41,782][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:46:42,282][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:46:42,785][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:46:43,284][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:46:43,782][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:46:44,285][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:46:44,797][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:46:45,296][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:46:45,797][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:46:46,296][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:46:46,795][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:46:47,296][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:46:47,794][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:46:48,294][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:46:48,794][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:46:49,292][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:46:49,793][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:46:50,292][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:46:50,790][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:46:51,289][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:46:51,787][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:46:52,287][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:46:52,789][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:46:53,285][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:46:53,784][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:46:54,283][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:46:54,782][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:46:55,280][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:46:55,778][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:46:56,282][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:46:56,780][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:46:57,282][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:46:57,783][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:46:58,281][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:46:58,781][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:46:59,287][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:46:59,789][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:47:00,300][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:47:00,802][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:47:01,300][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:47:01,802][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:47:02,306][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:47:02,805][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:47:03,304][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:47:03,804][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:47:04,303][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:47:04,804][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:47:05,300][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:47:05,796][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:47:06,294][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:47:06,792][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:47:07,290][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:47:07,787][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:47:08,285][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:47:08,780][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:47:09,279][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:47:09,775][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:47:10,273][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:47:10,773][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:47:11,268][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:47:11,765][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:47:12,262][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:47:12,759][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10139 tokens. +[2026-03-26 03:47:13,525][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.09%, Current % of VRAM taken: 60.51%, Block Peak % of device VRAM: 62.15%, ΔTime: 00:00:32 +[2026-03-26 03:47:14,274][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:47:14,276][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:47:14,277][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:47:15,568][__main__][INFO] - Iteration 431 took 51s (29.57% Gen, 67.92% Train). Generation: 15s, Training: 35s. Estimated remaining time: 36h 15m 28s. Estimated total time: 42h 57m 3s. Time estimates for 10 more iterations: 8m 35s, 100 more iterations: 1h 25m 54s, 500 more iterations: 7h 9m 30s. +[2026-03-26 03:47:15,570][__main__][INFO] - Starting iteration 431. +[2026-03-26 03:47:15,970][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 43 and human policies 1. +[2026-03-26 03:47:15,970][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:47:20,514][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 03:47:25,861][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 03:47:27,058][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 20 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 03:47:31,946][__main__][INFO] - Number of regex retries in iteration 431: 3 +[2026-03-26 03:47:31,946][__main__][INFO] - agents played in iteration 431 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:47:32,720][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:47:32,743][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:47:32,767][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:47:32,790][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:47:32,791][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:47:32,791][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:47:33,512][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:47:33,964][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:47:34,470][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:47:34,966][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:47:35,462][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:47:35,958][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:47:36,454][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:47:36,950][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:47:37,447][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:47:37,949][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:47:38,451][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:47:38,950][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:47:39,449][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:47:39,948][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:47:40,448][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:47:40,948][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:47:41,448][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:47:41,948][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:47:42,449][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:47:42,950][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:47:43,450][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:47:43,952][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:47:44,452][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:47:44,954][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:47:45,452][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:47:45,950][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:47:46,448][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:47:46,949][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:47:47,448][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:47:47,947][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:47:48,445][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:47:48,948][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:47:49,449][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:47:49,946][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:47:50,444][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:47:50,941][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:47:51,440][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:47:51,936][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:47:52,432][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:47:52,932][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:47:53,429][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:47:53,925][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:47:54,422][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:47:54,919][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:47:55,416][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:47:55,912][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:47:56,409][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:47:56,906][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:47:57,403][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:47:57,898][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:47:58,393][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:47:58,889][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:47:59,384][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:47:59,879][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:48:00,373][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:48:00,868][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:48:01,363][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:48:01,866][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:48:02,366][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:48:02,867][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:48:03,373][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:48:03,870][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:48:04,367][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:48:04,868][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:48:05,365][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10098 tokens. +[2026-03-26 03:48:06,140][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.02%, Current % of VRAM taken: 60.45%, Block Peak % of device VRAM: 62.45%, ΔTime: 00:00:32 +[2026-03-26 03:48:06,886][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:48:06,889][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:48:06,890][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:48:07,617][__main__][INFO] - Iteration 432 took 51s (30.93% Gen, 67.66% Train). Generation: 15s, Training: 34s. Estimated remaining time: 36h 19m 57s. Estimated total time: 43h 2m 24s. Time estimates for 10 more iterations: 8m 36s, 100 more iterations: 1h 26m 4s, 500 more iterations: 7h 10m 24s. +[2026-03-26 03:48:07,619][__main__][INFO] - Starting iteration 432. +[2026-03-26 03:48:08,020][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 43 and human policies 1. +[2026-03-26 03:48:08,021][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:48:25,160][__main__][INFO] - Number of regex retries in iteration 432: 0 +[2026-03-26 03:48:25,161][__main__][INFO] - agents played in iteration 432 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:48:25,922][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:48:25,946][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:48:25,972][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:48:25,997][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:48:25,998][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:48:25,999][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:48:26,741][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:48:27,201][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:48:27,710][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:48:28,210][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:48:28,720][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:48:29,224][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:48:29,727][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:48:30,225][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:48:30,725][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:48:31,224][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:48:31,729][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:48:32,253][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:48:32,750][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:48:33,247][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:48:33,748][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:48:34,246][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:48:34,742][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:48:35,238][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:48:35,735][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:48:36,232][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:48:36,728][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:48:37,225][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:48:37,722][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:48:38,218][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:48:38,713][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:48:39,217][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:48:39,723][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:48:40,226][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:48:40,728][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:48:41,225][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:48:41,723][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:48:42,227][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:48:42,736][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:48:43,240][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:48:43,743][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:48:44,245][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:48:44,746][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:48:45,248][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:48:45,749][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:48:46,247][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:48:46,746][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:48:47,246][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:48:47,744][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:48:48,243][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:48:48,742][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:48:49,240][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:48:49,752][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:48:50,252][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:48:50,755][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:48:51,256][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:48:51,755][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:48:52,256][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:48:52,754][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:48:53,253][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:48:53,758][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:48:54,266][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:48:54,765][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:48:55,264][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:48:55,762][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:48:56,258][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:48:56,754][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:48:57,252][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:48:57,752][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:48:58,248][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:48:58,744][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10147 tokens. +[2026-03-26 03:48:59,507][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 60.49%, Block Peak % of device VRAM: 62.21%, ΔTime: 00:00:32 +[2026-03-26 03:49:00,243][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:49:00,245][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:49:00,247][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:49:00,977][__main__][INFO] - Iteration 433 took 52s (32.37% Gen, 66.25% Train). Generation: 17s, Training: 35s. Estimated remaining time: 37h 24m 32s. Estimated total time: 44h 7m 53s. Time estimates for 10 more iterations: 8m 49s, 100 more iterations: 1h 28m 15s, 500 more iterations: 7h 21m 18s. +[2026-03-26 03:49:00,979][__main__][INFO] - Starting iteration 433. +[2026-03-26 03:49:01,377][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 43 and human policies 1. +[2026-03-26 03:49:01,378][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:49:17,351][__main__][INFO] - Number of regex retries in iteration 433: 0 +[2026-03-26 03:49:17,352][__main__][INFO] - agents played in iteration 433 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:49:18,113][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:49:18,137][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:49:18,161][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:49:18,184][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:49:18,185][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:49:18,185][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:49:18,915][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:49:19,369][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:49:19,875][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:49:20,373][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:49:20,871][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:49:21,380][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:49:21,890][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:49:22,392][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:49:22,892][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:49:23,391][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:49:23,890][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:49:24,391][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:49:24,888][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:49:25,390][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:49:25,888][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:49:26,386][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:49:26,885][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:49:27,386][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:49:27,883][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:49:28,384][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:49:28,888][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:49:29,401][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:49:29,901][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:49:30,401][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:49:30,901][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:49:31,402][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:49:31,904][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:49:32,407][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:49:32,906][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:49:33,405][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:49:33,903][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:49:34,401][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:49:34,904][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:49:35,403][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:49:35,901][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:49:36,403][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:49:36,904][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:49:37,402][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:49:37,900][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:49:38,398][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:49:38,895][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:49:39,392][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:49:39,889][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:49:40,385][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:49:40,883][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:49:41,379][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:49:41,875][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:49:42,371][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:49:42,867][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:49:43,365][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:49:43,864][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:49:44,362][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:49:44,864][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:49:45,361][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:49:45,863][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:49:46,359][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:49:46,858][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:49:47,364][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:49:47,864][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:49:48,363][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:49:48,863][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:49:49,365][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:49:49,865][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:49:50,366][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:49:50,870][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10107 tokens. +[2026-03-26 03:49:51,655][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 60.57%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:32 +[2026-03-26 03:49:52,429][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:49:52,432][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:49:52,434][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:49:53,159][__main__][INFO] - Iteration 434 took 51s (30.85% Gen, 67.75% Train). Generation: 15s, Training: 35s. Estimated remaining time: 36h 24m 55s. Estimated total time: 43h 9m 8s. Time estimates for 10 more iterations: 8m 37s, 100 more iterations: 1h 26m 18s, 500 more iterations: 7h 11m 31s. +[2026-03-26 03:49:53,161][__main__][INFO] - Starting iteration 434. +[2026-03-26 03:49:53,561][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 43 and human policies 1. +[2026-03-26 03:49:53,562][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:50:09,502][__main__][INFO] - Number of regex retries in iteration 434: 0 +[2026-03-26 03:50:09,503][__main__][INFO] - agents played in iteration 434 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:50:10,264][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:50:10,289][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:50:10,313][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:50:10,338][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:50:10,338][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:50:10,339][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:50:11,075][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:50:11,536][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:50:12,039][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:50:12,550][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:50:13,050][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:50:13,550][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:50:14,050][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:50:14,549][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:50:15,048][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:50:15,547][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:50:16,046][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:50:16,548][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:50:17,045][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:50:17,546][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:50:18,045][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:50:18,543][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:50:19,041][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:50:19,539][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:50:20,042][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:50:20,538][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:50:21,037][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:50:21,541][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:50:22,044][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:50:22,549][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:50:23,052][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:50:23,561][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:50:24,068][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:50:24,570][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:50:25,070][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:50:25,573][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:50:26,072][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:50:26,574][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:50:27,073][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:50:27,572][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:50:28,070][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:50:28,571][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:50:29,071][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:50:29,569][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:50:30,067][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:50:30,566][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:50:31,064][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:50:31,563][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:50:32,062][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:50:32,559][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:50:33,057][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:50:33,553][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:50:34,051][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:50:34,547][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:50:35,046][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:50:35,543][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:50:36,040][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:50:36,538][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:50:37,035][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:50:37,532][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:50:38,029][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:50:38,525][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:50:39,021][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:50:39,523][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:50:40,023][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:50:40,523][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:50:41,023][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:50:41,521][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:50:42,020][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:50:42,519][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:50:43,019][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10132 tokens. +[2026-03-26 03:50:43,792][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 +[2026-03-26 03:50:44,543][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:50:44,545][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:50:44,547][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:50:45,257][__main__][INFO] - Iteration 435 took 51s (30.84% Gen, 67.79% Train). Generation: 15s, Training: 35s. Estimated remaining time: 36h 19m 42s. Estimated total time: 43h 4m 47s. Time estimates for 10 more iterations: 8m 36s, 100 more iterations: 1h 26m 9s, 500 more iterations: 7h 10m 47s. +[2026-03-26 03:50:45,259][__main__][INFO] - Starting iteration 435. +[2026-03-26 03:50:45,659][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 43 and human policies 1. +[2026-03-26 03:50:45,660][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:50:53,366][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 03:51:03,620][__main__][INFO] - Number of regex retries in iteration 435: 1 +[2026-03-26 03:51:03,621][__main__][INFO] - agents played in iteration 435 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:51:04,383][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:51:04,408][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:51:04,433][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:51:04,457][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:51:04,458][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:51:04,459][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:51:05,208][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:51:05,671][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:51:06,179][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:51:06,685][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:51:07,187][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:51:07,702][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:51:08,207][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:51:08,709][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:51:09,209][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:51:09,712][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:51:10,216][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:51:10,715][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:51:11,217][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:51:11,717][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:51:12,217][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:51:12,718][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:51:13,218][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:51:13,715][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:51:14,214][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:51:14,716][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:51:15,222][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:51:15,727][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:51:16,227][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:51:16,727][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:51:17,227][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:51:17,725][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:51:18,222][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:51:18,729][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:51:19,234][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:51:19,734][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:51:20,233][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:51:20,731][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:51:21,229][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:51:21,728][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:51:22,230][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:51:22,728][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:51:23,227][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:51:23,726][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:51:24,225][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:51:24,726][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:51:25,227][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:51:25,726][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:51:26,223][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:51:26,720][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:51:27,222][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:51:27,718][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:51:28,215][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:51:28,715][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:51:29,212][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:51:29,710][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:51:30,208][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:51:30,710][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:51:31,210][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:51:31,707][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:51:32,206][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:51:32,705][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:51:33,201][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:51:33,704][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:51:34,205][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:51:34,711][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:51:35,222][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:51:35,722][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:51:36,225][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:51:36,725][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:51:37,227][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10233 tokens. +[2026-03-26 03:51:37,998][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 60.64%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:32 +[2026-03-26 03:51:38,753][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:51:38,756][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:51:38,757][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:51:39,477][__main__][INFO] - Iteration 436 took 53s (33.37% Gen, 65.29% Train). Generation: 17s, Training: 35s. Estimated remaining time: 38h 4m 55s. Estimated total time: 44h 50m 55s. Time estimates for 10 more iterations: 8m 58s, 100 more iterations: 1h 29m 41s, 500 more iterations: 7h 28m 29s. +[2026-03-26 03:51:39,479][__main__][INFO] - Starting iteration 436. +[2026-03-26 03:51:39,876][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 43 and human policies 1. +[2026-03-26 03:51:39,876][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:51:55,663][__main__][INFO] - Number of regex retries in iteration 436: 0 +[2026-03-26 03:51:55,664][__main__][INFO] - agents played in iteration 436 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:51:56,418][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:51:56,441][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:51:56,464][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:51:56,488][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:51:56,488][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:51:56,490][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:51:57,215][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:51:57,667][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:51:58,169][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:51:58,674][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:51:59,184][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:51:59,685][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:52:00,189][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:52:00,689][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:52:01,190][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:52:01,691][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:52:02,191][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:52:02,690][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:52:03,188][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:52:03,686][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:52:04,184][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:52:04,681][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:52:05,178][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:52:05,678][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:52:06,188][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:52:06,686][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:52:07,185][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:52:07,684][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:52:08,181][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:52:08,678][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:52:09,176][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:52:09,673][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:52:10,172][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:52:10,671][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:52:11,172][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:52:11,671][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:52:12,170][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:52:12,669][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:52:13,166][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:52:13,662][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:52:14,159][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:52:14,660][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:52:15,155][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:52:15,652][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:52:16,148][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:52:16,644][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:52:17,140][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:52:17,635][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:52:18,130][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:52:18,626][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:52:19,121][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:52:19,616][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:52:20,111][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:52:20,608][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:52:21,104][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:52:21,600][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:52:22,099][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:52:22,593][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:52:23,088][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:52:23,583][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:52:24,079][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:52:24,575][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:52:25,069][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:52:25,566][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:52:26,071][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:52:26,587][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:52:27,087][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:52:27,589][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:52:28,089][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:52:28,588][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:52:29,088][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9982 tokens. +[2026-03-26 03:52:29,862][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 60.57%, Block Peak % of device VRAM: 62.15%, ΔTime: 00:00:32 +[2026-03-26 03:52:30,613][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:52:30,616][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:52:30,617][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:52:31,341][__main__][INFO] - Iteration 437 took 51s (30.68% Gen, 67.92% Train). Generation: 15s, Training: 34s. Estimated remaining time: 36h 6m 26s. Estimated total time: 42h 53m 17s. Time estimates for 10 more iterations: 8m 34s, 100 more iterations: 1h 25m 46s, 500 more iterations: 7h 8m 52s. +[2026-03-26 03:52:31,343][__main__][INFO] - Starting iteration 437. +[2026-03-26 03:52:31,758][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 43 and human policies 1. +[2026-03-26 03:52:31,759][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:52:48,501][__main__][INFO] - Number of regex retries in iteration 437: 0 +[2026-03-26 03:52:48,502][__main__][INFO] - agents played in iteration 437 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:52:49,262][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:52:49,286][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:52:49,310][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:52:49,333][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:52:49,334][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:52:49,335][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:52:50,062][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:52:50,515][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:52:51,015][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:52:51,513][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:52:52,009][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:52:52,509][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:52:53,006][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:52:53,504][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:52:54,002][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:52:54,504][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:52:55,001][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:52:55,497][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:52:55,997][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:52:56,514][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:52:57,014][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:52:57,514][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:52:58,014][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:52:58,517][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:52:59,017][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:52:59,522][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:53:00,026][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:53:00,527][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:53:01,028][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:53:01,532][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:53:02,039][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:53:02,537][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:53:03,036][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:53:03,535][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:53:04,034][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:53:04,532][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:53:05,031][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:53:05,531][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:53:06,030][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:53:06,527][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:53:07,027][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:53:07,524][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:53:08,022][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:53:08,519][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:53:09,017][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:53:09,514][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:53:10,014][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:53:10,517][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:53:11,018][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:53:11,525][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:53:12,027][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:53:12,528][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:53:13,029][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:53:13,530][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:53:14,029][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:53:14,529][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:53:15,027][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:53:15,528][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:53:16,026][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:53:16,524][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:53:17,023][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:53:17,519][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:53:18,017][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:53:18,517][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:53:19,015][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:53:19,512][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:53:20,009][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:53:20,507][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:53:21,007][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:53:21,505][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:53:22,003][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10042 tokens. +[2026-03-26 03:53:22,761][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.58%, Block Peak % of device VRAM: 62.17%, ΔTime: 00:00:32 +[2026-03-26 03:53:23,508][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:53:23,510][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:53:23,512][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:53:24,279][__main__][INFO] - Iteration 438 took 52s (31.88% Gen, 66.66% Train). Generation: 16s, Training: 35s. Estimated remaining time: 36h 58m 19s. Estimated total time: 43h 46m 3s. Time estimates for 10 more iterations: 8m 45s, 100 more iterations: 1h 27m 32s, 500 more iterations: 7h 17m 40s. +[2026-03-26 03:53:24,281][__main__][INFO] - Starting iteration 438. +[2026-03-26 03:53:24,680][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 43 and human policies 1. +[2026-03-26 03:53:24,681][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:53:41,925][__main__][INFO] - Number of regex retries in iteration 438: 0 +[2026-03-26 03:53:41,926][__main__][INFO] - agents played in iteration 438 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:53:42,685][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:53:42,709][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:53:42,733][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:53:42,756][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:53:42,756][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:53:42,757][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:53:43,488][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:53:43,946][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:53:44,447][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:53:44,948][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:53:45,449][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:53:45,964][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:53:46,469][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:53:46,970][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:53:47,469][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:53:47,968][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:53:48,468][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:53:48,972][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:53:49,484][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:53:49,985][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:53:50,485][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:53:50,986][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:53:51,486][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:53:51,985][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:53:52,484][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:53:52,987][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:53:53,496][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:53:53,996][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:53:54,500][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:53:54,998][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:53:55,497][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:53:55,995][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:53:56,493][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:53:56,993][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:53:57,493][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:53:57,991][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:53:58,492][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:53:58,996][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:53:59,494][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:53:59,992][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:54:00,494][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:54:00,993][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:54:01,494][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:54:01,993][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:54:02,490][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:54:02,989][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:54:03,488][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:54:03,987][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:54:04,484][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:54:04,983][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:54:05,479][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:54:05,974][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:54:06,472][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:54:06,968][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:54:07,467][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:54:07,967][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:54:08,470][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:54:08,981][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:54:09,480][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:54:09,978][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:54:10,477][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:54:10,979][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:54:11,475][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:54:11,973][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:54:12,474][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:54:12,974][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:54:13,472][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:54:13,971][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:54:14,469][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:54:14,968][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:54:15,467][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10114 tokens. +[2026-03-26 03:54:16,238][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 60.53%, Block Peak % of device VRAM: 62.08%, ΔTime: 00:00:32 +[2026-03-26 03:54:17,028][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:54:17,030][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:54:17,032][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:54:17,750][__main__][INFO] - Iteration 439 took 53s (32.50% Gen, 66.15% Train). Generation: 17s, Training: 35s. Estimated remaining time: 37h 24m 55s. Estimated total time: 44h 13m 32s. Time estimates for 10 more iterations: 8m 50s, 100 more iterations: 1h 28m 27s, 500 more iterations: 7h 22m 15s. +[2026-03-26 03:54:17,752][__main__][INFO] - Starting iteration 439. +[2026-03-26 03:54:18,152][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 43 and human policies 1. +[2026-03-26 03:54:18,153][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:54:24,930][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 03:54:35,431][__main__][INFO] - Number of regex retries in iteration 439: 1 +[2026-03-26 03:54:35,431][__main__][INFO] - agents played in iteration 439 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:54:36,190][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:54:36,214][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:54:36,237][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:54:36,260][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:54:36,261][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:54:36,262][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:54:36,979][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:54:37,435][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:54:37,933][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:54:38,434][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:54:38,931][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:54:39,426][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:54:39,923][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:54:40,419][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:54:40,914][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:54:41,410][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:54:41,905][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:54:42,409][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:54:42,918][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:54:43,419][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:54:43,918][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:54:44,419][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:54:44,917][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:54:45,415][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:54:45,913][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:54:46,410][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:54:46,912][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:54:47,408][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:54:47,906][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:54:48,404][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:54:48,901][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:54:49,401][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:54:49,897][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:54:50,395][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:54:50,892][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:54:51,390][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:54:51,886][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:54:52,384][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:54:52,881][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:54:53,378][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:54:53,876][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:54:54,372][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:54:54,869][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:54:55,372][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:54:55,869][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:54:56,366][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:54:56,866][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:54:57,362][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:54:57,858][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:54:58,354][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:54:58,850][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:54:59,347][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:54:59,848][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:55:00,345][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:55:00,842][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:55:01,338][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:55:01,836][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:55:02,332][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:55:02,830][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:55:03,327][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:55:03,822][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:55:04,317][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:55:04,810][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:55:05,307][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:55:05,803][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:55:06,298][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:55:06,792][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:55:07,287][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:55:07,786][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:55:08,280][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:55:08,778][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10040 tokens. +[2026-03-26 03:55:09,533][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 60.56%, Block Peak % of device VRAM: 62.18%, ΔTime: 00:00:32 +[2026-03-26 03:55:10,337][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:55:10,339][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:55:10,341][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:55:11,072][__main__][INFO] - Iteration 440 took 52s (32.65% Gen, 65.97% Train). Generation: 17s, Training: 34s. Estimated remaining time: 37h 16m 30s. Estimated total time: 44h 6m 1s. Time estimates for 10 more iterations: 8m 49s, 100 more iterations: 1h 28m 12s, 500 more iterations: 7h 21m 0s. +[2026-03-26 03:55:11,074][__main__][INFO] - Starting iteration 440. +[2026-03-26 03:55:11,472][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 43 and human policies 1. +[2026-03-26 03:55:11,473][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:55:17,119][mllm.models.large_language_model_local][WARNING] - Response Proposition: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 03:55:28,682][__main__][INFO] - Number of regex retries in iteration 440: 1 +[2026-03-26 03:55:28,683][__main__][INFO] - agents played in iteration 440 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:55:29,442][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:55:29,466][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:55:29,489][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:55:29,512][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:55:29,513][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:55:29,513][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:55:30,231][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:55:30,689][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:55:31,189][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:55:31,691][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:55:32,200][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:55:32,710][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:55:33,213][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:55:33,712][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:55:34,212][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:55:34,712][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:55:35,218][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:55:35,723][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:55:36,225][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:55:36,726][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:55:37,227][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:55:37,727][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:55:38,226][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:55:38,736][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:55:39,245][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:55:39,750][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:55:40,253][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:55:40,755][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:55:41,258][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:55:41,758][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:55:42,258][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:55:42,757][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:55:43,269][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:55:43,770][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:55:44,270][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:55:44,772][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:55:45,271][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:55:45,770][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:55:46,271][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:55:46,773][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:55:47,271][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:55:47,769][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:55:48,268][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:55:48,771][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:55:49,278][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:55:49,777][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:55:50,276][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:55:50,777][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:55:51,277][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:55:51,776][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:55:52,275][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:55:52,776][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:55:53,274][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:55:53,776][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:55:54,274][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:55:54,771][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:55:55,270][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:55:55,775][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:55:56,286][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:55:56,788][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:55:57,287][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:55:57,790][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:55:58,291][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:55:58,790][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:55:59,289][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:55:59,788][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:56:00,288][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:56:00,786][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:56:01,287][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:56:01,788][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:56:02,291][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10143 tokens. +[2026-03-26 03:56:03,051][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 60.61%, Block Peak % of device VRAM: 62.27%, ΔTime: 00:00:32 +[2026-03-26 03:56:03,797][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:56:03,800][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:56:03,801][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:56:05,104][__main__][INFO] - Iteration 441 took 53s (32.09% Gen, 65.48% Train). Generation: 17s, Training: 35s. Estimated remaining time: 37h 51m 10s. Estimated total time: 44h 41m 35s. Time estimates for 10 more iterations: 8m 56s, 100 more iterations: 1h 29m 23s, 500 more iterations: 7h 26m 55s. +[2026-03-26 03:56:05,106][__main__][INFO] - Starting iteration 441. +[2026-03-26 03:56:05,505][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 44 and human policies 1. +[2026-03-26 03:56:05,506][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:56:21,544][__main__][INFO] - Number of regex retries in iteration 441: 0 +[2026-03-26 03:56:21,545][__main__][INFO] - agents played in iteration 441 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:56:22,299][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:56:22,323][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:56:22,348][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:56:22,373][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:56:22,373][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:56:22,374][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:56:23,114][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:56:23,575][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:56:24,083][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:56:24,585][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:56:25,085][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:56:25,585][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:56:26,085][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:56:26,591][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:56:27,094][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:56:27,596][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:56:28,095][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:56:28,594][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:56:29,098][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:56:29,597][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:56:30,104][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:56:30,603][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:56:31,106][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:56:31,604][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:56:32,102][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:56:32,599][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:56:33,096][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:56:33,594][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:56:34,096][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:56:34,601][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:56:35,101][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:56:35,628][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:56:36,130][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:56:36,632][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:56:37,132][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:56:37,633][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:56:38,135][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:56:38,633][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:56:39,132][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:56:39,630][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:56:40,130][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:56:40,631][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:56:41,132][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:56:41,633][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:56:42,133][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:56:42,633][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:56:43,133][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:56:43,631][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:56:44,131][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:56:44,629][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:56:45,127][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:56:45,627][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:56:46,125][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:56:46,622][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:56:47,119][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:56:47,617][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:56:48,118][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:56:48,624][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:56:49,124][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:56:49,623][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:56:50,126][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:56:50,623][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:56:51,122][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:56:51,624][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:56:52,121][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:56:52,619][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:56:53,117][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:56:53,614][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:56:54,113][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:56:54,611][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:56:55,110][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10085 tokens. +[2026-03-26 03:56:55,881][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 60.53%, Block Peak % of device VRAM: 62.16%, ΔTime: 00:00:32 +[2026-03-26 03:56:56,630][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:56:56,633][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:56:56,634][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:56:57,371][__main__][INFO] - Iteration 442 took 51s (30.92% Gen, 67.65% Train). Generation: 16s, Training: 35s. Estimated remaining time: 36h 22m 3s. Estimated total time: 43h 13m 20s. Time estimates for 10 more iterations: 8m 38s, 100 more iterations: 1h 26m 26s, 500 more iterations: 7h 12m 13s. +[2026-03-26 03:56:57,373][__main__][INFO] - Starting iteration 442. +[2026-03-26 03:56:57,771][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 44 and human policies 1. +[2026-03-26 03:56:57,772][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:57:09,906][mllm.models.large_language_model_local][WARNING] - Response Proposal: x hats, y books, 10 balls + +Given the values, I see that the balls are much more valuable to me (10) compared to hats and books (1 each). However, Bob values books much more than I do. Since we have to propound the amount of each item we will keep, and the total sum of our proposals can't exceed the item quantity, it's strategic to take as many balls as possible (10 of them) to maximize my points. For hats and books, I'll propose to take 0 of each, as it's better to keep more of the highly valued balls. did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 03:57:10,089][mllm.models.large_language_model_local][WARNING] - Response Proposal: x hats, y books, 10 balls + +Given the values: +- I value hats at 1, books at 1, and balls at 10. +- Bob values hats at 1, books at 10, and balls at 1. -Given the values, I see an opportunity to maximize my points by focusing on the items where my value is higher than Alice's. Specifically, hats are worth 10 points each for me, but only 1 for Alice. Books are valued equally at 10 each, and balls are worth 1 each for me, but 10 for Alice. +Since the item quantity for balls is 10 and I value them the highest at 10, I will propose to keep all 10 balls. For hats and books, since both value them equally at 1, I'll split them equally to avoid giving Bob an advantage in those items. Thus, I propose to keep 5 hats and 5 books. did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 03:57:15,394][__main__][INFO] - Number of regex retries in iteration 442: 2 +[2026-03-26 03:57:15,394][__main__][INFO] - agents played in iteration 442 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:57:16,155][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:57:16,179][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:57:16,203][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:57:16,227][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:57:16,227][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:57:16,228][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:57:16,972][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:57:17,432][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:57:17,935][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:57:18,434][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:57:18,940][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:57:19,451][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:57:19,950][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:57:20,449][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:57:20,947][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:57:21,449][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:57:21,952][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:57:22,458][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:57:22,959][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:57:23,457][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:57:23,955][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:57:24,452][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:57:24,951][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:57:25,450][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:57:25,948][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:57:26,448][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:57:26,947][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:57:27,446][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:57:27,944][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:57:28,441][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:57:28,938][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:57:29,459][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:57:29,960][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:57:30,464][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:57:30,963][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:57:31,466][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:57:31,969][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:57:32,473][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:57:32,975][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:57:33,476][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:57:33,977][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:57:34,477][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:57:34,979][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:57:35,475][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:57:35,973][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:57:36,475][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:57:36,972][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:57:37,472][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:57:37,969][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:57:38,467][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:57:38,965][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:57:39,465][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:57:39,962][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:57:40,459][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:57:40,956][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:57:41,459][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:57:41,963][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:57:42,464][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:57:42,966][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:57:43,467][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:57:43,967][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:57:44,467][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:57:44,966][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:57:45,465][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:57:45,966][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:57:46,469][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:57:46,979][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:57:47,479][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:57:47,979][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:57:48,483][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:57:48,986][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10144 tokens. +[2026-03-26 03:57:49,755][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 60.63%, Block Peak % of device VRAM: 62.18%, ΔTime: 00:00:32 +[2026-03-26 03:57:50,505][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:57:50,508][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:57:50,509][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:57:51,231][__main__][INFO] - Iteration 443 took 53s (32.96% Gen, 65.68% Train). Generation: 17s, Training: 35s. Estimated remaining time: 37h 40m 49s. Estimated total time: 44h 33m 0s. Time estimates for 10 more iterations: 8m 54s, 100 more iterations: 1h 29m 6s, 500 more iterations: 7h 25m 30s. +[2026-03-26 03:57:51,233][__main__][INFO] - Starting iteration 443. +[2026-03-26 03:57:51,634][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 44 and human policies 1. +[2026-03-26 03:57:51,634][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:58:08,609][__main__][INFO] - Number of regex retries in iteration 443: 0 +[2026-03-26 03:58:08,610][__main__][INFO] - agents played in iteration 443 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:58:09,370][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:58:09,394][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:58:09,418][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:58:09,442][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:58:09,443][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:58:09,443][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:58:10,186][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:58:10,648][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:58:11,152][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:58:11,651][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:58:12,150][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:58:12,651][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:58:13,155][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:58:13,662][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:58:14,160][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:58:14,661][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:58:15,161][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:58:15,661][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:58:16,160][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:58:16,665][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:58:17,165][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:58:17,664][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:58:18,172][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:58:18,685][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:58:19,184][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:58:19,688][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:58:20,190][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:58:20,688][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:58:21,187][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:58:21,684][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:58:22,181][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:58:22,681][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:58:23,180][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:58:23,678][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:58:24,182][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:58:24,688][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:58:25,191][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:58:25,694][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:58:26,194][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:58:26,698][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:58:27,201][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:58:27,704][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:58:28,203][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:58:28,703][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:58:29,205][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:58:29,703][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:58:30,206][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:58:30,704][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:58:31,209][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:58:31,720][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:58:32,222][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:58:32,722][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:58:33,222][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:58:33,719][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:58:34,218][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:58:34,715][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:58:35,216][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:58:35,712][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:58:36,213][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:58:36,713][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:58:37,216][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:58:37,714][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:58:38,212][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:58:38,712][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:58:39,210][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:58:39,712][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:58:40,214][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:58:40,712][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:58:41,213][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:58:41,713][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:58:42,212][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10181 tokens. +[2026-03-26 03:58:42,977][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.58%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:32 +[2026-03-26 03:58:43,727][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:58:43,729][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:58:43,730][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:58:44,491][__main__][INFO] - Iteration 444 took 52s (32.12% Gen, 66.44% Train). Generation: 16s, Training: 35s. Estimated remaining time: 37h 9m 50s. Estimated total time: 44h 2m 54s. Time estimates for 10 more iterations: 8m 48s, 100 more iterations: 1h 28m 5s, 500 more iterations: 7h 20m 29s. +[2026-03-26 03:58:44,493][__main__][INFO] - Starting iteration 444. +[2026-03-26 03:58:44,893][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 44 and human policies 1. +[2026-03-26 03:58:44,894][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:59:00,872][__main__][INFO] - Number of regex retries in iteration 444: 0 +[2026-03-26 03:59:00,873][__main__][INFO] - agents played in iteration 444 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:59:01,634][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:59:01,658][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:59:01,681][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:59:01,705][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:59:01,705][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:59:01,706][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:59:02,438][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:59:02,890][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:59:03,396][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:59:03,895][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:59:04,396][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:59:04,902][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:59:05,401][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:59:05,905][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:59:06,409][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:59:06,909][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:59:07,409][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:59:07,913][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:59:08,414][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:59:08,914][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 03:59:09,414][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 03:59:09,914][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 03:59:10,413][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 03:59:10,913][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 03:59:11,415][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 03:59:11,915][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 03:59:12,413][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 03:59:12,913][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 03:59:13,414][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 03:59:13,914][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 03:59:14,414][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 03:59:14,911][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 03:59:15,409][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 03:59:15,909][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 03:59:16,406][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 03:59:16,911][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 03:59:17,410][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 03:59:17,910][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 03:59:18,416][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 03:59:18,919][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 03:59:19,418][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 03:59:19,920][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 03:59:20,423][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 03:59:20,922][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 03:59:21,420][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 03:59:21,921][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 03:59:22,418][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 03:59:22,920][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 03:59:23,417][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 03:59:23,917][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 03:59:24,415][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 03:59:24,913][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 03:59:25,412][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 03:59:25,911][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 03:59:26,409][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 03:59:26,906][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 03:59:27,406][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 03:59:27,903][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 03:59:28,402][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 03:59:28,898][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 03:59:29,394][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 03:59:29,891][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 03:59:30,387][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 03:59:30,891][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 03:59:31,391][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 03:59:31,894][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 03:59:32,395][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 03:59:32,894][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 03:59:33,392][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 03:59:33,891][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 03:59:34,391][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10146 tokens. +[2026-03-26 03:59:35,162][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.58%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:32 +[2026-03-26 03:59:35,909][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 03:59:35,911][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 03:59:35,913][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 03:59:36,637][__main__][INFO] - Iteration 445 took 51s (30.88% Gen, 67.72% Train). Generation: 15s, Training: 35s. Estimated remaining time: 36h 13m 17s. Estimated total time: 43h 7m 13s. Time estimates for 10 more iterations: 8m 37s, 100 more iterations: 1h 26m 14s, 500 more iterations: 7h 11m 12s. +[2026-03-26 03:59:36,639][__main__][INFO] - Starting iteration 445. +[2026-03-26 03:59:37,039][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 44 and human policies 1. +[2026-03-26 03:59:37,040][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 03:59:41,176][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 03:59:51,716][__main__][INFO] - Number of regex retries in iteration 445: 1 +[2026-03-26 03:59:51,717][__main__][INFO] - agents played in iteration 445 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 03:59:52,476][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:59:52,503][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:59:52,529][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:59:52,555][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 03:59:52,556][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 03:59:52,557][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 03:59:53,290][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 03:59:53,744][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 03:59:54,254][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 03:59:54,753][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 03:59:55,252][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 03:59:55,751][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 03:59:56,254][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 03:59:56,765][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 03:59:57,265][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 03:59:57,763][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 03:59:58,262][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 03:59:58,761][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 03:59:59,259][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 03:59:59,761][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:00:00,261][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:00:00,761][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:00:01,257][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:00:01,752][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:00:02,250][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:00:02,749][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:00:03,247][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:00:03,745][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:00:04,242][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:00:04,738][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:00:05,234][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:00:05,731][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:00:06,234][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:00:06,742][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:00:07,246][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:00:07,744][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:00:08,243][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:00:08,741][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:00:09,238][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:00:09,738][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:00:10,235][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:00:10,737][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:00:11,240][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:00:11,738][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:00:12,236][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:00:12,733][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:00:13,231][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:00:13,728][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:00:14,232][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:00:14,730][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:00:15,229][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:00:15,728][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:00:16,230][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:00:16,739][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:00:17,239][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:00:17,739][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:00:18,240][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:00:18,742][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:00:19,245][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:00:19,752][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:00:20,251][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:00:20,750][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:00:21,252][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:00:21,749][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:00:22,250][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:00:22,752][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:00:23,262][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:00:23,760][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:00:24,258][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:00:24,756][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:00:25,255][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10061 tokens. +[2026-03-26 04:00:26,028][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 60.49%, Block Peak % of device VRAM: 62.22%, ΔTime: 00:00:32 +[2026-03-26 04:00:26,737][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:00:26,740][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:00:26,741][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:00:27,465][__main__][INFO] - Iteration 446 took 50s (29.11% Gen, 69.46% Train). Generation: 14s, Training: 35s. Estimated remaining time: 35h 6m 31s. Estimated total time: 42h 1m 18s. Time estimates for 10 more iterations: 8m 24s, 100 more iterations: 1h 24m 2s, 500 more iterations: 7h 0m 13s. +[2026-03-26 04:00:27,467][__main__][INFO] - Starting iteration 446. +[2026-03-26 04:00:27,867][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 44 and human policies 1. +[2026-03-26 04:00:27,867][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:00:44,311][__main__][INFO] - Number of regex retries in iteration 446: 0 +[2026-03-26 04:00:44,312][__main__][INFO] - agents played in iteration 446 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:00:45,071][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:00:45,095][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:00:45,118][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:00:45,142][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:00:45,142][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:00:45,143][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:00:45,869][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:00:46,321][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:00:46,829][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:00:47,332][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:00:47,832][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:00:48,336][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:00:48,836][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:00:49,334][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:00:49,831][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:00:50,331][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:00:50,837][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:00:51,351][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:00:51,850][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:00:52,353][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:00:52,854][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:00:53,355][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:00:53,856][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:00:54,380][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:00:54,881][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:00:55,377][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:00:55,879][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:00:56,381][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:00:56,881][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:00:57,384][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:00:57,886][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:00:58,383][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:00:58,888][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:00:59,389][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:00:59,888][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:01:00,389][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:01:00,887][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:01:01,391][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:01:01,893][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:01:02,403][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:01:02,903][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:01:03,401][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:01:03,904][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:01:04,401][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:01:04,905][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:01:05,405][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:01:05,903][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:01:06,402][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:01:06,900][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:01:07,399][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:01:07,896][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:01:08,395][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:01:08,893][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:01:09,392][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:01:09,890][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:01:10,389][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:01:10,886][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:01:11,383][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:01:11,885][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:01:12,382][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:01:12,886][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:01:13,385][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:01:13,884][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:01:14,381][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:01:14,878][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:01:15,377][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:01:15,875][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:01:16,373][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:01:16,872][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:01:17,369][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:01:17,867][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10144 tokens. +[2026-03-26 04:01:18,623][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:32 +[2026-03-26 04:01:19,393][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:01:19,395][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:01:19,397][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:01:20,214][__main__][INFO] - Iteration 447 took 52s (31.41% Gen, 67.02% Train). Generation: 16s, Training: 35s. Estimated remaining time: 36h 41m 42s. Estimated total time: 43h 37m 22s. Time estimates for 10 more iterations: 8m 43s, 100 more iterations: 1h 27m 14s, 500 more iterations: 7h 16m 13s. +[2026-03-26 04:01:20,216][__main__][INFO] - Starting iteration 447. +[2026-03-26 04:01:20,616][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 44 and human policies 1. +[2026-03-26 04:01:20,616][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:01:39,291][__main__][INFO] - Number of regex retries in iteration 447: 0 +[2026-03-26 04:01:39,291][__main__][INFO] - agents played in iteration 447 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:01:40,051][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:01:40,074][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:01:40,098][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:01:40,121][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:01:40,122][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:01:40,122][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:01:40,843][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:01:41,296][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:01:41,803][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:01:42,302][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:01:42,800][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:01:43,296][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:01:43,798][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:01:44,294][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:01:44,791][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:01:45,291][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:01:45,790][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:01:46,294][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:01:46,795][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:01:47,294][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:01:47,793][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:01:48,292][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:01:48,792][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:01:49,290][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:01:49,787][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:01:50,284][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:01:50,781][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:01:51,281][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:01:51,776][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:01:52,275][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:01:52,771][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:01:53,269][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:01:53,766][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:01:54,267][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:01:54,766][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:01:55,264][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:01:55,762][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:01:56,260][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:01:56,758][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:01:57,255][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:01:57,750][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:01:58,247][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:01:58,746][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:01:59,243][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:01:59,739][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:02:00,239][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:02:00,738][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:02:01,239][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:02:01,740][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:02:02,240][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:02:02,740][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:02:03,239][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:02:03,738][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:02:04,238][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:02:04,738][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:02:05,237][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:02:05,731][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:02:06,228][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:02:06,727][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:02:07,227][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:02:07,723][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:02:08,219][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:02:08,715][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:02:09,215][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:02:09,713][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:02:10,212][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:02:10,712][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:02:11,212][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:02:11,708][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:02:12,206][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:02:12,703][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10273 tokens. +[2026-03-26 04:02:13,469][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 60.49%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:32 +[2026-03-26 04:02:14,217][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:02:14,220][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:02:14,221][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:02:14,930][__main__][INFO] - Iteration 448 took 54s (34.38% Gen, 64.31% Train). Generation: 18s, Training: 34s. Estimated remaining time: 38h 19m 12s. Estimated total time: 45h 15m 47s. Time estimates for 10 more iterations: 9m 3s, 100 more iterations: 1h 30m 31s, 500 more iterations: 7h 32m 37s. +[2026-03-26 04:02:14,932][__main__][INFO] - Starting iteration 448. +[2026-03-26 04:02:15,333][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 44 and human policies 1. +[2026-03-26 04:02:15,333][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:02:21,566][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 04:02:29,908][__main__][INFO] - Number of regex retries in iteration 448: 1 +[2026-03-26 04:02:29,908][__main__][INFO] - agents played in iteration 448 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:02:30,663][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:02:30,687][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:02:30,710][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:02:30,734][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:02:30,735][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:02:30,735][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:02:31,489][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:02:31,948][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:02:32,452][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:02:32,951][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:02:33,451][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:02:33,953][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:02:34,451][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:02:34,953][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:02:35,458][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:02:35,959][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:02:36,459][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:02:36,963][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:02:37,462][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:02:37,962][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:02:38,461][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:02:38,962][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:02:39,465][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:02:39,980][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:02:40,483][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:02:40,983][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:02:41,487][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:02:41,987][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:02:42,490][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:02:42,991][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:02:43,492][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:02:43,990][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:02:44,492][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:02:44,996][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:02:45,507][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:02:46,009][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:02:46,508][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:02:47,007][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:02:47,510][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:02:48,010][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:02:48,510][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:02:49,013][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:02:49,513][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:02:50,013][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:02:50,513][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:02:51,016][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:02:51,519][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:02:52,019][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:02:52,524][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:02:53,035][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:02:53,536][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:02:54,037][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:02:54,538][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:02:55,040][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:02:55,542][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:02:56,039][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:02:56,537][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:02:57,036][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:02:57,537][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:02:58,035][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:02:58,536][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:02:59,033][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:02:59,531][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:03:00,029][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:03:00,526][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:03:01,023][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:03:01,520][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:03:02,023][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:03:02,520][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:03:03,018][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:03:03,515][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10085 tokens. +[2026-03-26 04:03:04,299][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 60.49%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 +[2026-03-26 04:03:05,071][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:03:05,074][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:03:05,076][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:03:05,806][__main__][INFO] - Iteration 449 took 50s (28.88% Gen, 69.67% Train). Generation: 14s, Training: 35s. Estimated remaining time: 35h 6m 16s. Estimated total time: 42h 3m 42s. Time estimates for 10 more iterations: 8m 24s, 100 more iterations: 1h 24m 7s, 500 more iterations: 7h 0m 37s. +[2026-03-26 04:03:05,808][__main__][INFO] - Starting iteration 449. +[2026-03-26 04:03:06,210][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 44 and human policies 1. +[2026-03-26 04:03:06,211][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:03:21,076][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 04:03:22,239][__main__][INFO] - Number of regex retries in iteration 449: 1 +[2026-03-26 04:03:22,240][__main__][INFO] - agents played in iteration 449 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:03:23,001][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:03:23,025][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:03:23,049][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:03:23,073][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:03:23,073][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:03:23,074][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:03:23,812][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:03:24,274][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:03:24,782][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:03:25,284][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:03:25,784][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:03:26,284][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:03:26,784][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:03:27,301][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:03:27,804][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:03:28,304][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:03:28,804][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:03:29,303][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:03:29,801][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:03:30,301][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:03:30,812][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:03:31,313][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:03:31,816][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:03:32,316][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:03:32,813][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:03:33,311][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:03:33,808][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:03:34,305][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:03:34,802][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:03:35,305][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:03:35,810][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:03:36,310][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:03:36,809][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:03:37,312][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:03:37,811][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:03:38,312][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:03:38,814][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:03:39,310][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:03:39,808][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:03:40,305][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:03:40,808][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:03:41,315][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:03:41,815][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:03:42,314][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:03:42,813][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:03:43,311][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:03:43,809][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:03:44,309][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:03:44,808][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:03:45,307][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:03:45,806][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:03:46,303][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:03:46,804][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:03:47,301][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:03:47,803][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:03:48,315][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:03:48,818][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:03:49,319][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:03:49,824][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:03:50,326][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:03:50,827][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:03:51,325][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:03:51,824][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:03:52,322][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:03:52,818][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:03:53,314][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:03:53,813][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:03:54,310][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:03:54,808][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:03:55,303][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:03:55,799][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10012 tokens. +[2026-03-26 04:03:56,573][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.02%, Current % of VRAM taken: 60.44%, Block Peak % of device VRAM: 62.06%, ΔTime: 00:00:32 +[2026-03-26 04:03:57,325][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:03:57,327][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:03:57,329][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:03:58,158][__main__][INFO] - Iteration 450 took 51s (30.86% Gen, 67.54% Train). Generation: 16s, Training: 35s. Estimated remaining time: 36h 19m 9s. Estimated total time: 43h 17m 27s. Time estimates for 10 more iterations: 8m 39s, 100 more iterations: 1h 26m 34s, 500 more iterations: 7h 12m 54s. +[2026-03-26 04:03:58,160][__main__][INFO] - Starting iteration 450. +[2026-03-26 04:03:58,560][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 44 and human policies 1. +[2026-03-26 04:03:58,560][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:04:14,379][__main__][INFO] - Number of regex retries in iteration 450: 0 +[2026-03-26 04:04:14,379][__main__][INFO] - agents played in iteration 450 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:04:15,134][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:04:15,158][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:04:15,182][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:04:15,205][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:04:15,206][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:04:15,206][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:04:15,944][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:04:16,401][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:04:16,913][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:04:17,413][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:04:17,922][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:04:18,422][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:04:18,922][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:04:19,420][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:04:19,920][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:04:20,420][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:04:20,918][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:04:21,419][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:04:21,917][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:04:22,414][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:04:22,914][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:04:23,416][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:04:23,913][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:04:24,416][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:04:24,915][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:04:25,414][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:04:25,914][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:04:26,413][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:04:26,912][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:04:27,423][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:04:27,925][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:04:28,436][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:04:28,938][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:04:29,439][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:04:29,940][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:04:30,439][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:04:30,937][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:04:31,437][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:04:31,940][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:04:32,439][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:04:32,937][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:04:33,436][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:04:33,937][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:04:34,435][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:04:34,935][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:04:35,433][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:04:35,932][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:04:36,432][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:04:36,930][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:04:37,427][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:04:37,922][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:04:38,418][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:04:38,915][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:04:39,412][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:04:39,908][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:04:40,406][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:04:40,904][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:04:41,405][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:04:41,904][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:04:42,401][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:04:42,898][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:04:43,397][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:04:43,894][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:04:44,390][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:04:44,886][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:04:45,382][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:04:45,881][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:04:46,376][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:04:46,872][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:04:47,368][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:04:47,867][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10145 tokens. +[2026-03-26 04:04:48,638][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 60.60%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:32 +[2026-03-26 04:04:49,400][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:04:49,402][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:04:49,404][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:04:50,866][__main__][INFO] - Iteration 451 took 52s (30.24% Gen, 66.96% Train). Generation: 15s, Training: 35s. Estimated remaining time: 36h 36m 9s. Estimated total time: 43h 35m 20s. Time estimates for 10 more iterations: 8m 43s, 100 more iterations: 1h 27m 10s, 500 more iterations: 7h 15m 53s. +[2026-03-26 04:04:50,868][__main__][INFO] - Starting iteration 451. +[2026-03-26 04:04:51,267][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 45 and human policies 1. +[2026-03-26 04:04:51,267][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:05:06,396][__main__][INFO] - Number of regex retries in iteration 451: 0 +[2026-03-26 04:05:06,397][__main__][INFO] - agents played in iteration 451 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:05:07,160][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:05:07,184][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:05:07,207][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:05:07,230][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:05:07,231][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:05:07,232][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:05:07,965][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:05:08,420][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:05:08,927][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:05:09,427][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:05:09,924][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:05:10,420][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:05:10,917][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:05:11,415][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:05:11,918][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:05:12,414][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:05:12,913][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:05:13,415][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:05:13,918][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:05:14,417][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:05:14,915][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:05:15,420][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:05:15,923][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:05:16,427][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:05:16,928][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:05:17,431][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:05:17,932][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:05:18,435][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:05:18,935][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:05:19,440][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:05:19,941][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:05:20,440][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:05:20,938][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:05:21,436][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:05:21,935][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:05:22,433][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:05:22,930][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:05:23,427][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:05:23,925][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:05:24,426][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:05:24,926][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:05:25,429][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:05:25,928][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:05:26,427][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:05:26,927][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:05:27,426][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:05:27,927][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:05:28,425][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:05:28,922][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:05:29,423][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:05:29,920][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:05:30,417][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:05:30,917][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:05:31,415][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:05:31,915][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:05:32,414][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:05:32,919][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:05:33,418][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:05:33,917][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:05:34,423][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:05:34,924][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:05:35,424][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:05:35,923][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:05:36,422][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:05:36,920][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:05:37,421][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:05:37,923][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:05:38,422][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:05:38,922][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:05:39,420][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:05:39,918][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10171 tokens. +[2026-03-26 04:05:40,687][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 +[2026-03-26 04:05:41,441][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:05:41,443][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:05:41,445][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:05:42,171][__main__][INFO] - Iteration 452 took 50s (29.72% Gen, 68.85% Train). Generation: 15s, Training: 35s. Estimated remaining time: 35h 25m 14s. Estimated total time: 42h 25m 16s. Time estimates for 10 more iterations: 8m 29s, 100 more iterations: 1h 24m 50s, 500 more iterations: 7h 4m 12s. +[2026-03-26 04:05:42,174][__main__][INFO] - Starting iteration 452. +[2026-03-26 04:05:42,574][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 45 and human policies 1. +[2026-03-26 04:05:42,575][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:05:57,210][__main__][INFO] - Number of regex retries in iteration 452: 0 +[2026-03-26 04:05:57,211][__main__][INFO] - agents played in iteration 452 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:05:57,964][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:05:57,988][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:05:58,011][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:05:58,035][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:05:58,035][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:05:58,036][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:05:58,760][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:05:59,213][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:05:59,713][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:06:00,218][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:06:00,725][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:06:01,226][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:06:01,727][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:06:02,227][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:06:02,726][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:06:03,237][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:06:03,737][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:06:04,237][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:06:04,737][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:06:05,235][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:06:05,734][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:06:06,236][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:06:06,737][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:06:07,240][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:06:07,746][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:06:08,248][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:06:08,748][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:06:09,249][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:06:09,749][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:06:10,251][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:06:10,772][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:06:11,273][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:06:11,771][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:06:12,269][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:06:12,769][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:06:13,268][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:06:13,766][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:06:14,264][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:06:14,761][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:06:15,260][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:06:15,758][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:06:16,256][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:06:16,755][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:06:17,253][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:06:17,752][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:06:18,250][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:06:18,748][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:06:19,246][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:06:19,746][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:06:20,246][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:06:20,744][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:06:21,245][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:06:21,745][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:06:22,243][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:06:22,740][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:06:23,241][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:06:23,737][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:06:24,234][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:06:24,734][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:06:25,233][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:06:25,732][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:06:26,231][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:06:26,728][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:06:27,228][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:06:27,723][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:06:28,219][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:06:28,716][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:06:29,212][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:06:29,709][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:06:30,208][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:06:30,707][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10051 tokens. +[2026-03-26 04:06:31,467][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.57%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 +[2026-03-26 04:06:32,225][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:06:32,227][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:06:32,229][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:06:32,958][__main__][INFO] - Iteration 453 took 50s (29.05% Gen, 69.50% Train). Generation: 14s, Training: 35s. Estimated remaining time: 34h 58m 19s. Estimated total time: 41h 59m 12s. Time estimates for 10 more iterations: 8m 23s, 100 more iterations: 1h 23m 58s, 500 more iterations: 6h 59m 52s. +[2026-03-26 04:06:32,960][__main__][INFO] - Starting iteration 453. +[2026-03-26 04:06:33,357][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 45 and human policies 1. +[2026-03-26 04:06:33,357][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:06:50,518][__main__][INFO] - Number of regex retries in iteration 453: 0 +[2026-03-26 04:06:50,519][__main__][INFO] - agents played in iteration 453 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:06:51,278][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:06:51,302][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:06:51,326][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:06:51,351][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:06:51,351][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:06:51,352][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:06:52,090][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:06:52,549][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:06:53,053][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:06:53,561][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:06:54,073][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:06:54,574][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:06:55,075][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:06:55,577][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:06:56,079][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:06:56,576][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:06:57,078][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:06:57,580][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:06:58,079][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:06:58,578][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:06:59,080][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:06:59,587][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:07:00,090][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:07:00,590][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:07:01,091][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:07:01,590][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:07:02,092][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:07:02,592][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:07:03,091][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:07:03,594][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:07:04,093][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:07:04,593][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:07:05,098][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:07:05,602][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:07:06,101][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:07:06,601][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:07:07,098][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:07:07,601][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:07:08,098][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:07:08,596][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:07:09,097][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:07:09,597][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:07:10,098][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:07:10,595][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:07:11,097][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:07:11,597][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:07:12,098][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:07:12,600][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:07:13,118][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:07:13,620][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:07:14,122][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:07:14,624][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:07:15,122][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:07:15,622][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:07:16,122][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:07:16,620][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:07:17,119][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:07:17,617][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:07:18,116][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:07:18,616][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:07:19,113][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:07:19,612][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:07:20,110][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:07:20,608][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:07:21,106][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:07:21,607][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:07:22,104][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:07:22,601][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:07:23,098][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:07:23,597][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:07:24,094][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10207 tokens. +[2026-03-26 04:07:24,864][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 60.50%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 +[2026-03-26 04:07:25,564][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:07:25,567][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:07:25,568][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:07:26,297][__main__][INFO] - Iteration 454 took 52s (32.42% Gen, 66.21% Train). Generation: 17s, Training: 35s. Estimated remaining time: 37h 5m 15s. Estimated total time: 44h 7m 1s. Time estimates for 10 more iterations: 8m 49s, 100 more iterations: 1h 28m 14s, 500 more iterations: 7h 21m 10s. +[2026-03-26 04:07:26,299][__main__][INFO] - Starting iteration 454. +[2026-03-26 04:07:26,699][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 45 and human policies 1. +[2026-03-26 04:07:26,700][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:07:42,477][__main__][INFO] - Number of regex retries in iteration 454: 0 +[2026-03-26 04:07:42,478][__main__][INFO] - agents played in iteration 454 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:07:43,234][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:07:43,258][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:07:43,281][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:07:43,304][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:07:43,305][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:07:43,305][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:07:44,043][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:07:44,496][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:07:44,999][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:07:45,496][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:07:45,999][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:07:46,502][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:07:47,010][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:07:47,510][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:07:48,010][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:07:48,513][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:07:49,016][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:07:49,518][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:07:50,023][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:07:50,523][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:07:51,025][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:07:51,524][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:07:52,026][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:07:52,528][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:07:53,049][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:07:53,551][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:07:54,052][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:07:54,553][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:07:55,056][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:07:55,557][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:07:56,058][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:07:56,554][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:07:57,052][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:07:57,550][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:07:58,047][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:07:58,547][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:07:59,043][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:07:59,547][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:08:00,050][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:08:00,559][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:08:01,058][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:08:01,557][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:08:02,060][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:08:02,562][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:08:03,061][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:08:03,560][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:08:04,061][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:08:04,564][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:08:05,065][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:08:05,569][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:08:06,070][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:08:06,569][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:08:07,071][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:08:07,571][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:08:08,068][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:08:08,567][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:08:09,065][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:08:09,562][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:08:10,061][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:08:10,558][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:08:11,059][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:08:11,556][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:08:12,057][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:08:12,555][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:08:13,051][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:08:13,550][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:08:14,047][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:08:14,544][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:08:15,040][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:08:15,537][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:08:16,034][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10146 tokens. +[2026-03-26 04:08:16,811][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 60.61%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 +[2026-03-26 04:08:17,547][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:08:17,550][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:08:17,551][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:08:18,417][__main__][INFO] - Iteration 455 took 51s (30.51% Gen, 67.82% Train). Generation: 15s, Training: 35s. Estimated remaining time: 36h 3m 16s. Estimated total time: 43h 5m 54s. Time estimates for 10 more iterations: 8m 37s, 100 more iterations: 1h 26m 11s, 500 more iterations: 7h 10m 59s. +[2026-03-26 04:08:18,419][__main__][INFO] - Starting iteration 455. +[2026-03-26 04:08:18,818][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 45 and human policies 1. +[2026-03-26 04:08:18,819][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:08:33,803][__main__][INFO] - Number of regex retries in iteration 455: 0 +[2026-03-26 04:08:33,803][__main__][INFO] - agents played in iteration 455 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:08:34,557][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:08:34,580][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:08:34,604][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:08:34,627][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:08:34,628][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:08:34,628][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:08:35,361][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:08:35,813][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:08:36,320][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:08:36,817][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:08:37,313][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:08:37,810][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:08:38,308][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:08:38,806][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:08:39,303][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:08:39,800][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:08:40,296][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:08:40,798][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:08:41,297][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:08:41,806][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:08:42,307][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:08:42,810][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:08:43,310][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:08:43,811][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:08:44,312][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:08:44,811][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:08:45,311][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:08:45,814][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:08:46,313][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:08:46,815][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:08:47,312][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:08:47,812][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:08:48,310][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:08:48,808][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:08:49,306][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:08:49,803][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:08:50,298][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:08:50,794][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:08:51,294][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:08:51,791][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:08:52,290][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:08:52,789][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:08:53,287][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:08:53,784][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:08:54,286][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:08:54,782][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:08:55,279][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:08:55,775][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:08:56,272][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:08:56,770][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:08:57,269][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:08:57,774][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:08:58,274][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:08:58,772][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:08:59,278][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:08:59,777][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:09:00,277][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:09:00,776][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:09:01,276][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:09:01,776][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:09:02,279][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:09:02,776][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:09:03,277][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:09:03,776][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:09:04,277][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:09:04,773][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:09:05,270][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:09:05,767][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:09:06,266][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:09:06,765][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:09:07,261][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10017 tokens. +[2026-03-26 04:09:08,030][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.04%, Current % of VRAM taken: 60.47%, Block Peak % of device VRAM: 62.11%, ΔTime: 00:00:32 +[2026-03-26 04:09:08,779][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:09:08,781][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:09:08,783][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:09:09,486][__main__][INFO] - Iteration 456 took 50s (29.57% Gen, 69.04% Train). Generation: 14s, Training: 34s. Estimated remaining time: 35h 9m 55s. Estimated total time: 42h 13m 24s. Time estimates for 10 more iterations: 8m 26s, 100 more iterations: 1h 24m 26s, 500 more iterations: 7h 2m 14s. +[2026-03-26 04:09:09,488][__main__][INFO] - Starting iteration 456. +[2026-03-26 04:09:09,886][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 45 and human policies 1. +[2026-03-26 04:09:09,886][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:09:15,667][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 04:09:24,812][__main__][INFO] - Number of regex retries in iteration 456: 1 +[2026-03-26 04:09:24,813][__main__][INFO] - agents played in iteration 456 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:09:25,576][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:09:25,600][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:09:25,623][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:09:25,646][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:09:25,647][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:09:25,647][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:09:26,378][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:09:26,832][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:09:27,335][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:09:27,842][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:09:28,358][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:09:28,859][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:09:29,359][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:09:29,859][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:09:30,357][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:09:30,864][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:09:31,371][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:09:31,873][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:09:32,374][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:09:32,874][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:09:33,372][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:09:33,871][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:09:34,376][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:09:34,889][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:09:35,389][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:09:35,892][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:09:36,395][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:09:36,895][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:09:37,395][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:09:37,896][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:09:38,395][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:09:38,899][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:09:39,407][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:09:39,911][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:09:40,412][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:09:40,911][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:09:41,411][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:09:41,912][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:09:42,411][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:09:42,908][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:09:43,405][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:09:43,900][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:09:44,397][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:09:44,897][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:09:45,407][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:09:45,905][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:09:46,402][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:09:46,903][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:09:47,406][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:09:47,904][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:09:48,401][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:09:48,899][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:09:49,396][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:09:49,896][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:09:50,396][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:09:50,895][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:09:51,396][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:09:51,900][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:09:52,413][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:09:52,915][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:09:53,417][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:09:53,944][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:09:54,450][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:09:54,947][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:09:55,444][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:09:55,940][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:09:56,438][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:09:56,934][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:09:57,432][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:09:57,929][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:09:58,425][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10108 tokens. +[2026-03-26 04:09:59,191][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 60.49%, Block Peak % of device VRAM: 62.36%, ΔTime: 00:00:32 +[2026-03-26 04:09:59,979][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:09:59,981][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:09:59,983][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:10:00,758][__main__][INFO] - Iteration 457 took 50s (29.34% Gen, 69.13% Train). Generation: 14s, Training: 35s. Estimated remaining time: 35h 19m 17s. Estimated total time: 42h 23m 38s. Time estimates for 10 more iterations: 8m 28s, 100 more iterations: 1h 24m 47s, 500 more iterations: 7h 3m 56s. +[2026-03-26 04:10:00,760][__main__][INFO] - Starting iteration 457. +[2026-03-26 04:10:01,157][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 45 and human policies 1. +[2026-03-26 04:10:01,158][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:10:10,270][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 04:10:16,599][__main__][INFO] - Number of regex retries in iteration 457: 1 +[2026-03-26 04:10:16,600][__main__][INFO] - agents played in iteration 457 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:10:17,353][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:10:17,377][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:10:17,400][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:10:17,423][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:10:17,424][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:10:17,425][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:10:18,165][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:10:18,620][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:10:19,123][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:10:19,620][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:10:20,123][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:10:20,623][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:10:21,128][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:10:21,630][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:10:22,130][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:10:22,632][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:10:23,140][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:10:23,640][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:10:24,140][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:10:24,644][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:10:25,143][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:10:25,643][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:10:26,142][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:10:26,641][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:10:27,140][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:10:27,657][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:10:28,162][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:10:28,662][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:10:29,162][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:10:29,669][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:10:30,170][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:10:30,676][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:10:31,182][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:10:31,681][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:10:32,181][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:10:32,681][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:10:33,183][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:10:33,685][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:10:34,185][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:10:34,687][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:10:35,184][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:10:35,681][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:10:36,178][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:10:36,681][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:10:37,179][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:10:37,676][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:10:38,173][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:10:38,671][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:10:39,168][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:10:39,665][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:10:40,164][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:10:40,662][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:10:41,159][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:10:41,657][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:10:42,154][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:10:42,652][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:10:43,150][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:10:43,648][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:10:44,148][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:10:44,647][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:10:45,145][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:10:45,643][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:10:46,144][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:10:46,641][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:10:47,141][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:10:47,640][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:10:48,136][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:10:48,633][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:10:49,132][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:10:49,629][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:10:50,130][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10196 tokens. +[2026-03-26 04:10:50,904][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 60.66%, Block Peak % of device VRAM: 62.35%, ΔTime: 00:00:32 +[2026-03-26 04:10:51,631][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:10:51,633][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:10:51,635][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:10:52,468][__main__][INFO] - Iteration 458 took 51s (30.09% Gen, 68.28% Train). Generation: 15s, Training: 35s. Estimated remaining time: 35h 40m 20s. Estimated total time: 42h 45m 32s. Time estimates for 10 more iterations: 8m 33s, 100 more iterations: 1h 25m 31s, 500 more iterations: 7h 7m 35s. +[2026-03-26 04:10:52,470][__main__][INFO] - Starting iteration 458. +[2026-03-26 04:10:52,868][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 45 and human policies 1. +[2026-03-26 04:10:52,869][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:11:07,670][__main__][INFO] - Number of regex retries in iteration 458: 0 +[2026-03-26 04:11:07,671][__main__][INFO] - agents played in iteration 458 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:11:08,425][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:11:08,449][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:11:08,473][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:11:08,496][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:11:08,497][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:11:08,497][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:11:09,241][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:11:09,695][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:11:10,202][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:11:10,699][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:11:11,197][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:11:11,695][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:11:12,192][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:11:12,689][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:11:13,186][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:11:13,681][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:11:14,175][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:11:14,672][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:11:15,177][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:11:15,674][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:11:16,181][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:11:16,682][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:11:17,180][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:11:17,677][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:11:18,175][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:11:18,672][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:11:19,170][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:11:19,672][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:11:20,169][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:11:20,666][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:11:21,163][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:11:21,669][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:11:22,171][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:11:22,671][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:11:23,170][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:11:23,669][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:11:24,169][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:11:24,668][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:11:25,170][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:11:25,669][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:11:26,167][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:11:26,665][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:11:27,162][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:11:27,659][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:11:28,156][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:11:28,653][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:11:29,151][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:11:29,654][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:11:30,155][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:11:30,662][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:11:31,163][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:11:31,663][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:11:32,169][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:11:32,673][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:11:33,172][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:11:33,674][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:11:34,172][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:11:34,672][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:11:35,174][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:11:35,673][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:11:36,172][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:11:36,670][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:11:37,169][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:11:37,670][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:11:38,171][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:11:38,691][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:11:39,191][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:11:39,690][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:11:40,191][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:11:40,691][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:11:41,191][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10154 tokens. +[2026-03-26 04:11:41,961][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 60.65%, Block Peak % of device VRAM: 62.35%, ΔTime: 00:00:32 +[2026-03-26 04:11:42,699][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:11:42,701][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:11:42,702][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:11:43,438][__main__][INFO] - Iteration 459 took 50s (29.27% Gen, 69.27% Train). Generation: 14s, Training: 35s. Estimated remaining time: 35h 2m 28s. Estimated total time: 42h 8m 31s. Time estimates for 10 more iterations: 8m 25s, 100 more iterations: 1h 24m 17s, 500 more iterations: 7h 1m 25s. +[2026-03-26 04:11:43,441][__main__][INFO] - Starting iteration 459. +[2026-03-26 04:11:43,837][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 45 and human policies 1. +[2026-03-26 04:11:43,838][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:11:48,459][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 04:11:59,449][__main__][INFO] - Number of regex retries in iteration 459: 1 +[2026-03-26 04:11:59,450][__main__][INFO] - agents played in iteration 459 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:12:00,210][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:12:00,234][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:12:00,257][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:12:00,280][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:12:00,281][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:12:00,282][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:12:01,012][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:12:01,465][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:12:01,970][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:12:02,465][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:12:02,959][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:12:03,454][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:12:03,958][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:12:04,456][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:12:04,957][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:12:05,457][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:12:05,957][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:12:06,459][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:12:06,959][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:12:07,458][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:12:07,962][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:12:08,461][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:12:08,959][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:12:09,481][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:12:09,978][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:12:10,478][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:12:10,981][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:12:11,479][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:12:11,980][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:12:12,479][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:12:12,977][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:12:13,474][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:12:13,976][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:12:14,471][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:12:14,968][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:12:15,464][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:12:15,959][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:12:16,457][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:12:16,953][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:12:17,450][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:12:17,947][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:12:18,445][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:12:18,942][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:12:19,439][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:12:19,937][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:12:20,434][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:12:20,932][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:12:21,428][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:12:21,924][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:12:22,422][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:12:22,919][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:12:23,421][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:12:23,923][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:12:24,432][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:12:24,930][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:12:25,430][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:12:25,933][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:12:26,437][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:12:26,939][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:12:27,440][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:12:27,938][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:12:28,439][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:12:28,938][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:12:29,439][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:12:29,938][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:12:30,441][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:12:30,948][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:12:31,449][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:12:31,952][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:12:32,454][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:12:32,957][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10151 tokens. +[2026-03-26 04:12:33,741][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.28%, Current % of VRAM taken: 60.71%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 +[2026-03-26 04:12:34,489][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:12:34,491][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:12:34,493][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:12:35,214][__main__][INFO] - Iteration 460 took 51s (30.39% Gen, 68.21% Train). Generation: 15s, Training: 35s. Estimated remaining time: 35h 41m 56s. Estimated total time: 42h 48m 51s. Time estimates for 10 more iterations: 8m 33s, 100 more iterations: 1h 25m 37s, 500 more iterations: 7h 8m 8s. +[2026-03-26 04:12:35,216][__main__][INFO] - Starting iteration 460. +[2026-03-26 04:12:35,616][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 45 and human policies 1. +[2026-03-26 04:12:35,617][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:12:50,303][__main__][INFO] - Number of regex retries in iteration 460: 0 +[2026-03-26 04:12:50,304][__main__][INFO] - agents played in iteration 460 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:12:51,055][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:12:51,078][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:12:51,102][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:12:51,125][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:12:51,126][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:12:51,126][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:12:51,853][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:12:52,307][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:12:52,811][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:12:53,310][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:12:53,809][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:12:54,308][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:12:54,808][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:12:55,311][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:12:55,810][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:12:56,309][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:12:56,806][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:12:57,304][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:12:57,799][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:12:58,300][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:12:58,795][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:12:59,292][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:12:59,792][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:13:00,289][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:13:00,784][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:13:01,277][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:13:01,776][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:13:02,273][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:13:02,769][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:13:03,267][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:13:03,762][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:13:04,258][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:13:04,754][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:13:05,253][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:13:05,754][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:13:06,253][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:13:06,751][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:13:07,260][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:13:07,758][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:13:08,256][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:13:08,753][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:13:09,253][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:13:09,751][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:13:10,249][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:13:10,748][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:13:11,246][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:13:11,744][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:13:12,241][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:13:12,739][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:13:13,235][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:13:13,731][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:13:14,226][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:13:14,723][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:13:15,219][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:13:15,715][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:13:16,215][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:13:16,715][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:13:17,215][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:13:17,715][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:13:18,212][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:13:18,712][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:13:19,210][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:13:19,710][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:13:20,206][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:13:20,703][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:13:21,199][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:13:21,697][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:13:22,194][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:13:22,692][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:13:23,189][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:13:23,688][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10070 tokens. +[2026-03-26 04:13:24,458][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 60.50%, Block Peak % of device VRAM: 62.02%, ΔTime: 00:00:32 +[2026-03-26 04:13:25,203][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:13:25,205][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:13:25,207][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:13:26,634][__main__][INFO] - Iteration 461 took 51s (28.79% Gen, 68.41% Train). Generation: 14s, Training: 34s. Estimated remaining time: 35h 23m 9s. Estimated total time: 42h 30m 56s. Time estimates for 10 more iterations: 8m 30s, 100 more iterations: 1h 25m 1s, 500 more iterations: 7h 5m 9s. +[2026-03-26 04:13:26,637][__main__][INFO] - Starting iteration 461. +[2026-03-26 04:13:27,034][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 46 and human policies 1. +[2026-03-26 04:13:27,035][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:13:42,810][__main__][INFO] - Number of regex retries in iteration 461: 0 +[2026-03-26 04:13:42,811][__main__][INFO] - agents played in iteration 461 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:13:43,563][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:13:43,587][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:13:43,610][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:13:43,633][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:13:43,634][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:13:43,635][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:13:44,375][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:13:44,829][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:13:45,333][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:13:45,830][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:13:46,330][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:13:46,827][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:13:47,342][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:13:47,844][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:13:48,349][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:13:48,851][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:13:49,357][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:13:49,859][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:13:50,382][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:13:50,900][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:13:51,404][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:13:51,911][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:13:52,416][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:13:52,916][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:13:53,417][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:13:53,917][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:13:54,421][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:13:54,932][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:13:55,438][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:13:55,938][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:13:56,443][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:13:56,941][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:13:57,439][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:13:57,941][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:13:58,442][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:13:58,941][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:13:59,442][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:13:59,948][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:14:00,451][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:14:00,952][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:14:01,452][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:14:01,954][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:14:02,455][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:14:02,957][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:14:03,456][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:14:03,955][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:14:04,455][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:14:04,968][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:14:05,475][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:14:05,975][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:14:06,476][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:14:06,977][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:14:07,476][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:14:07,976][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:14:08,477][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:14:08,980][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:14:09,482][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:14:09,983][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:14:10,480][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:14:10,980][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:14:11,481][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:14:11,983][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:14:12,486][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:14:12,983][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:14:13,483][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:14:13,981][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:14:14,482][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:14:14,986][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:14:15,488][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:14:15,991][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:14:16,493][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10258 tokens. +[2026-03-26 04:14:17,274][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:32 +[2026-03-26 04:14:18,040][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:14:18,043][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:14:18,044][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:14:18,751][__main__][INFO] - Iteration 462 took 51s (30.50% Gen, 68.13% Train). Generation: 15s, Training: 35s. Estimated remaining time: 35h 57m 13s. Estimated total time: 43h 5m 51s. Time estimates for 10 more iterations: 8m 37s, 100 more iterations: 1h 26m 11s, 500 more iterations: 7h 10m 58s. +[2026-03-26 04:14:18,753][__main__][INFO] - Starting iteration 462. +[2026-03-26 04:14:19,151][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 46 and human policies 1. +[2026-03-26 04:14:19,152][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:14:23,723][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 04:14:35,287][__main__][INFO] - Number of regex retries in iteration 462: 1 +[2026-03-26 04:14:35,288][__main__][INFO] - agents played in iteration 462 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:14:36,042][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:14:36,065][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:14:36,089][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:14:36,112][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:14:36,113][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:14:36,113][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:14:36,840][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:14:37,292][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:14:37,797][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:14:38,295][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:14:38,791][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:14:39,289][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:14:39,786][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:14:40,282][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:14:40,778][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:14:41,275][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:14:41,778][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:14:42,280][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:14:42,792][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:14:43,296][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:14:43,798][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:14:44,300][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:14:44,802][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:14:45,301][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:14:45,800][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:14:46,297][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:14:46,796][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:14:47,293][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:14:47,793][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:14:48,290][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:14:48,788][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:14:49,286][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:14:49,787][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:14:50,288][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:14:50,785][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:14:51,286][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:14:51,787][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:14:52,295][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:14:52,798][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:14:53,300][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:14:53,798][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:14:54,296][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:14:54,793][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:14:55,296][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:14:55,794][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:14:56,290][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:14:56,791][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:14:57,289][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:14:57,787][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:14:58,285][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:14:58,782][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:14:59,280][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:14:59,777][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:15:00,274][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:15:00,773][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:15:01,274][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:15:01,775][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:15:02,277][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:15:02,776][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:15:03,275][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:15:03,801][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:15:04,299][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:15:04,799][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:15:05,299][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:15:05,796][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:15:06,295][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:15:06,796][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:15:07,293][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:15:07,791][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:15:08,291][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:15:08,791][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10203 tokens. +[2026-03-26 04:15:09,561][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 60.67%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 +[2026-03-26 04:15:10,327][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:15:10,330][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:15:10,331][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:15:11,043][__main__][INFO] - Iteration 463 took 51s (31.09% Gen, 67.53% Train). Generation: 16s, Training: 35s. Estimated remaining time: 36h 5m 6s. Estimated total time: 43h 14m 36s. Time estimates for 10 more iterations: 8m 38s, 100 more iterations: 1h 26m 29s, 500 more iterations: 7h 12m 26s. +[2026-03-26 04:15:11,045][__main__][INFO] - Starting iteration 463. +[2026-03-26 04:15:11,443][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 46 and human policies 1. +[2026-03-26 04:15:11,443][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:15:27,991][__main__][INFO] - Number of regex retries in iteration 463: 0 +[2026-03-26 04:15:27,992][__main__][INFO] - agents played in iteration 463 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:15:28,746][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:15:28,770][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:15:28,793][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:15:28,816][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:15:28,817][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:15:28,818][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:15:29,548][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:15:30,005][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:15:30,510][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:15:31,010][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:15:31,509][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:15:32,010][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:15:32,510][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:15:33,011][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:15:33,512][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:15:34,010][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:15:34,508][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:15:35,005][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:15:35,500][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:15:35,997][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:15:36,496][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:15:36,999][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:15:37,498][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:15:38,001][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:15:38,503][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:15:39,001][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:15:39,498][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:15:40,000][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:15:40,498][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:15:41,000][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:15:41,498][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:15:41,999][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:15:42,496][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:15:42,998][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:15:43,500][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:15:43,996][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:15:44,497][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:15:44,995][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:15:45,493][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:15:45,991][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:15:46,490][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:15:46,992][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:15:47,490][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:15:47,989][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:15:48,491][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:15:48,994][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:15:49,503][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:15:50,008][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:15:50,513][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:15:51,016][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:15:51,516][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:15:52,015][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:15:52,519][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:15:53,022][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:15:53,524][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:15:54,026][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:15:54,529][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:15:55,031][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:15:55,532][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:15:56,030][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:15:56,529][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:15:57,027][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:15:57,527][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:15:58,025][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:15:58,529][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:15:59,029][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:15:59,528][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:16:00,030][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:16:00,526][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:16:01,026][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:16:01,523][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10308 tokens. +[2026-03-26 04:16:02,299][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.58%, Block Peak % of device VRAM: 62.21%, ΔTime: 00:00:32 +[2026-03-26 04:16:03,047][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:16:03,049][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:16:03,051][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:16:03,774][__main__][INFO] - Iteration 464 took 52s (31.62% Gen, 66.99% Train). Generation: 16s, Training: 35s. Estimated remaining time: 36h 26m 11s. Estimated total time: 43h 36m 34s. Time estimates for 10 more iterations: 8m 43s, 100 more iterations: 1h 27m 13s, 500 more iterations: 7h 16m 5s. +[2026-03-26 04:16:03,776][__main__][INFO] - Starting iteration 464. +[2026-03-26 04:16:04,173][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 46 and human policies 1. +[2026-03-26 04:16:04,174][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:16:10,224][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 04:16:17,054][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 ballots did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 04:16:20,250][__main__][INFO] - Number of regex retries in iteration 464: 2 +[2026-03-26 04:16:20,250][__main__][INFO] - agents played in iteration 464 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:16:21,004][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:16:21,028][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:16:21,052][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:16:21,075][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:16:21,075][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:16:21,076][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:16:21,804][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:16:22,262][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:16:22,766][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:16:23,268][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:16:23,772][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:16:24,273][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:16:24,778][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:16:25,279][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:16:25,788][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:16:26,291][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:16:26,793][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:16:27,292][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:16:27,793][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:16:28,295][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:16:28,794][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:16:29,294][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:16:29,792][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:16:30,295][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:16:30,793][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:16:31,293][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:16:31,793][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:16:32,293][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:16:32,791][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:16:33,288][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:16:33,786][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:16:34,284][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:16:34,786][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:16:35,284][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:16:35,780][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:16:36,281][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:16:36,783][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:16:37,279][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:16:37,778][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:16:38,278][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:16:38,777][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:16:39,275][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:16:39,777][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:16:40,281][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:16:40,782][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:16:41,291][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:16:41,793][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:16:42,298][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:16:42,799][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:16:43,300][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:16:43,799][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:16:44,300][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:16:44,799][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:16:45,301][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:16:45,806][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:16:46,305][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:16:46,807][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:16:47,309][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:16:47,811][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:16:48,310][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:16:48,812][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:16:49,314][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:16:49,815][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:16:50,315][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:16:50,815][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:16:51,315][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:16:51,815][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:16:52,315][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:16:52,815][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:16:53,315][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:16:53,815][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10233 tokens. +[2026-03-26 04:16:54,590][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.25%, Current % of VRAM taken: 60.67%, Block Peak % of device VRAM: 62.45%, ΔTime: 00:00:32 +[2026-03-26 04:16:55,340][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:16:55,343][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:16:55,344][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:16:56,066][__main__][INFO] - Iteration 465 took 51s (30.98% Gen, 67.63% Train). Generation: 16s, Training: 35s. Estimated remaining time: 36h 3m 24s. Estimated total time: 43h 14m 40s. Time estimates for 10 more iterations: 8m 38s, 100 more iterations: 1h 26m 29s, 500 more iterations: 7h 12m 26s. +[2026-03-26 04:16:56,068][__main__][INFO] - Starting iteration 465. +[2026-03-26 04:16:56,467][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 46 and human policies 1. +[2026-03-26 04:16:56,468][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:17:00,595][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 04:17:11,927][__main__][INFO] - Number of regex retries in iteration 465: 1 +[2026-03-26 04:17:11,928][__main__][INFO] - agents played in iteration 465 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:17:12,679][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:17:12,703][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:17:12,727][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:17:12,750][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:17:12,751][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:17:12,751][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:17:13,477][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:17:13,930][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:17:14,436][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:17:14,933][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:17:15,430][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:17:15,926][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:17:16,425][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:17:16,927][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:17:17,429][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:17:17,927][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:17:18,425][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:17:18,923][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:17:19,420][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:17:19,924][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:17:20,428][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:17:20,926][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:17:21,424][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:17:21,925][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:17:22,421][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:17:22,920][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:17:23,419][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:17:23,918][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:17:24,418][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:17:24,914][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:17:25,414][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:17:25,911][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:17:26,409][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:17:26,906][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:17:27,404][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:17:27,902][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:17:28,402][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:17:28,897][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:17:29,393][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:17:29,896][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:17:30,395][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:17:30,893][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:17:31,403][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:17:31,904][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:17:32,402][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:17:32,907][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:17:33,404][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:17:33,905][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:17:34,406][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:17:34,909][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:17:35,410][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:17:35,911][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:17:36,413][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:17:36,915][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:17:37,418][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:17:37,916][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:17:38,417][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:17:38,914][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:17:39,411][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:17:39,908][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:17:40,406][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:17:40,903][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:17:41,399][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:17:41,895][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:17:42,393][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:17:42,890][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:17:43,389][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:17:43,886][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:17:44,382][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:17:44,878][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:17:45,375][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10173 tokens. +[2026-03-26 04:17:46,155][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.15%, ΔTime: 00:00:32 +[2026-03-26 04:17:46,905][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:17:46,908][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:17:46,909][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:17:47,676][__main__][INFO] - Iteration 466 took 51s (30.19% Gen, 68.31% Train). Generation: 15s, Training: 34s. Estimated remaining time: 35h 28m 23s. Estimated total time: 42h 40m 30s. Time estimates for 10 more iterations: 8m 32s, 100 more iterations: 1h 25m 21s, 500 more iterations: 7h 6m 45s. +[2026-03-26 04:17:47,679][__main__][INFO] - Starting iteration 466. +[2026-03-26 04:17:48,079][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 46 and human policies 1. +[2026-03-26 04:17:48,080][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:18:01,850][__main__][INFO] - Number of regex retries in iteration 466: 0 +[2026-03-26 04:18:01,851][__main__][INFO] - agents played in iteration 466 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:18:02,604][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:18:02,627][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:18:02,651][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:18:02,674][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:18:02,675][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:18:02,675][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:18:03,401][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:18:03,853][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:18:04,357][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:18:04,855][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:18:05,357][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:18:05,856][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:18:06,357][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:18:06,859][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:18:07,359][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:18:07,856][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:18:08,353][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:18:08,850][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:18:09,346][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:18:09,842][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:18:10,338][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:18:10,834][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:18:11,331][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:18:11,830][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:18:12,329][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:18:12,827][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:18:13,324][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:18:13,824][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:18:14,320][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:18:14,822][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:18:15,320][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:18:15,819][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:18:16,317][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:18:16,815][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:18:17,313][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:18:17,812][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:18:18,310][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:18:18,809][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:18:19,308][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:18:19,805][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:18:20,301][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:18:20,796][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:18:21,292][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:18:21,789][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:18:22,285][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:18:22,782][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:18:23,277][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:18:23,774][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:18:24,272][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:18:24,769][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:18:25,267][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:18:25,765][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:18:26,262][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:18:26,765][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:18:27,264][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:18:27,766][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:18:28,280][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:18:28,781][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:18:29,281][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:18:29,780][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:18:30,282][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:18:30,780][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:18:31,280][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:18:31,783][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:18:32,283][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:18:32,782][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:18:33,280][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:18:33,779][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:18:34,277][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:18:34,777][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:18:35,276][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10019 tokens. +[2026-03-26 04:18:36,065][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 60.54%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:32 +[2026-03-26 04:18:36,741][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:18:36,744][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:18:36,747][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:18:37,620][__main__][INFO] - Iteration 467 took 49s (27.80% Gen, 70.44% Train). Generation: 13s, Training: 34s. Estimated remaining time: 34h 4m 8s. Estimated total time: 41h 17m 5s. Time estimates for 10 more iterations: 8m 15s, 100 more iterations: 1h 22m 34s, 500 more iterations: 6h 52m 50s. +[2026-03-26 04:18:37,623][__main__][INFO] - Starting iteration 467. +[2026-03-26 04:18:38,022][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 46 and human policies 1. +[2026-03-26 04:18:38,023][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:18:51,441][mllm.models.large_language_model_local][WARNING] - Response Propsoal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 04:18:54,723][__main__][INFO] - Number of regex retries in iteration 467: 1 +[2026-03-26 04:18:54,724][__main__][INFO] - agents played in iteration 467 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:18:55,476][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:18:55,500][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:18:55,523][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:18:55,547][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:18:55,547][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:18:55,548][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:18:56,283][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:18:56,735][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:18:57,238][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:18:57,736][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:18:58,236][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:18:58,735][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:18:59,237][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:18:59,733][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:19:00,230][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:19:00,731][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:19:01,228][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:19:01,726][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:19:02,228][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:19:02,725][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:19:03,223][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:19:03,720][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:19:04,216][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:19:04,714][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:19:05,235][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:19:05,732][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:19:06,230][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:19:06,728][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:19:07,229][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:19:07,724][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:19:08,221][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:19:08,719][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:19:09,216][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:19:09,713][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:19:10,211][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:19:10,709][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:19:11,205][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:19:11,707][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:19:12,203][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:19:12,699][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:19:13,199][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:19:13,702][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:19:14,199][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:19:14,697][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:19:15,195][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:19:15,695][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:19:16,192][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:19:16,690][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:19:17,191][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:19:17,691][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:19:18,191][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:19:18,695][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:19:19,198][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:19:19,699][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:19:20,197][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:19:20,705][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:19:21,204][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:19:21,702][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:19:22,201][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:19:22,700][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:19:23,201][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:19:23,699][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:19:24,198][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:19:24,697][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:19:25,196][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:19:25,693][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:19:26,189][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:19:26,686][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:19:27,185][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:19:27,684][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:19:28,182][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10124 tokens. +[2026-03-26 04:19:28,968][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 60.54%, Block Peak % of device VRAM: 62.11%, ΔTime: 00:00:32 +[2026-03-26 04:19:29,705][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:19:29,707][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:19:29,709][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:19:30,557][__main__][INFO] - Iteration 468 took 52s (31.79% Gen, 66.59% Train). Generation: 16s, Training: 34s. Estimated remaining time: 36h 32m 56s. Estimated total time: 43h 46m 47s. Time estimates for 10 more iterations: 8m 45s, 100 more iterations: 1h 27m 33s, 500 more iterations: 7h 17m 47s. +[2026-03-26 04:19:30,559][__main__][INFO] - Starting iteration 468. +[2026-03-26 04:19:30,959][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 46 and human policies 1. +[2026-03-26 04:19:30,960][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:19:46,717][__main__][INFO] - Number of regex retries in iteration 468: 0 +[2026-03-26 04:19:46,718][__main__][INFO] - agents played in iteration 468 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:19:47,479][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:19:47,503][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:19:47,526][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:19:47,549][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:19:47,550][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:19:47,550][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:19:48,285][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:19:48,741][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:19:49,250][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:19:49,751][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:19:50,251][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:19:50,750][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:19:51,264][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:19:51,774][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:19:52,278][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:19:52,778][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:19:53,278][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:19:53,778][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:19:54,277][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:19:54,779][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:19:55,278][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:19:55,782][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:19:56,281][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:19:56,783][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:19:57,280][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:19:57,778][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:19:58,293][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:19:58,800][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:19:59,303][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:19:59,801][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:20:00,303][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:20:00,802][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:20:01,300][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:20:01,805][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:20:02,304][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:20:02,820][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:20:03,324][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:20:03,824][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:20:04,326][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:20:04,823][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:20:05,320][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:20:05,817][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:20:06,315][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:20:06,813][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:20:07,312][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:20:07,811][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:20:08,312][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:20:08,819][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:20:09,320][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:20:09,819][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:20:10,318][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:20:10,818][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:20:11,322][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:20:11,820][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:20:12,319][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:20:12,820][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:20:13,317][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:20:13,818][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:20:14,318][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:20:14,825][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:20:15,332][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:20:15,839][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:20:16,342][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:20:16,840][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:20:17,341][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:20:17,840][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:20:18,339][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:20:18,839][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:20:19,339][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:20:19,838][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:20:20,339][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10099 tokens. +[2026-03-26 04:20:21,116][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 60.66%, Block Peak % of device VRAM: 62.38%, ΔTime: 00:00:32 +[2026-03-26 04:20:21,868][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:20:21,871][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:20:21,872][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:20:22,593][__main__][INFO] - Iteration 469 took 51s (30.52% Gen, 68.08% Train). Generation: 15s, Training: 35s. Estimated remaining time: 35h 47m 2s. Estimated total time: 43h 1m 44s. Time estimates for 10 more iterations: 8m 36s, 100 more iterations: 1h 26m 3s, 500 more iterations: 7h 10m 17s. +[2026-03-26 04:20:22,596][__main__][INFO] - Starting iteration 469. +[2026-03-26 04:20:22,994][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 46 and human policies 1. +[2026-03-26 04:20:22,994][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:20:37,606][__main__][INFO] - Number of regex retries in iteration 469: 0 +[2026-03-26 04:20:37,607][__main__][INFO] - agents played in iteration 469 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:20:38,365][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:20:38,389][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:20:38,412][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:20:38,436][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:20:38,436][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:20:38,437][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:20:39,181][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:20:39,634][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:20:40,142][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:20:40,640][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:20:41,137][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:20:41,635][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:20:42,150][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:20:42,653][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:20:43,157][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:20:43,657][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:20:44,158][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:20:44,662][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:20:45,165][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:20:45,667][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:20:46,175][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:20:46,677][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:20:47,180][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:20:47,678][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:20:48,177][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:20:48,674][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:20:49,171][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:20:49,670][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:20:50,168][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:20:50,666][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:20:51,164][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:20:51,663][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:20:52,159][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:20:52,654][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:20:53,152][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:20:53,655][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:20:54,161][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:20:54,662][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:20:55,162][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:20:55,662][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:20:56,162][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:20:56,662][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:20:57,166][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:20:57,664][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:20:58,164][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:20:58,665][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:20:59,167][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:20:59,671][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:21:00,173][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:21:00,675][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:21:01,175][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:21:01,674][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:21:02,179][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:21:02,677][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:21:03,176][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:21:03,677][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:21:04,178][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:21:04,684][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:21:05,196][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:21:05,703][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:21:06,207][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:21:06,711][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:21:07,215][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:21:07,715][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:21:08,214][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:21:08,713][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:21:09,212][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:21:09,711][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:21:10,209][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:21:10,705][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:21:11,202][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10033 tokens. +[2026-03-26 04:21:11,982][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.01%, Current % of VRAM taken: 60.43%, Block Peak % of device VRAM: 62.11%, ΔTime: 00:00:32 +[2026-03-26 04:21:12,694][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:21:12,697][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:21:12,698][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:21:13,455][__main__][INFO] - Iteration 470 took 50s (28.96% Gen, 69.54% Train). Generation: 14s, Training: 35s. Estimated remaining time: 34h 47m 32s. Estimated total time: 42h 3m 5s. Time estimates for 10 more iterations: 8m 24s, 100 more iterations: 1h 24m 6s, 500 more iterations: 7h 0m 30s. +[2026-03-26 04:21:13,457][__main__][INFO] - Starting iteration 470. +[2026-03-26 04:21:13,856][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 46 and human policies 1. +[2026-03-26 04:21:13,857][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:21:18,767][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 04:21:21,221][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 04:21:29,148][__main__][INFO] - Number of regex retries in iteration 470: 2 +[2026-03-26 04:21:29,149][__main__][INFO] - agents played in iteration 470 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:21:29,907][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.45%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:21:29,931][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.45%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:21:29,954][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.45%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:21:29,978][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.45%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:21:29,978][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:21:29,979][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:21:30,718][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:21:31,173][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:21:31,676][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:21:32,174][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:21:32,671][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:21:33,178][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:21:33,687][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:21:34,191][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:21:34,695][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:21:35,194][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:21:35,698][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:21:36,201][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:21:36,701][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:21:37,202][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:21:37,704][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:21:38,203][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:21:38,704][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:21:39,202][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:21:39,700][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:21:40,199][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:21:40,701][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:21:41,204][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:21:41,707][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:21:42,217][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:21:42,718][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:21:43,225][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:21:43,724][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:21:44,228][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:21:44,726][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:21:45,229][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:21:45,729][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:21:46,229][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:21:46,732][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:21:47,237][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:21:47,738][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:21:48,238][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:21:48,740][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:21:49,239][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:21:49,742][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:21:50,241][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:21:50,740][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:21:51,238][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:21:51,739][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:21:52,238][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:21:52,737][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:21:53,236][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:21:53,733][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:21:54,232][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:21:54,729][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:21:55,229][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:21:55,725][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:21:56,223][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:21:56,720][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:21:57,219][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:21:57,718][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:21:58,216][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:21:58,717][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:21:59,216][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:21:59,715][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:22:00,213][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:22:00,715][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:22:01,214][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:22:01,715][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:22:02,214][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:22:02,712][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10242 tokens. +[2026-03-26 04:22:03,490][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 60.56%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 +[2026-03-26 04:22:17,617][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:22:17,619][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:22:17,621][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:22:19,005][__main__][INFO] - Iteration 471 took 1m 5s (23.47% Gen, 74.40% Train). Generation: 15s, Training: 48s. Estimated remaining time: 47h 0m 47s. Estimated total time: 54h 17m 26s. Time estimates for 10 more iterations: 10m 51s, 100 more iterations: 1h 48m 34s, 500 more iterations: 9h 2m 54s. +[2026-03-26 04:22:19,007][__main__][INFO] - Starting iteration 471. +[2026-03-26 04:22:19,408][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 47 and human policies 1. +[2026-03-26 04:22:19,409][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:22:35,530][__main__][INFO] - Number of regex retries in iteration 471: 0 +[2026-03-26 04:22:35,531][__main__][INFO] - agents played in iteration 471 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:22:36,297][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:22:36,320][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:22:36,344][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:22:36,368][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:22:36,368][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:22:36,369][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:22:37,110][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:22:37,562][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:22:38,070][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:22:38,583][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:22:39,088][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:22:39,587][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:22:40,084][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:22:40,583][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:22:41,080][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:22:41,579][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:22:42,080][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:22:42,585][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:22:43,087][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:22:43,585][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:22:44,084][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:22:44,582][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:22:45,094][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:22:45,597][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:22:46,098][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:22:46,596][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:22:47,093][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:22:47,590][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:22:48,099][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:22:48,601][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:22:49,100][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:22:49,602][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:22:50,103][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:22:50,604][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:22:51,106][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:22:51,612][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:22:52,121][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:22:52,645][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:22:53,152][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:22:53,652][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:22:54,152][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:22:54,651][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:22:55,154][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:22:55,651][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:22:56,163][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:22:56,667][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:22:57,166][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:22:57,671][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:22:58,172][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:22:58,672][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:22:59,174][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:22:59,673][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:23:00,176][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:23:00,673][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:23:01,178][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:23:01,686][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:23:02,190][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:23:02,690][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:23:03,189][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:23:03,688][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:23:04,186][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:23:04,684][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:23:05,182][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:23:05,679][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:23:06,178][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:23:06,676][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:23:07,175][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:23:07,676][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:23:08,177][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:23:08,675][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:23:09,172][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10123 tokens. +[2026-03-26 04:23:09,948][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.05%, Current % of VRAM taken: 60.47%, Block Peak % of device VRAM: 62.11%, ΔTime: 00:00:32 +[2026-03-26 04:23:30,204][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:23:30,206][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:23:30,208][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:23:31,016][__main__][INFO] - Iteration 472 took 1m 11s (22.51% Gen, 76.36% Train). Generation: 16s, Training: 54s. Estimated remaining time: 52h 22m 34s. Estimated total time: 59h 40m 25s. Time estimates for 10 more iterations: 11m 56s, 100 more iterations: 1h 59m 20s, 500 more iterations: 9h 56m 44s. +[2026-03-26 04:23:31,018][__main__][INFO] - Starting iteration 472. +[2026-03-26 04:23:31,415][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 47 and human policies 1. +[2026-03-26 04:23:31,415][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:23:35,928][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 04:23:37,024][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 04:23:47,245][__main__][INFO] - Number of regex retries in iteration 472: 2 +[2026-03-26 04:23:47,246][__main__][INFO] - agents played in iteration 472 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:23:48,039][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:23:48,063][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:23:48,086][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:23:48,110][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:23:48,110][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:23:48,111][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:23:48,851][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:23:49,304][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:23:49,804][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:23:50,303][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:23:50,800][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:23:51,299][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:23:51,804][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:23:52,306][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:23:52,805][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:23:53,305][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:23:53,804][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:23:54,302][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:23:54,800][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:23:55,302][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:23:55,800][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:23:56,298][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:23:56,796][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:23:57,294][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:23:57,792][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:23:58,291][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:23:58,791][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:23:59,294][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:23:59,794][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:24:00,300][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:24:00,804][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:24:01,303][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:24:01,810][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:24:02,310][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:24:02,810][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:24:03,309][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:24:03,807][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:24:04,305][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:24:04,807][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:24:05,309][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:24:05,811][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:24:06,335][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:24:06,838][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:24:07,339][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:24:07,839][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:24:08,339][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:24:08,838][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:24:09,339][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:24:09,837][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:24:10,339][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:24:10,840][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:24:11,336][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:24:11,834][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:24:12,332][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:24:12,830][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:24:13,327][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:24:13,824][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:24:14,321][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:24:14,818][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:24:15,316][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:24:15,814][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:24:16,312][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:24:16,809][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:24:17,305][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:24:17,801][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:24:18,296][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:24:18,793][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:24:19,290][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:24:19,785][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:24:20,280][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:24:20,778][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10139 tokens. +[2026-03-26 04:24:21,555][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 60.61%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:32 +[2026-03-26 04:24:22,259][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:24:22,261][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:24:22,263][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:24:22,993][__main__][INFO] - Iteration 473 took 51s (30.69% Gen, 67.89% Train). Generation: 15s, Training: 35s. Estimated remaining time: 35h 40m 15s. Estimated total time: 42h 58m 57s. Time estimates for 10 more iterations: 8m 35s, 100 more iterations: 1h 25m 57s, 500 more iterations: 7h 9m 49s. +[2026-03-26 04:24:22,996][__main__][INFO] - Starting iteration 473. +[2026-03-26 04:24:23,393][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 47 and human policies 1. +[2026-03-26 04:24:23,394][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:24:27,590][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 04:24:37,879][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 1 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 04:24:38,863][__main__][INFO] - Number of regex retries in iteration 473: 2 +[2026-03-26 04:24:38,864][__main__][INFO] - agents played in iteration 473 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:24:39,613][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:24:39,637][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:24:39,661][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:24:39,684][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:24:39,685][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:24:39,685][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:24:40,433][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:24:40,888][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:24:41,406][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:24:41,913][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:24:42,418][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:24:42,921][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:24:43,427][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:24:43,931][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:24:44,448][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:24:44,948][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:24:45,452][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:24:45,951][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:24:46,455][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:24:46,954][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:24:47,455][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:24:47,954][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:24:48,453][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:24:48,955][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:24:49,453][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:24:49,953][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:24:50,453][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:24:50,952][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:24:51,453][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:24:51,952][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:24:52,454][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:24:52,952][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:24:53,453][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:24:53,951][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:24:54,452][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:24:54,954][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:24:55,453][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:24:55,952][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:24:56,449][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:24:56,948][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:24:57,445][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:24:57,946][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:24:58,442][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:24:58,939][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:24:59,434][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:24:59,932][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:25:00,430][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:25:00,927][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:25:01,422][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:25:01,920][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:25:02,418][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:25:02,918][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:25:03,418][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:25:03,914][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:25:04,412][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:25:04,909][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:25:05,405][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:25:05,904][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:25:06,400][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:25:06,900][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:25:07,397][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:25:07,896][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:25:08,394][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:25:08,891][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:25:09,388][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:25:09,887][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:25:10,383][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:25:10,879][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:25:11,375][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:25:11,876][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:25:12,375][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10166 tokens. +[2026-03-26 04:25:13,152][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.21%, ΔTime: 00:00:32 +[2026-03-26 04:25:13,912][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:25:13,914][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:25:13,916][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:25:14,640][__main__][INFO] - Iteration 474 took 51s (30.19% Gen, 68.40% Train). Generation: 15s, Training: 35s. Estimated remaining time: 35h 22m 49s. Estimated total time: 42h 42m 23s. Time estimates for 10 more iterations: 8m 32s, 100 more iterations: 1h 25m 24s, 500 more iterations: 7h 7m 3s. +[2026-03-26 04:25:14,642][__main__][INFO] - Starting iteration 474. +[2026-03-26 04:25:15,040][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 47 and human policies 1. +[2026-03-26 04:25:15,041][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:25:31,519][__main__][INFO] - Number of regex retries in iteration 474: 0 +[2026-03-26 04:25:31,520][__main__][INFO] - agents played in iteration 474 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:25:32,284][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:25:32,309][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:25:32,333][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:25:32,357][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:25:32,358][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:25:32,359][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:25:33,120][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:25:33,585][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:25:34,096][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:25:34,605][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:25:35,108][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:25:35,613][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:25:36,120][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:25:36,635][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:25:37,140][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:25:37,643][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:25:38,143][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:25:38,643][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:25:39,143][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:25:39,647][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:25:40,155][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:25:40,658][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:25:41,162][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:25:41,662][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:25:42,162][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:25:42,660][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:25:43,160][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:25:43,660][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:25:44,174][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:25:44,679][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:25:45,182][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:25:45,687][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:25:46,190][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:25:46,693][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:25:47,194][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:25:47,696][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:25:48,197][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:25:48,698][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:25:49,199][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:25:49,699][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:25:50,197][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:25:50,695][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:25:51,194][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:25:51,696][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:25:52,195][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:25:52,700][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:25:53,208][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:25:53,713][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:25:54,214][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:25:54,716][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:25:55,220][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:25:55,722][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:25:56,224][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:25:56,722][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:25:57,224][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:25:57,723][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:25:58,221][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:25:58,724][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:25:59,233][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:25:59,736][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:26:00,235][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:26:00,737][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:26:01,238][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:26:01,742][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:26:02,247][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:26:02,749][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:26:03,251][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:26:03,755][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:26:04,258][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:26:04,759][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:26:05,265][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10313 tokens. +[2026-03-26 04:26:06,073][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 60.64%, Block Peak % of device VRAM: 62.28%, ΔTime: 00:00:32 +[2026-03-26 04:26:15,212][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:26:15,215][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:26:15,217][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:26:15,946][__main__][INFO] - Iteration 475 took 1m 0s (27.06% Gen, 71.74% Train). Generation: 16s, Training: 43s. Estimated remaining time: 43h 24m 43s. Estimated total time: 50h 45m 19s. Time estimates for 10 more iterations: 10m 9s, 100 more iterations: 1h 41m 30s, 500 more iterations: 8h 27m 33s. +[2026-03-26 04:26:15,948][__main__][INFO] - Starting iteration 475. +[2026-03-26 04:26:16,346][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 47 and human policies 1. +[2026-03-26 04:26:16,347][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:26:32,451][__main__][INFO] - Number of regex retries in iteration 475: 0 +[2026-03-26 04:26:32,452][__main__][INFO] - agents played in iteration 475 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:26:33,208][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:26:33,232][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:26:33,255][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:26:33,279][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:26:33,279][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:26:33,280][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:26:34,027][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:26:34,485][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:26:34,990][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:26:35,495][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:26:36,006][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:26:36,506][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:26:37,011][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:26:37,512][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:26:38,014][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:26:38,514][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:26:39,014][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:26:39,516][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:26:40,016][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:26:40,514][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:26:41,017][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:26:41,517][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:26:42,015][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:26:42,513][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:26:43,012][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:26:43,512][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:26:44,019][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:26:44,518][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:26:45,026][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:26:45,532][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:26:46,034][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:26:46,534][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:26:47,036][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:26:47,536][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:26:48,035][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:26:48,535][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:26:49,035][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:26:49,534][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:26:50,035][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:26:50,537][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:26:51,038][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:26:51,537][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:26:52,037][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:26:52,536][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:26:53,034][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:26:53,532][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:26:54,032][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:26:54,530][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:26:55,027][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:26:55,525][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:26:56,023][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:26:56,520][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:26:57,019][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:26:57,521][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:26:58,021][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:26:58,518][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:26:59,015][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:26:59,514][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:27:00,012][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:27:00,512][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:27:01,010][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:27:01,508][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:27:02,008][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:27:02,504][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:27:03,000][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:27:03,497][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:27:03,995][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:27:04,494][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:27:04,991][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:27:05,486][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:27:05,981][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10120 tokens. +[2026-03-26 04:27:06,769][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 60.49%, Block Peak % of device VRAM: 62.15%, ΔTime: 00:00:32 +[2026-03-26 04:27:07,532][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:27:07,534][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:27:07,536][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:27:08,303][__main__][INFO] - Iteration 476 took 51s (31.00% Gen, 67.52% Train). Generation: 16s, Training: 35s. Estimated remaining time: 35h 56m 26s. Estimated total time: 43h 17m 54s. Time estimates for 10 more iterations: 8m 39s, 100 more iterations: 1h 26m 35s, 500 more iterations: 7h 12m 59s. +[2026-03-26 04:27:08,305][__main__][INFO] - Starting iteration 476. +[2026-03-26 04:27:08,703][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 47 and human policies 1. +[2026-03-26 04:27:08,704][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:27:24,655][__main__][INFO] - Number of regex retries in iteration 476: 0 +[2026-03-26 04:27:24,655][__main__][INFO] - agents played in iteration 476 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:27:25,408][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:27:25,431][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:27:25,455][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:27:25,478][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:27:25,479][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:27:25,479][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:27:26,229][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:27:26,683][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:27:27,185][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:27:27,681][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:27:28,178][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:27:28,676][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:27:29,174][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:27:29,675][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:27:30,184][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:27:30,688][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:27:31,193][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:27:31,694][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:27:32,196][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:27:32,701][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:27:33,199][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:27:33,698][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:27:34,201][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:27:34,703][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:27:35,203][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:27:35,723][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:27:36,226][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:27:36,727][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:27:37,226][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:27:37,729][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:27:38,229][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:27:38,727][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:27:39,226][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:27:39,723][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:27:40,222][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:27:40,721][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:27:41,220][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:27:41,718][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:27:42,217][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:27:42,716][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:27:43,214][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:27:43,713][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:27:44,212][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:27:44,720][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:27:45,219][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:27:45,727][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:27:46,233][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:27:46,740][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:27:47,241][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:27:47,746][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:27:48,248][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:27:48,749][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:27:49,252][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:27:49,752][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:27:50,253][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:27:50,753][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:27:51,249][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:27:51,749][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:27:52,251][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:27:52,747][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:27:53,243][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:27:53,739][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:27:54,237][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:27:54,735][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:27:55,233][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:27:55,730][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:27:56,226][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:27:56,722][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:27:57,218][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:27:57,715][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:27:58,212][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10110 tokens. +[2026-03-26 04:27:58,997][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 60.51%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 +[2026-03-26 04:27:59,752][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:27:59,754][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:27:59,756][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:28:00,541][__main__][INFO] - Iteration 477 took 51s (30.77% Gen, 67.71% Train). Generation: 15s, Training: 35s. Estimated remaining time: 35h 49m 35s. Estimated total time: 43h 11m 56s. Time estimates for 10 more iterations: 8m 38s, 100 more iterations: 1h 26m 23s, 500 more iterations: 7h 11m 59s. +[2026-03-26 04:28:00,543][__main__][INFO] - Starting iteration 477. +[2026-03-26 04:28:00,943][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 47 and human policies 1. +[2026-03-26 04:28:00,944][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:28:16,574][__main__][INFO] - Number of regex retries in iteration 477: 0 +[2026-03-26 04:28:16,575][__main__][INFO] - agents played in iteration 477 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:28:17,329][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:28:17,354][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:28:17,379][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:28:17,404][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:28:17,404][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:28:17,405][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:28:18,163][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:28:18,628][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:28:19,136][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:28:19,642][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:28:20,145][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:28:20,643][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:28:21,156][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:28:21,662][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:28:22,166][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:28:22,668][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:28:23,170][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:28:23,673][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:28:24,178][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:28:24,688][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:28:25,189][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:28:25,690][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:28:26,194][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:28:26,696][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:28:27,199][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:28:27,698][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:28:28,203][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:28:28,709][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:28:29,210][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:28:29,714][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:28:30,215][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:28:30,714][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:28:31,213][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:28:31,714][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:28:32,215][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:28:32,714][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:28:33,219][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:28:33,725][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:28:34,227][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:28:34,726][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:28:35,227][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:28:35,727][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:28:36,229][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:28:36,732][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:28:37,231][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:28:37,730][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:28:38,229][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:28:38,732][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:28:39,233][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:28:39,736][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:28:40,237][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:28:40,737][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:28:41,238][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:28:41,739][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:28:42,240][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:28:42,737][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:28:43,236][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:28:43,740][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:28:44,248][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:28:44,750][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:28:45,249][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:28:45,749][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:28:46,252][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:28:46,756][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:28:47,256][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:28:47,756][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:28:48,256][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:28:48,758][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:28:49,257][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:28:49,756][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:28:50,255][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10164 tokens. +[2026-03-26 04:28:51,052][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.34%, ΔTime: 00:00:32 +[2026-03-26 04:28:51,857][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:28:51,859][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:28:51,860][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:28:52,631][__main__][INFO] - Iteration 478 took 51s (30.24% Gen, 68.26% Train). Generation: 15s, Training: 35s. Estimated remaining time: 35h 41m 15s. Estimated total time: 43h 4m 28s. Time estimates for 10 more iterations: 8m 36s, 100 more iterations: 1h 26m 8s, 500 more iterations: 7h 10m 44s. +[2026-03-26 04:28:52,634][__main__][INFO] - Starting iteration 478. +[2026-03-26 04:28:53,031][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 47 and human policies 1. +[2026-03-26 04:28:53,032][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:29:09,577][__main__][INFO] - Number of regex retries in iteration 478: 0 +[2026-03-26 04:29:09,578][__main__][INFO] - agents played in iteration 478 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:29:10,333][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:29:10,356][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:29:10,380][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:29:10,403][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:29:10,404][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:29:10,404][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:29:11,156][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:29:11,608][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:29:12,113][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:29:12,620][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:29:13,128][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:29:13,634][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:29:14,134][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:29:14,634][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:29:15,134][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:29:15,637][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:29:16,139][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:29:16,640][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:29:17,140][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:29:17,640][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:29:18,144][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:29:18,645][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:29:19,145][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:29:19,642][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:29:20,141][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:29:20,643][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:29:21,145][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:29:21,643][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:29:22,145][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:29:22,647][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:29:23,146][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:29:23,644][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:29:24,141][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:29:24,641][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:29:25,139][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:29:25,637][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:29:26,135][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:29:26,637][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:29:27,138][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:29:27,635][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:29:28,134][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:29:28,632][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:29:29,132][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:29:29,632][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:29:30,131][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:29:30,628][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:29:31,129][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:29:31,627][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:29:32,124][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:29:32,619][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:29:33,120][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:29:33,616][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:29:34,113][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:29:34,611][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:29:35,108][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:29:35,607][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:29:36,102][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:29:36,597][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:29:37,093][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:29:37,589][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:29:38,088][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:29:38,586][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:29:39,080][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:29:39,579][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:29:40,077][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:29:40,574][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:29:41,070][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:29:41,566][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:29:42,064][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:29:42,562][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:29:43,061][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10194 tokens. +[2026-03-26 04:29:43,847][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 60.63%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:32 +[2026-03-26 04:29:44,599][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:29:44,602][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:29:44,603][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:29:45,330][__main__][INFO] - Iteration 479 took 52s (31.64% Gen, 66.97% Train). Generation: 16s, Training: 35s. Estimated remaining time: 36h 10m 53s. Estimated total time: 43h 34m 58s. Time estimates for 10 more iterations: 8m 42s, 100 more iterations: 1h 27m 9s, 500 more iterations: 7h 15m 49s. +[2026-03-26 04:29:45,332][__main__][INFO] - Starting iteration 479. +[2026-03-26 04:29:45,732][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 47 and human policies 1. +[2026-03-26 04:29:45,733][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:30:00,701][__main__][INFO] - Number of regex retries in iteration 479: 0 +[2026-03-26 04:30:00,702][__main__][INFO] - agents played in iteration 479 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:30:01,450][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:30:01,474][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:30:01,497][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:30:01,521][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:30:01,521][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:30:01,522][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:30:02,276][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:30:02,737][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:30:03,245][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:30:03,745][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:30:04,245][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:30:04,748][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:30:05,256][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:30:05,757][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:30:06,260][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:30:06,760][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:30:07,257][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:30:07,761][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:30:08,265][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:30:08,765][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:30:09,264][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:30:09,764][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:30:10,262][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:30:10,761][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:30:11,259][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:30:11,758][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:30:12,255][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:30:12,754][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:30:13,253][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:30:13,752][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:30:14,250][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:30:14,747][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:30:15,244][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:30:15,741][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:30:16,237][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:30:16,735][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:30:17,233][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:30:17,731][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:30:18,228][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:30:18,729][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:30:19,226][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:30:19,726][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:30:20,227][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:30:20,729][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:30:21,231][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:30:21,729][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:30:22,227][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:30:22,732][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:30:23,232][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:30:23,731][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:30:24,231][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:30:24,731][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:30:25,230][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:30:25,728][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:30:26,227][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:30:26,736][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:30:27,245][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:30:27,747][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:30:28,249][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:30:28,750][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:30:29,251][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:30:29,753][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:30:30,257][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:30:30,759][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:30:31,267][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:30:31,769][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:30:32,274][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:30:32,775][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:30:33,274][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:30:33,772][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:30:34,275][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10143 tokens. +[2026-03-26 04:30:35,075][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 60.65%, Block Peak % of device VRAM: 62.28%, ΔTime: 00:00:32 +[2026-03-26 04:30:43,902][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:30:43,904][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:30:43,906][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:30:44,681][__main__][INFO] - Iteration 480 took 58s (25.39% Gen, 73.29% Train). Generation: 14s, Training: 43s. Estimated remaining time: 41h 42m 25s. Estimated total time: 49h 7m 29s. Time estimates for 10 more iterations: 9m 49s, 100 more iterations: 1h 38m 14s, 500 more iterations: 8h 11m 14s. +[2026-03-26 04:30:44,684][__main__][INFO] - Starting iteration 480. +[2026-03-26 04:30:45,084][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 47 and human policies 1. +[2026-03-26 04:30:45,085][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:31:01,436][__main__][INFO] - Number of regex retries in iteration 480: 0 +[2026-03-26 04:31:01,437][__main__][INFO] - agents played in iteration 480 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:31:02,196][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:31:02,222][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:31:02,248][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:31:02,274][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:31:02,274][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:31:02,275][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:31:03,037][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:31:03,496][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:31:04,000][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:31:04,501][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:31:04,999][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:31:05,502][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:31:06,000][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:31:06,499][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:31:06,999][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:31:07,501][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:31:07,999][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:31:08,499][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:31:09,001][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:31:09,507][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:31:10,012][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:31:10,513][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:31:11,016][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:31:11,514][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:31:12,011][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:31:12,508][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:31:13,004][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:31:13,500][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:31:13,999][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:31:14,496][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:31:14,992][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:31:15,488][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:31:15,983][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:31:16,477][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:31:16,972][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:31:17,468][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:31:17,962][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:31:18,456][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:31:18,951][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:31:19,449][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:31:19,956][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:31:20,455][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:31:20,975][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:31:21,477][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:31:21,975][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:31:22,474][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:31:22,973][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:31:23,475][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:31:23,973][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:31:24,470][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:31:24,967][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:31:25,464][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:31:25,962][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:31:26,459][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:31:26,960][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:31:27,457][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:31:27,954][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:31:28,451][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:31:28,949][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:31:29,451][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:31:29,947][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:31:30,445][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:31:30,945][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:31:31,442][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:31:31,950][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:31:32,457][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:31:32,957][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:31:33,458][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:31:33,957][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:31:34,456][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:31:34,956][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10003 tokens. +[2026-03-26 04:31:35,763][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.09%, Current % of VRAM taken: 60.52%, Block Peak % of device VRAM: 62.24%, ΔTime: 00:00:32 +[2026-03-26 04:31:36,497][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:31:36,500][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:31:36,502][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:31:37,769][__main__][INFO] - Iteration 481 took 52s (31.04% Gen, 66.55% Train). Generation: 16s, Training: 35s. Estimated remaining time: 36h 28m 19s. Estimated total time: 43h 54m 16s. Time estimates for 10 more iterations: 8m 46s, 100 more iterations: 1h 27m 48s, 500 more iterations: 7h 19m 2s. +[2026-03-26 04:31:37,772][__main__][INFO] - Starting iteration 481. +[2026-03-26 04:31:38,169][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 48 and human policies 1. +[2026-03-26 04:31:38,170][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:31:52,649][__main__][INFO] - Number of regex retries in iteration 481: 0 +[2026-03-26 04:31:52,650][__main__][INFO] - agents played in iteration 481 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:31:53,396][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:31:53,420][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:31:53,444][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:31:53,468][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:31:53,469][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:31:53,469][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:31:54,224][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:31:54,679][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:31:55,182][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:31:55,680][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:31:56,178][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:31:56,681][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:31:57,182][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:31:57,679][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:31:58,177][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:31:58,674][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:31:59,174][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:31:59,670][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:32:00,166][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:32:00,662][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:32:01,160][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:32:01,656][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:32:02,154][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:32:02,653][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:32:03,153][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:32:03,651][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:32:04,158][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:32:04,656][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:32:05,162][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:32:05,666][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:32:06,168][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:32:06,666][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:32:07,167][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:32:07,665][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:32:08,165][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:32:08,662][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:32:09,160][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:32:09,661][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:32:10,159][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:32:10,657][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:32:11,155][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:32:11,655][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:32:12,154][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:32:12,654][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:32:13,153][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:32:13,652][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:32:14,150][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:32:14,650][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:32:15,151][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:32:15,651][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:32:16,153][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:32:16,653][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:32:17,158][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:32:17,659][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:32:18,167][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:32:18,671][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:32:19,171][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:32:19,670][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:32:20,168][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:32:20,665][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:32:21,167][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:32:21,664][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:32:22,161][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:32:22,659][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:32:23,158][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:32:23,655][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:32:24,155][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:32:24,653][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:32:25,150][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:32:25,652][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:32:26,150][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10131 tokens. +[2026-03-26 04:32:26,940][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 60.53%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:32 +[2026-03-26 04:32:27,804][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:32:27,806][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:32:27,808][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:32:28,532][__main__][INFO] - Iteration 482 took 50s (28.75% Gen, 69.81% Train). Generation: 14s, Training: 35s. Estimated remaining time: 34h 31m 22s. Estimated total time: 41h 58m 10s. Time estimates for 10 more iterations: 8m 23s, 100 more iterations: 1h 23m 56s, 500 more iterations: 6h 59m 41s. +[2026-03-26 04:32:28,534][__main__][INFO] - Starting iteration 482. +[2026-03-26 04:32:28,932][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 48 and human policies 1. +[2026-03-26 04:32:28,933][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:32:42,734][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 04:32:45,812][__main__][INFO] - Number of regex retries in iteration 482: 1 +[2026-03-26 04:32:45,812][__main__][INFO] - agents played in iteration 482 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:32:46,573][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:32:46,596][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:32:46,620][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:32:46,643][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:32:46,644][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:32:46,644][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:32:47,394][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:32:47,853][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:32:48,357][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:32:48,856][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:32:49,354][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:32:49,856][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:32:50,366][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:32:50,868][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:32:51,366][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:32:51,870][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:32:52,371][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:32:52,877][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:32:53,383][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:32:53,894][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:32:54,395][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:32:54,897][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:32:55,401][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:32:55,900][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:32:56,400][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:32:56,900][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:32:57,401][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:32:57,902][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:32:58,402][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:32:58,901][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:32:59,400][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:32:59,900][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:33:00,405][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:33:00,914][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:33:01,418][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:33:01,923][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:33:02,423][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:33:02,924][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:33:03,428][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:33:03,926][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:33:04,424][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:33:04,922][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:33:05,420][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:33:05,918][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:33:06,419][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:33:06,923][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:33:07,420][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:33:07,918][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:33:08,416][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:33:08,915][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:33:09,414][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:33:09,912][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:33:10,410][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:33:10,912][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:33:11,409][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:33:11,910][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:33:12,411][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:33:12,912][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:33:13,413][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:33:13,912][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:33:14,412][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:33:14,937][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:33:15,437][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:33:15,934][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:33:16,432][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:33:16,930][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:33:17,427][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:33:17,924][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:33:18,424][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:33:18,923][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:33:19,419][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10078 tokens. +[2026-03-26 04:33:20,205][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:32 +[2026-03-26 04:33:21,296][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:33:21,298][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:33:21,300][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:33:22,026][__main__][INFO] - Iteration 483 took 53s (31.79% Gen, 66.84% Train). Generation: 16s, Training: 35s. Estimated remaining time: 36h 47m 0s. Estimated total time: 44h 14m 42s. Time estimates for 10 more iterations: 8m 50s, 100 more iterations: 1h 28m 29s, 500 more iterations: 7h 22m 27s. +[2026-03-26 04:33:22,028][__main__][INFO] - Starting iteration 483. +[2026-03-26 04:33:22,426][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 48 and human policies 1. +[2026-03-26 04:33:22,426][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:33:26,299][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 04:33:36,148][__main__][INFO] - Number of regex retries in iteration 483: 1 +[2026-03-26 04:33:36,149][__main__][INFO] - agents played in iteration 483 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:33:36,958][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:33:36,984][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:33:37,009][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:33:37,035][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:33:37,035][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:33:37,036][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:33:37,797][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:33:38,254][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:33:38,765][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:33:39,269][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:33:39,771][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:33:40,281][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:33:40,784][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:33:41,285][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:33:41,789][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:33:42,286][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:33:42,783][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:33:43,288][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:33:43,789][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:33:44,287][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:33:44,786][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:33:45,286][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:33:45,785][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:33:46,285][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:33:46,794][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:33:47,299][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:33:47,798][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:33:48,298][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:33:48,798][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:33:49,299][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:33:49,799][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:33:50,301][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:33:50,803][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:33:51,311][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:33:51,811][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:33:52,311][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:33:52,811][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:33:53,310][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:33:53,807][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:33:54,306][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:33:54,806][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:33:55,310][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:33:55,814][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:33:56,318][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:33:56,819][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:33:57,319][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:33:57,819][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:33:58,318][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:33:58,817][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:33:59,317][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:33:59,814][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:34:00,313][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:34:00,818][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:34:01,328][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:34:01,834][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:34:02,334][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:34:02,834][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:34:03,336][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:34:03,835][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:34:04,334][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:34:04,831][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:34:05,328][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:34:05,828][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:34:06,332][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:34:06,834][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:34:07,340][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:34:07,849][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:34:08,357][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:34:08,860][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:34:09,362][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:34:09,863][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10072 tokens. +[2026-03-26 04:34:10,662][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.15%, ΔTime: 00:00:32 +[2026-03-26 04:34:11,398][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:34:11,401][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:34:11,402][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:34:12,172][__main__][INFO] - Iteration 484 took 49s (27.59% Gen, 70.86% Train). Generation: 13s, Training: 35s. Estimated remaining time: 33h 58m 48s. Estimated total time: 41h 27m 20s. Time estimates for 10 more iterations: 8m 17s, 100 more iterations: 1h 22m 54s, 500 more iterations: 6h 54m 33s. +[2026-03-26 04:34:12,174][__main__][INFO] - Starting iteration 484. +[2026-03-26 04:34:12,573][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 48 and human policies 1. +[2026-03-26 04:34:12,574][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:34:28,485][__main__][INFO] - Number of regex retries in iteration 484: 0 +[2026-03-26 04:34:28,486][__main__][INFO] - agents played in iteration 484 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:34:29,252][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:34:29,276][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:34:29,300][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:34:29,323][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:34:29,324][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:34:29,324][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:34:30,088][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:34:30,541][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:34:31,057][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:34:31,562][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:34:32,066][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:34:32,569][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:34:33,069][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:34:33,570][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:34:34,069][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:34:34,569][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:34:35,069][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:34:35,570][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:34:36,070][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:34:36,572][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:34:37,072][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:34:37,572][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:34:38,070][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:34:38,567][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:34:39,065][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:34:39,562][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:34:40,059][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:34:40,557][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:34:41,065][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:34:41,567][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:34:42,069][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:34:42,574][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:34:43,075][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:34:43,576][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:34:44,076][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:34:44,575][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:34:45,074][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:34:45,572][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:34:46,070][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:34:46,567][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:34:47,064][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:34:47,561][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:34:48,058][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:34:48,556][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:34:49,053][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:34:49,552][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:34:50,051][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:34:50,548][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:34:51,053][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:34:51,576][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:34:52,073][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:34:52,572][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:34:53,072][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:34:53,577][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:34:54,079][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:34:54,578][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:34:55,075][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:34:55,573][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:34:56,070][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:34:56,567][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:34:57,065][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:34:57,562][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:34:58,063][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:34:58,562][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:34:59,059][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:34:59,559][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:35:00,056][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:35:00,553][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:35:01,050][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:35:01,551][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:35:02,052][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10102 tokens. +[2026-03-26 04:35:02,850][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 60.60%, Block Peak % of device VRAM: 62.23%, ΔTime: 00:00:32 +[2026-03-26 04:35:03,606][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:35:03,609][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:35:03,610][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:35:04,327][__main__][INFO] - Iteration 485 took 51s (30.75% Gen, 67.87% Train). Generation: 15s, Training: 35s. Estimated remaining time: 35h 38m 21s. Estimated total time: 43h 7m 45s. Time estimates for 10 more iterations: 8m 37s, 100 more iterations: 1h 26m 15s, 500 more iterations: 7h 11m 17s. +[2026-03-26 04:35:04,330][__main__][INFO] - Starting iteration 485. +[2026-03-26 04:35:04,727][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 48 and human policies 1. +[2026-03-26 04:35:04,728][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:35:19,526][__main__][INFO] - Number of regex retries in iteration 485: 0 +[2026-03-26 04:35:19,528][__main__][INFO] - agents played in iteration 485 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:35:20,292][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:35:20,321][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:35:20,347][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:35:20,374][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:35:20,374][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:35:20,375][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:35:21,140][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:35:21,596][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:35:22,103][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:35:22,603][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:35:23,103][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:35:23,603][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:35:24,108][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:35:24,617][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:35:25,120][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:35:25,620][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:35:26,120][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:35:26,620][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:35:27,133][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:35:27,638][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:35:28,143][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:35:28,643][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:35:29,146][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:35:29,644][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:35:30,148][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:35:30,653][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:35:31,159][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:35:31,663][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:35:32,166][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:35:32,666][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:35:33,165][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:35:33,664][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:35:34,165][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:35:34,664][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:35:35,162][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:35:35,661][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:35:36,162][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:35:36,660][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:35:37,158][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:35:37,655][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:35:38,164][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:35:38,668][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:35:39,167][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:35:39,668][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:35:40,167][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:35:40,666][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:35:41,165][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:35:41,667][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:35:42,170][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:35:42,671][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:35:43,176][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:35:43,688][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:35:44,191][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:35:44,692][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:35:45,197][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:35:45,697][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:35:46,198][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:35:46,701][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:35:47,200][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:35:47,703][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:35:48,206][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:35:48,704][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:35:49,213][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:35:49,723][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:35:50,226][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:35:50,734][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:35:51,257][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:35:51,762][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:35:52,263][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:35:52,764][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:35:53,269][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10154 tokens. +[2026-03-26 04:35:54,068][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 60.64%, Block Peak % of device VRAM: 62.38%, ΔTime: 00:00:32 +[2026-03-26 04:36:16,715][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:36:16,718][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:36:16,719][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:36:17,517][__main__][INFO] - Iteration 486 took 1m 12s (20.33% Gen, 78.57% Train). Generation: 14s, Training: 57s. Estimated remaining time: 53h 8m 54s. Estimated total time: 60h 39m 32s. Time estimates for 10 more iterations: 12m 7s, 100 more iterations: 2h 1m 19s, 500 more iterations: 10h 6m 35s. +[2026-03-26 04:36:17,519][__main__][INFO] - Starting iteration 486. +[2026-03-26 04:36:17,917][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 48 and human policies 1. +[2026-03-26 04:36:17,917][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:36:32,179][__main__][INFO] - Number of regex retries in iteration 486: 0 +[2026-03-26 04:36:32,180][__main__][INFO] - agents played in iteration 486 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:36:32,942][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:36:32,966][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:36:32,989][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:36:33,013][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:36:33,013][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:36:33,014][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:36:33,775][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:36:34,239][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:36:34,749][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:36:35,259][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:36:35,762][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:36:36,266][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:36:36,773][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:36:37,278][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:36:37,780][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:36:38,279][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:36:38,777][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:36:39,278][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:36:39,776][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:36:40,275][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:36:40,776][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:36:41,275][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:36:41,772][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:36:42,274][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:36:42,772][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:36:43,278][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:36:43,790][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:36:44,291][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:36:44,792][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:36:45,293][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:36:45,796][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:36:46,300][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:36:46,806][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:36:47,317][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:36:47,820][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:36:48,328][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:36:48,833][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:36:49,335][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:36:49,837][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:36:50,340][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:36:50,838][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:36:51,340][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:36:51,837][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:36:52,335][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:36:52,834][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:36:53,332][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:36:53,830][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:36:54,329][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:36:54,830][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:36:55,329][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:36:55,828][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:36:56,330][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:36:56,828][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:36:57,326][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:36:57,824][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:36:58,322][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:36:58,820][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:36:59,318][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:36:59,821][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:37:00,328][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:37:00,832][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:37:01,333][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:37:01,836][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:37:02,338][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:37:02,840][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:37:03,341][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:37:03,840][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:37:04,338][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:37:04,850][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:37:05,361][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:37:05,865][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10152 tokens. +[2026-03-26 04:37:06,704][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.25%, Current % of VRAM taken: 60.67%, Block Peak % of device VRAM: 62.47%, ΔTime: 00:00:32 +[2026-03-26 04:37:07,424][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:37:07,426][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:37:07,428][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:37:08,157][__main__][INFO] - Iteration 487 took 50s (28.39% Gen, 70.16% Train). Generation: 14s, Training: 35s. Estimated remaining time: 34h 20m 35s. Estimated total time: 41h 52m 3s. Time estimates for 10 more iterations: 8m 22s, 100 more iterations: 1h 23m 44s, 500 more iterations: 6h 58m 40s. +[2026-03-26 04:37:08,159][__main__][INFO] - Starting iteration 487. +[2026-03-26 04:37:08,559][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 48 and human policies 1. +[2026-03-26 04:37:08,560][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:37:25,965][__main__][INFO] - Number of regex retries in iteration 487: 0 +[2026-03-26 04:37:25,966][__main__][INFO] - agents played in iteration 487 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:37:26,734][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:37:26,758][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:37:26,782][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:37:26,806][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:37:26,807][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:37:26,807][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:37:27,574][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:37:28,032][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:37:28,547][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:37:29,050][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:37:29,550][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:37:30,051][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:37:30,550][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:37:31,064][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:37:31,567][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:37:32,071][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:37:32,571][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:37:33,071][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:37:33,570][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:37:34,071][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:37:34,569][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:37:35,074][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:37:35,577][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:37:36,075][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:37:36,581][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:37:37,080][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:37:37,589][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:37:38,095][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:37:38,597][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:37:39,098][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:37:39,599][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:37:40,098][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:37:40,598][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:37:41,102][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:37:41,608][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:37:42,111][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:37:42,616][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:37:43,117][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:37:43,618][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:37:44,122][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:37:44,621][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:37:45,121][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:37:45,626][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:37:46,131][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:37:46,642][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:37:47,144][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:37:47,648][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:37:48,145][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:37:48,643][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:37:49,143][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:37:49,645][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:37:50,148][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:37:50,652][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:37:51,161][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:37:51,670][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:37:52,168][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:37:52,668][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:37:53,166][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:37:53,663][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:37:54,160][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:37:54,660][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:37:55,160][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:37:55,657][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:37:56,160][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:37:56,661][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:37:57,173][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:37:57,683][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:37:58,207][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:37:58,710][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:37:59,215][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:37:59,717][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10115 tokens. +[2026-03-26 04:38:00,519][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 60.62%, Block Peak % of device VRAM: 62.35%, ΔTime: 00:00:32 +[2026-03-26 04:38:01,253][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:38:01,255][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:38:01,257][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:38:01,984][__main__][INFO] - Iteration 488 took 53s (32.58% Gen, 66.06% Train). Generation: 17s, Training: 35s. Estimated remaining time: 36h 58m 55s. Estimated total time: 44h 31m 17s. Time estimates for 10 more iterations: 8m 54s, 100 more iterations: 1h 29m 2s, 500 more iterations: 7h 25m 12s. +[2026-03-26 04:38:01,987][__main__][INFO] - Starting iteration 488. +[2026-03-26 04:38:02,383][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 48 and human policies 1. +[2026-03-26 04:38:02,384][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:38:10,852][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 04:38:17,185][__main__][INFO] - Number of regex retries in iteration 488: 1 +[2026-03-26 04:38:17,186][__main__][INFO] - agents played in iteration 488 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:38:17,945][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:38:17,969][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:38:17,993][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:38:18,017][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:38:18,017][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:38:18,018][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:38:18,781][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:38:19,235][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:38:19,738][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:38:20,237][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:38:20,737][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:38:21,240][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:38:21,746][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:38:22,247][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:38:22,756][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:38:23,259][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:38:23,761][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:38:24,262][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:38:24,763][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:38:25,265][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:38:25,763][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:38:26,264][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:38:26,766][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:38:27,264][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:38:27,762][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:38:28,260][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:38:28,762][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:38:29,260][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:38:29,757][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:38:30,260][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:38:30,768][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:38:31,276][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:38:31,779][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:38:32,284][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:38:32,786][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:38:33,288][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:38:33,788][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:38:34,289][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:38:34,791][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:38:35,289][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:38:35,787][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:38:36,288][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:38:36,785][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:38:37,282][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:38:37,779][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:38:38,279][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:38:38,776][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:38:39,273][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:38:39,769][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:38:40,265][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:38:40,761][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:38:41,258][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:38:41,753][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:38:42,250][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:38:42,745][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:38:43,240][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:38:43,734][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:38:44,233][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:38:44,728][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:38:45,221][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:38:45,715][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:38:46,210][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:38:46,704][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:38:47,200][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:38:47,696][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:38:48,193][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:38:48,689][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:38:49,184][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:38:49,679][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:38:50,177][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:38:50,675][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10037 tokens. +[2026-03-26 04:38:51,471][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 60.52%, Block Peak % of device VRAM: 62.33%, ΔTime: 00:00:32 +[2026-03-26 04:38:52,217][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:38:52,220][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:38:52,221][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:38:52,940][__main__][INFO] - Iteration 489 took 50s (29.28% Gen, 69.30% Train). Generation: 14s, Training: 35s. Estimated remaining time: 34h 34m 38s. Estimated total time: 42h 7m 50s. Time estimates for 10 more iterations: 8m 25s, 100 more iterations: 1h 24m 15s, 500 more iterations: 7h 1m 18s. +[2026-03-26 04:38:52,942][__main__][INFO] - Starting iteration 489. +[2026-03-26 04:38:53,340][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 48 and human policies 1. +[2026-03-26 04:38:53,340][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:39:07,696][__main__][INFO] - Number of regex retries in iteration 489: 0 +[2026-03-26 04:39:07,697][__main__][INFO] - agents played in iteration 489 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:39:08,456][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:39:08,480][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:39:08,503][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:39:08,527][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:39:08,528][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:39:08,528][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:39:09,286][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:39:09,740][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:39:10,242][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:39:10,739][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:39:11,237][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:39:11,735][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:39:12,234][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:39:12,733][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:39:13,230][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:39:13,730][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:39:14,232][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:39:14,730][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:39:15,230][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:39:15,729][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:39:16,226][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:39:16,725][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:39:17,224][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:39:17,725][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:39:18,224][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:39:18,721][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:39:19,220][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:39:19,726][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:39:20,224][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:39:20,723][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:39:21,221][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:39:21,721][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:39:22,241][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:39:22,745][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:39:23,246][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:39:23,745][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:39:24,251][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:39:24,756][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:39:25,262][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:39:25,763][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:39:26,269][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:39:26,769][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:39:27,268][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:39:27,767][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:39:28,267][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:39:28,766][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:39:29,266][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:39:29,767][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:39:30,268][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:39:30,768][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:39:31,270][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:39:31,770][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:39:32,269][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:39:32,768][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:39:33,266][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:39:33,765][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:39:34,263][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:39:34,761][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:39:35,260][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:39:35,758][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:39:36,256][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:39:36,755][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:39:37,254][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:39:37,749][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:39:38,245][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:39:38,742][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:39:39,239][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:39:39,736][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:39:40,233][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:39:40,730][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:39:41,226][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10057 tokens. +[2026-03-26 04:39:42,021][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 60.53%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 +[2026-03-26 04:39:42,763][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:39:42,766][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:39:42,767][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:39:43,485][__main__][INFO] - Iteration 490 took 50s (28.63% Gen, 69.94% Train). Generation: 14s, Training: 35s. Estimated remaining time: 34h 13m 14s. Estimated total time: 41h 47m 18s. Time estimates for 10 more iterations: 8m 21s, 100 more iterations: 1h 23m 34s, 500 more iterations: 6h 57m 53s. +[2026-03-26 04:39:43,487][__main__][INFO] - Starting iteration 490. +[2026-03-26 04:39:43,885][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 48 and human policies 1. +[2026-03-26 04:39:43,885][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:39:57,198][__main__][INFO] - Number of regex retries in iteration 490: 0 +[2026-03-26 04:39:57,198][__main__][INFO] - agents played in iteration 490 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:39:57,953][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:39:57,977][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:39:58,001][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:39:58,024][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:39:58,025][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:39:58,026][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:39:58,805][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:39:59,263][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:39:59,771][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:40:00,272][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:40:00,774][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:40:01,275][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:40:01,790][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:40:02,295][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:40:02,797][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:40:03,300][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:40:03,800][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:40:04,300][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:40:04,802][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:40:05,309][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:40:05,814][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:40:06,315][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:40:06,817][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:40:07,316][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:40:07,818][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:40:08,323][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:40:08,822][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:40:09,324][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:40:09,833][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:40:10,333][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:40:10,835][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:40:11,336][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:40:11,835][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:40:12,333][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:40:12,831][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:40:13,328][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:40:13,830][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:40:14,333][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:40:14,840][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:40:15,340][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:40:15,867][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:40:16,371][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:40:16,870][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:40:17,373][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:40:17,872][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:40:18,370][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:40:18,875][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:40:19,381][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:40:19,887][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:40:20,387][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:40:20,890][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:40:21,388][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:40:21,890][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:40:22,391][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:40:22,889][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:40:23,386][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:40:23,884][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:40:24,382][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:40:24,880][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:40:25,379][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:40:25,876][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:40:26,375][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:40:26,873][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:40:27,370][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:40:27,868][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:40:28,370][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:40:28,879][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:40:29,385][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:40:29,884][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:40:30,384][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:40:30,885][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10100 tokens. +[2026-03-26 04:40:31,696][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.21%, ΔTime: 00:00:32 +[2026-03-26 04:40:32,406][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:40:32,409][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:40:32,410][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:40:33,666][__main__][INFO] - Iteration 491 took 49s (26.74% Gen, 70.73% Train). Generation: 13s, Training: 35s. Estimated remaining time: 33h 54m 13s. Estimated total time: 41h 29m 7s. Time estimates for 10 more iterations: 8m 17s, 100 more iterations: 1h 22m 58s, 500 more iterations: 6h 54m 51s. +[2026-03-26 04:40:33,669][__main__][INFO] - Starting iteration 491. +[2026-03-26 04:40:34,067][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 49 and human policies 1. +[2026-03-26 04:40:34,068][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:40:50,190][__main__][INFO] - Number of regex retries in iteration 491: 0 +[2026-03-26 04:40:50,191][__main__][INFO] - agents played in iteration 491 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:40:50,942][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:40:50,966][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:40:50,990][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:40:51,013][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:40:51,013][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:40:51,014][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:40:51,779][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:40:52,232][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:40:52,740][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:40:53,248][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:40:53,756][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:40:54,258][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:40:54,765][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:40:55,268][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:40:55,771][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:40:56,270][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:40:56,768][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:40:57,265][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:40:57,763][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:40:58,260][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:40:58,759][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:40:59,257][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:40:59,755][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:41:00,259][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:41:00,765][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:41:01,267][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:41:01,770][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:41:02,272][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:41:02,776][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:41:03,276][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:41:03,775][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:41:04,274][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:41:04,774][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:41:05,272][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:41:05,774][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:41:06,273][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:41:06,775][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:41:07,273][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:41:07,775][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:41:08,273][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:41:08,771][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:41:09,270][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:41:09,768][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:41:10,267][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:41:10,766][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:41:11,264][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:41:11,762][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:41:12,259][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:41:12,756][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:41:13,253][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:41:13,750][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:41:14,248][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:41:14,744][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:41:15,245][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:41:15,741][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:41:16,239][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:41:16,736][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:41:17,233][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:41:17,730][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:41:18,228][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:41:18,725][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:41:19,223][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:41:19,725][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:41:20,222][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:41:20,720][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:41:21,217][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:41:21,717][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:41:22,216][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:41:22,720][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:41:23,229][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:41:23,730][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10039 tokens. +[2026-03-26 04:41:24,545][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 60.53%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 +[2026-03-26 04:41:25,302][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:41:25,305][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:41:25,306][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:41:26,109][__main__][INFO] - Iteration 492 took 52s (30.98% Gen, 67.47% Train). Generation: 16s, Training: 35s. Estimated remaining time: 35h 46m 22s. Estimated total time: 43h 22m 8s. Time estimates for 10 more iterations: 8m 40s, 100 more iterations: 1h 26m 44s, 500 more iterations: 7h 13m 41s. +[2026-03-26 04:41:26,112][__main__][INFO] - Starting iteration 492. +[2026-03-26 04:41:26,511][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 49 and human policies 1. +[2026-03-26 04:41:26,512][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:41:31,428][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 04:41:41,825][__main__][INFO] - Number of regex retries in iteration 492: 1 +[2026-03-26 04:41:41,825][__main__][INFO] - agents played in iteration 492 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:41:42,578][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:41:42,602][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:41:42,631][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:41:42,657][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:41:42,658][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:41:42,658][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:41:43,431][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:41:43,895][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:41:44,404][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:41:44,907][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:41:45,410][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:41:45,918][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:41:46,431][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:41:46,935][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:41:47,439][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:41:47,938][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:41:48,438][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:41:48,937][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:41:49,437][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:41:49,942][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:41:50,441][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:41:50,940][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:41:51,442][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:41:51,942][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:41:52,440][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:41:52,937][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:41:53,434][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:41:53,932][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:41:54,429][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:41:54,936][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:41:55,439][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:41:55,942][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:41:56,445][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:41:56,944][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:41:57,444][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:41:57,943][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:41:58,444][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:41:58,944][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:41:59,446][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:41:59,946][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:42:00,445][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:42:00,946][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:42:01,448][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:42:01,947][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:42:02,454][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:42:02,960][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:42:03,461][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:42:03,962][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:42:04,465][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:42:04,963][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:42:05,465][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:42:05,966][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:42:06,468][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:42:06,974][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:42:07,478][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:42:07,981][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:42:08,481][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:42:08,983][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:42:09,482][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:42:09,981][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:42:10,480][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:42:10,979][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:42:11,479][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:42:11,984][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:42:12,482][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:42:12,981][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:42:13,478][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:42:13,976][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:42:14,475][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:42:14,973][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:42:15,472][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10181 tokens. +[2026-03-26 04:42:16,271][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:32 +[2026-03-26 04:42:17,041][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:42:17,043][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:42:17,045][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:42:17,767][__main__][INFO] - Iteration 493 took 51s (29.88% Gen, 68.71% Train). Generation: 15s, Training: 35s. Estimated remaining time: 35h 6m 10s. Estimated total time: 42h 42m 48s. Time estimates for 10 more iterations: 8m 32s, 100 more iterations: 1h 25m 25s, 500 more iterations: 7h 7m 8s. +[2026-03-26 04:42:17,769][__main__][INFO] - Starting iteration 493. +[2026-03-26 04:42:18,169][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 49 and human policies 1. +[2026-03-26 04:42:18,169][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:42:34,204][__main__][INFO] - Number of regex retries in iteration 493: 0 +[2026-03-26 04:42:34,205][__main__][INFO] - agents played in iteration 493 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:42:34,968][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:42:34,992][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:42:35,016][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:42:35,039][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:42:35,040][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:42:35,041][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:42:35,814][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:42:36,268][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:42:36,786][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:42:37,292][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:42:37,797][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:42:38,300][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:42:38,805][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:42:39,312][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:42:39,819][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:42:40,322][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:42:40,824][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:42:41,326][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:42:41,826][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:42:42,330][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:42:42,828][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:42:43,328][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:42:43,842][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:42:44,340][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:42:44,839][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:42:45,346][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:42:46,009][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:42:46,511][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:42:47,017][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:42:47,517][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:42:48,022][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:42:48,529][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:42:49,048][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:42:49,554][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:42:50,082][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:42:50,587][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:42:51,090][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:42:51,593][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:42:52,101][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:42:52,602][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:42:53,104][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:42:53,605][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:42:54,104][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:42:54,603][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:42:55,103][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:42:55,605][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:42:56,110][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:42:56,612][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:42:57,113][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:42:57,613][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:42:58,114][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:42:58,616][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:42:59,117][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:42:59,628][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:43:00,142][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:43:00,644][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:43:01,149][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:43:01,655][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:43:02,159][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:43:02,659][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:43:03,162][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:43:03,664][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:43:04,164][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:43:04,669][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:43:05,179][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:43:05,681][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:43:06,184][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:43:06,686][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:43:07,186][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:43:07,689][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:43:08,192][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10168 tokens. +[2026-03-26 04:43:09,074][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 60.64%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:33 +[2026-03-26 04:43:09,846][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:43:09,848][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:43:09,850][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:43:10,589][__main__][INFO] - Iteration 494 took 52s (30.59% Gen, 68.00% Train). Generation: 16s, Training: 35s. Estimated remaining time: 36h 3m 33s. Estimated total time: 43h 41m 3s. Time estimates for 10 more iterations: 8m 44s, 100 more iterations: 1h 27m 22s, 500 more iterations: 7h 16m 50s. +[2026-03-26 04:43:10,594][__main__][INFO] - Starting iteration 494. +[2026-03-26 04:43:10,997][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 49 and human policies 1. +[2026-03-26 04:43:10,997][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:45:47,653][mllm.models.large_language_model_local][WARNING] - Response 提案: 0 帽子, 10 书籍, 0 球 did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 04:45:59,095][__main__][INFO] - Number of regex retries in iteration 494: 1 +[2026-03-26 04:45:59,098][__main__][INFO] - agents played in iteration 494 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:46:00,216][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:46:00,241][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:46:00,265][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:46:00,289][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:46:00,290][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:46:00,290][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:46:01,115][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:46:01,788][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:46:02,293][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:46:02,799][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:46:03,298][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:46:03,801][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:46:04,302][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:46:04,802][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:46:05,303][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:46:05,810][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:46:06,321][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:46:06,851][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:46:07,358][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:46:07,863][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:46:08,369][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:46:08,875][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:46:09,381][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:46:09,885][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:46:10,389][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:46:10,891][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:46:11,394][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:46:11,896][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:46:12,397][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:46:12,898][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:46:13,400][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:46:13,898][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:46:14,398][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:46:14,896][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:46:15,393][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:46:15,890][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:46:16,387][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:46:16,886][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:46:17,386][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:46:17,885][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:46:18,384][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:46:18,884][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:46:19,384][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:46:19,885][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:46:20,386][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:46:20,885][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:46:21,384][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:46:21,879][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:46:22,376][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:46:22,875][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:46:23,375][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:46:30,929][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:46:31,435][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:46:31,935][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:46:32,437][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:46:32,936][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:46:33,436][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:46:33,936][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:46:34,435][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:46:34,936][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:46:35,434][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:46:35,934][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:46:36,433][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:46:36,932][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:46:37,434][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:46:37,932][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:46:38,430][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:46:38,928][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:46:39,428][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:46:39,928][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:46:40,434][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10249 tokens. +[2026-03-26 04:46:42,232][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 60.67%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:41 +[2026-03-26 04:46:42,986][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:46:42,988][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:46:42,990][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:46:43,766][__main__][INFO] - Iteration 495 took 3m 32s (79.01% Gen, 20.63% Train). Generation: 2m 48s, Training: 43s. Estimated remaining time: 169h 37m 26s. Estimated total time: 177h 18m 29s. Time estimates for 10 more iterations: 35m 27s, 100 more iterations: 5h 54m 36s, 500 more iterations: 29h 33m 4s. +[2026-03-26 04:46:44,384][__main__][INFO] - Starting iteration 495. +[2026-03-26 04:46:44,787][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 49 and human policies 1. +[2026-03-26 04:46:44,788][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:48:33,986][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 04:48:43,721][__main__][INFO] - Number of regex retries in iteration 495: 1 +[2026-03-26 04:48:43,722][__main__][INFO] - agents played in iteration 495 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:48:44,937][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.59%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:48:44,961][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.59%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:48:44,986][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.59%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:48:45,009][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.59%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:48:45,010][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:48:45,011][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:48:46,060][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:48:46,653][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:48:47,158][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:48:47,663][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:48:48,176][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:48:48,683][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:48:49,184][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:48:49,683][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:48:50,182][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:48:50,681][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:48:51,181][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:48:51,679][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:48:52,178][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:48:52,679][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:48:53,181][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:48:53,680][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:48:54,178][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:48:54,687][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:48:55,201][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:48:55,702][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:48:56,203][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:48:56,703][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:48:57,203][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:48:57,709][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:48:58,222][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:48:58,725][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:48:59,226][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:48:59,725][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:49:00,226][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:49:00,728][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:49:01,226][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:49:01,727][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:49:02,228][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:49:02,728][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:49:03,226][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:49:03,724][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:49:04,223][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:49:04,721][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:49:05,224][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:49:05,722][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:49:06,222][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:49:06,726][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:49:07,238][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:49:07,739][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:49:08,242][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:49:08,745][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:49:09,243][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:49:09,743][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:49:10,243][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:49:10,744][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:49:11,248][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:49:11,750][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:49:12,252][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:49:12,749][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:49:13,249][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:49:13,752][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:49:14,267][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:49:14,769][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:49:15,269][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:49:15,768][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:49:16,268][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:49:16,767][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:49:17,270][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:49:17,768][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:49:18,271][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10163 tokens. +[2026-03-26 04:49:19,063][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 60.64%, Block Peak % of device VRAM: 62.17%, ΔTime: 00:00:33 +[2026-03-26 04:49:19,817][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:49:19,819][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:49:19,821][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:49:20,547][__main__][INFO] - Iteration 496 took 2m 35s (76.36% Gen, 23.18% Train). Generation: 1m 58s, Training: 36s. Estimated remaining time: 122h 4m 20s. Estimated total time: 129h 48m 0s. Time estimates for 10 more iterations: 25m 57s, 100 more iterations: 4h 19m 36s, 500 more iterations: 21h 38m 0s. +[2026-03-26 04:49:20,549][__main__][INFO] - Starting iteration 496. +[2026-03-26 04:49:20,949][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 49 and human policies 1. +[2026-03-26 04:49:20,950][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:51:27,122][__main__][INFO] - Number of regex retries in iteration 496: 0 +[2026-03-26 04:51:27,123][__main__][INFO] - agents played in iteration 496 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:51:27,892][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:51:27,916][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:51:27,939][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:51:27,963][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:51:27,964][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:51:27,964][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:51:28,716][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:51:29,173][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:51:29,681][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:51:30,189][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:51:30,691][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:51:31,192][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:51:31,692][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:51:32,193][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:51:32,693][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:51:33,194][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:51:33,695][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:51:34,200][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:51:34,707][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:51:35,214][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:51:35,717][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:51:36,220][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:51:36,721][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:51:37,223][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:51:37,728][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:51:38,230][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:51:38,732][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:51:39,233][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:51:39,733][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:51:40,234][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:51:40,735][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:51:41,236][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:51:41,735][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:51:42,237][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:51:42,734][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:51:43,232][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:51:43,730][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:51:44,228][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:51:44,726][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:51:45,225][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:51:45,724][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:51:46,223][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:51:46,722][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:51:47,221][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:51:47,720][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:51:48,220][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:51:48,719][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:51:49,219][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:51:49,719][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:51:50,222][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:51:50,721][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:51:51,221][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:51:51,722][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:51:52,227][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:51:52,734][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:51:53,239][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:51:53,739][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:51:54,240][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:51:54,740][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:51:55,239][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:51:55,743][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:51:56,241][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:51:56,740][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:51:57,242][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:51:57,742][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:51:58,247][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:51:58,754][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:51:59,263][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:51:59,765][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:52:00,265][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:52:00,766][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10110 tokens. +[2026-03-26 04:52:01,567][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 60.57%, Block Peak % of device VRAM: 62.17%, ΔTime: 00:00:32 +[2026-03-26 04:52:02,317][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:52:02,320][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:52:02,321][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:52:03,052][__main__][INFO] - Iteration 497 took 2m 42s (77.83% Gen, 21.71% Train). Generation: 2m 6s, Training: 35s. Estimated remaining time: 127h 18m 47s. Estimated total time: 135h 5m 10s. Time estimates for 10 more iterations: 27m 1s, 100 more iterations: 4h 30m 10s, 500 more iterations: 22h 30m 51s. +[2026-03-26 04:52:03,054][__main__][INFO] - Starting iteration 497. +[2026-03-26 04:52:03,454][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 49 and human policies 1. +[2026-03-26 04:52:03,454][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:53:00,698][__main__][INFO] - Number of regex retries in iteration 497: 0 +[2026-03-26 04:53:00,700][__main__][INFO] - agents played in iteration 497 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:53:01,465][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:53:01,489][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:53:01,513][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:53:01,537][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:53:01,538][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:53:01,538][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:53:02,292][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:53:02,751][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:53:03,261][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:53:03,764][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:53:04,269][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:53:04,770][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:53:05,274][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:53:05,778][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:53:06,289][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:53:06,791][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:53:07,291][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:53:07,791][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:53:08,292][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:53:08,791][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:53:09,291][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:53:09,797][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:53:10,311][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:53:10,813][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:53:11,313][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:53:11,813][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:53:12,312][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:53:12,811][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:53:13,311][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:53:13,817][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:53:14,326][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:53:14,827][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:53:15,327][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:53:15,827][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:53:16,326][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:53:16,825][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:53:17,323][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:53:17,822][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:53:18,329][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:53:18,832][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:53:19,339][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:53:19,839][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:53:20,343][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:53:20,843][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:53:21,345][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:53:21,847][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:53:22,346][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:53:22,846][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:53:23,358][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:53:23,866][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:53:24,373][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:53:24,878][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:53:25,383][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:53:25,883][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:53:26,386][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:53:26,888][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:53:27,386][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:53:27,888][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:53:28,386][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:53:28,883][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:53:29,381][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:53:29,879][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:53:30,377][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:53:30,874][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:53:31,372][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:53:31,871][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:53:32,370][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:53:32,870][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:53:33,371][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:53:33,873][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:53:34,372][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10086 tokens. +[2026-03-26 04:53:35,178][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.00%, Current % of VRAM taken: 60.43%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 +[2026-03-26 04:53:35,952][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:53:35,954][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:53:35,956][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:53:36,741][__main__][INFO] - Iteration 498 took 1m 33s (61.36% Gen, 37.79% Train). Generation: 57s, Training: 35s. Estimated remaining time: 69h 56m 27s. Estimated total time: 77h 44m 23s. Time estimates for 10 more iterations: 15m 32s, 100 more iterations: 2h 35m 28s, 500 more iterations: 12h 57m 23s. +[2026-03-26 04:53:36,743][__main__][INFO] - Starting iteration 498. +[2026-03-26 04:53:37,143][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 49 and human policies 1. +[2026-03-26 04:53:37,144][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:53:42,031][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 04:53:52,792][__main__][INFO] - Number of regex retries in iteration 498: 1 +[2026-03-26 04:53:52,793][__main__][INFO] - agents played in iteration 498 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:53:53,564][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:53:53,587][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:53:53,611][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:53:53,634][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:53:53,635][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:53:53,635][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:53:54,398][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:53:54,858][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:53:55,368][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:53:55,869][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:53:56,369][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:53:56,880][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:53:57,388][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:53:57,889][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:53:58,391][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:53:58,890][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:53:59,390][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:53:59,888][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:54:00,387][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:54:00,886][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:54:01,386][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:54:01,888][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:54:02,387][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:54:02,885][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:54:03,387][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:54:03,901][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:54:04,413][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:54:04,913][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:54:05,414][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:54:05,914][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:54:06,420][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:54:06,932][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:54:07,439][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:54:07,958][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:54:08,465][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:54:08,968][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:54:09,472][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:54:09,993][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:54:10,502][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:54:11,001][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:54:11,502][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:54:12,002][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:54:12,503][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:54:13,006][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:54:13,519][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:54:14,023][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:54:14,523][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:54:15,027][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:54:15,531][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:54:16,031][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:54:16,532][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:54:17,033][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:54:17,534][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:54:18,037][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:54:18,538][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:54:19,041][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:54:19,544][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:54:20,051][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:54:20,552][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:54:21,056][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:54:21,557][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:54:22,059][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:54:22,558][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:54:23,058][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:54:23,558][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:54:24,060][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:54:24,560][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:54:25,062][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:54:25,560][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:54:26,060][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:54:26,559][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10156 tokens. +[2026-03-26 04:54:27,341][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 60.57%, Block Peak % of device VRAM: 62.50%, ΔTime: 00:00:32 +[2026-03-26 04:54:28,091][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:54:28,093][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:54:28,095][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:54:28,917][__main__][INFO] - Iteration 499 took 51s (30.23% Gen, 68.19% Train). Generation: 15s, Training: 35s. Estimated remaining time: 35h 19m 53s. Estimated total time: 43h 8m 41s. Time estimates for 10 more iterations: 8m 37s, 100 more iterations: 1h 26m 17s, 500 more iterations: 7h 11m 26s. +[2026-03-26 04:54:28,919][__main__][INFO] - Starting iteration 499. +[2026-03-26 04:54:29,316][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 49 and human policies 1. +[2026-03-26 04:54:29,317][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:54:58,961][__main__][INFO] - Number of regex retries in iteration 499: 0 +[2026-03-26 04:54:58,962][__main__][INFO] - agents played in iteration 499 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:54:59,732][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:54:59,755][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:54:59,779][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:54:59,802][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:54:59,803][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:54:59,803][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:55:00,559][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:55:01,016][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:55:01,523][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:55:02,028][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:55:02,530][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:55:03,030][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:55:03,534][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:55:04,036][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:55:04,537][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:55:05,041][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:55:05,540][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:55:06,048][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:55:06,560][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:55:07,062][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:55:07,563][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:55:08,063][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:55:08,564][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:55:09,064][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:55:09,564][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:55:10,063][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:55:10,565][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:55:11,066][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:55:11,566][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:55:12,066][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:55:12,566][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:55:13,068][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:55:13,565][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:55:14,062][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:55:14,563][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:55:15,064][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:55:15,562][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:55:16,061][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:55:16,563][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:55:17,060][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:55:17,565][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:55:18,064][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:55:18,572][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:55:19,074][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:55:19,583][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:55:20,083][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:55:20,582][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:55:21,083][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:55:21,580][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:55:22,077][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:55:22,577][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:55:23,075][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:55:23,572][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:55:24,072][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:55:24,570][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:55:25,072][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:55:25,569][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:55:26,066][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:55:26,567][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:55:27,067][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:55:27,564][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:55:28,062][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:55:28,562][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:55:29,061][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:55:29,561][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:55:30,061][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:55:30,560][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:55:31,059][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:55:31,558][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:55:32,060][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:55:32,559][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10312 tokens. +[2026-03-26 04:55:33,351][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 60.63%, Block Peak % of device VRAM: 62.48%, ΔTime: 00:00:32 +[2026-03-26 04:55:34,095][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:55:34,097][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:55:34,098][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:55:34,797][__main__][INFO] - Iteration 500 took 1m 5s (45.27% Gen, 53.66% Train). Generation: 29s, Training: 35s. Estimated remaining time: 46h 44m 11s. Estimated total time: 54h 34m 6s. Time estimates for 10 more iterations: 10m 54s, 100 more iterations: 1h 49m 8s, 500 more iterations: 9h 5m 41s. +[2026-03-26 04:55:34,799][__main__][INFO] - Starting iteration 500. +[2026-03-26 04:55:35,197][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 49 and human policies 1. +[2026-03-26 04:55:35,198][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:55:58,238][__main__][INFO] - Number of regex retries in iteration 500: 0 +[2026-03-26 04:55:58,239][__main__][INFO] - agents played in iteration 500 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:55:59,004][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:55:59,027][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:55:59,051][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:55:59,074][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:55:59,075][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:55:59,075][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:55:59,831][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:56:00,284][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:56:00,790][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:56:01,287][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:56:01,787][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:56:02,287][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:56:02,785][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:56:03,291][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:56:03,790][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:56:04,294][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:56:04,799][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:56:05,307][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:56:05,809][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:56:06,310][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:56:06,808][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:56:07,307][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:56:07,806][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:56:08,304][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:56:08,803][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:56:09,300][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:56:09,798][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:56:10,299][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:56:10,795][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:56:11,294][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:56:11,795][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:56:12,292][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:56:12,791][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:56:13,292][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:56:13,790][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:56:14,287][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:56:14,784][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:56:15,284][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:56:15,780][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:56:16,277][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:56:16,774][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:56:17,271][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:56:17,767][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:56:18,264][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:56:18,764][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:56:19,260][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:56:19,763][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:56:20,262][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:56:20,765][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:56:21,265][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:56:21,764][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:56:22,266][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:56:22,765][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:56:23,266][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:56:23,764][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:56:24,264][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:56:24,763][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:56:25,262][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:56:25,762][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:56:26,260][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:56:26,760][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:56:27,257][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:56:27,758][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:56:28,254][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:56:28,750][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:56:29,246][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:56:29,742][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:56:30,237][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:56:30,732][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:56:31,228][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:56:31,724][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10034 tokens. +[2026-03-26 04:56:32,519][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.02%, Current % of VRAM taken: 60.44%, Block Peak % of device VRAM: 62.08%, ΔTime: 00:00:32 +[2026-03-26 04:56:33,296][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:56:33,298][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:56:33,299][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:56:34,588][__main__][INFO] - Iteration 501 took 59s (38.80% Gen, 59.03% Train). Generation: 23s, Training: 35s. Estimated remaining time: 41h 38m 39s. Estimated total time: 49h 29m 33s. Time estimates for 10 more iterations: 9m 53s, 100 more iterations: 1h 38m 59s, 500 more iterations: 8h 14m 55s. +[2026-03-26 04:56:34,590][__main__][INFO] - Starting iteration 501. +[2026-03-26 04:56:34,989][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 50 and human policies 1. +[2026-03-26 04:56:34,989][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:57:03,334][__main__][INFO] - Number of regex retries in iteration 501: 0 +[2026-03-26 04:57:03,335][__main__][INFO] - agents played in iteration 501 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:57:04,101][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:57:04,125][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:57:04,148][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:57:04,173][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:57:04,173][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:57:04,174][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:57:04,934][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:57:05,386][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:57:05,888][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:57:06,387][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:57:06,886][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:57:07,390][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:57:07,903][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:57:08,405][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:57:08,907][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:57:09,411][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:57:09,911][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:57:10,411][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:57:10,918][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:57:11,429][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:57:11,934][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:57:12,439][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:57:12,938][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:57:13,440][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:57:13,942][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:57:14,444][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:57:14,956][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:57:15,465][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:57:15,966][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:57:16,467][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:57:16,968][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:57:17,469][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:57:17,969][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:57:19,337][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:57:19,838][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:57:20,344][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:57:20,850][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:57:21,355][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:57:21,863][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:57:22,629][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:57:23,133][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:57:23,633][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:57:24,134][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:57:24,638][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:57:25,138][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:57:25,641][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:57:26,147][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:57:26,655][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:57:27,165][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:57:27,666][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:57:28,167][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:57:28,669][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:57:29,168][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:57:29,667][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:57:30,166][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:57:30,665][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:57:31,168][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:57:31,669][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:57:32,172][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:57:32,673][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:57:33,174][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:57:33,671][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:57:34,172][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:57:34,668][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:57:35,165][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:57:35,662][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:57:36,159][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:57:36,658][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:57:37,155][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:57:37,655][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:57:38,155][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10156 tokens. +[2026-03-26 04:57:39,054][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.58%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:34 +[2026-03-26 04:57:39,817][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:57:39,830][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:57:39,831][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:57:40,615][__main__][INFO] - Iteration 502 took 1m 5s (43.19% Gen, 55.61% Train). Generation: 28s, Training: 36s. Estimated remaining time: 46h 49m 20s. Estimated total time: 54h 41m 20s. Time estimates for 10 more iterations: 10m 56s, 100 more iterations: 1h 49m 22s, 500 more iterations: 9h 6m 53s. +[2026-03-26 04:57:40,635][__main__][INFO] - Starting iteration 502. +[2026-03-26 04:57:41,041][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 50 and human policies 1. +[2026-03-26 04:57:41,042][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:58:05,081][__main__][INFO] - Number of regex retries in iteration 502: 0 +[2026-03-26 04:58:05,082][__main__][INFO] - agents played in iteration 502 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:58:05,930][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:58:05,955][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:58:05,981][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:58:06,006][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:58:06,006][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:58:06,007][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:58:06,774][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:58:07,250][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:58:07,753][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:58:08,253][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:58:08,753][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:58:09,253][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:58:09,759][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:58:10,272][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:58:10,775][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:58:11,276][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:58:11,778][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:58:12,280][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:58:12,780][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:58:13,281][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:58:13,782][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:58:14,282][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:58:14,785][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:58:15,284][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:58:15,783][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:58:16,284][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:58:16,782][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:58:17,279][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:58:17,777][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:58:18,278][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:58:18,779][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:58:19,277][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:58:19,775][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:58:20,279][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:58:20,777][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:58:21,274][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:58:21,772][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:58:22,282][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:58:22,797][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:58:23,299][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:58:23,801][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:58:24,304][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:58:24,828][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:58:25,339][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:58:25,845][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:58:26,349][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:58:26,850][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:58:27,351][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:58:27,855][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:58:28,357][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:58:28,867][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:58:29,371][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:58:29,873][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:58:30,379][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:58:30,880][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:58:31,380][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:58:31,881][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:58:32,380][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:58:32,883][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:58:33,387][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:58:33,889][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:58:34,389][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:58:34,893][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:58:35,397][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:58:35,902][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:58:36,407][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:58:36,910][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:58:37,413][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:58:37,914][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:58:38,416][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:58:38,918][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10092 tokens. +[2026-03-26 04:58:39,745][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 60.60%, Block Peak % of device VRAM: 62.21%, ΔTime: 00:00:32 +[2026-03-26 04:58:40,572][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:58:40,575][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:58:40,576][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:58:41,310][__main__][INFO] - Iteration 503 took 1m 0s (39.89% Gen, 58.89% Train). Generation: 24s, Training: 35s. Estimated remaining time: 42h 20m 28s. Estimated total time: 50h 13m 29s. Time estimates for 10 more iterations: 10m 2s, 100 more iterations: 1h 40m 26s, 500 more iterations: 8h 22m 14s. +[2026-03-26 04:58:41,317][__main__][INFO] - Starting iteration 503. +[2026-03-26 04:58:41,715][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 50 and human policies 1. +[2026-03-26 04:58:41,716][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 04:59:05,552][__main__][INFO] - Number of regex retries in iteration 503: 0 +[2026-03-26 04:59:05,553][__main__][INFO] - agents played in iteration 503 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 04:59:06,311][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:59:06,335][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:59:06,358][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:59:06,382][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 04:59:06,383][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 04:59:06,383][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 04:59:07,139][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 04:59:07,597][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 04:59:08,099][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 04:59:08,598][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 04:59:09,100][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 04:59:09,597][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 04:59:10,095][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 04:59:10,592][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 04:59:11,089][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 04:59:11,584][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 04:59:12,080][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 04:59:12,578][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 04:59:13,078][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 04:59:13,584][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 04:59:14,088][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 04:59:14,587][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 04:59:15,084][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 04:59:15,591][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 04:59:16,095][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 04:59:16,598][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 04:59:17,100][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 04:59:17,601][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 04:59:18,102][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 04:59:18,603][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 04:59:19,104][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 04:59:19,601][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 04:59:20,103][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 04:59:20,609][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 04:59:21,112][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 04:59:21,610][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 04:59:22,110][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 04:59:22,608][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 04:59:23,106][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 04:59:23,605][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 04:59:24,104][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 04:59:24,604][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 04:59:25,104][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 04:59:25,603][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 04:59:26,101][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 04:59:26,600][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 04:59:27,099][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 04:59:27,598][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 04:59:28,099][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 04:59:28,597][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 04:59:29,103][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 04:59:29,601][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 04:59:30,099][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 04:59:30,602][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 04:59:31,106][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 04:59:31,611][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 04:59:32,110][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 04:59:32,609][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 04:59:33,107][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 04:59:33,605][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 04:59:34,103][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 04:59:34,601][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 04:59:35,099][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 04:59:35,597][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 04:59:36,095][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 04:59:36,593][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 04:59:37,091][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 04:59:37,587][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 04:59:38,087][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 04:59:38,595][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 04:59:39,103][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10072 tokens. +[2026-03-26 04:59:39,908][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 60.56%, Block Peak % of device VRAM: 62.18%, ΔTime: 00:00:32 +[2026-03-26 04:59:40,692][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 04:59:40,694][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 04:59:40,696][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 04:59:41,422][__main__][INFO] - Iteration 504 took 59s (39.92% Gen, 58.86% Train). Generation: 23s, Training: 35s. Estimated remaining time: 41h 51m 20s. Estimated total time: 49h 45m 21s. Time estimates for 10 more iterations: 9m 57s, 100 more iterations: 1h 39m 30s, 500 more iterations: 8h 17m 33s. +[2026-03-26 04:59:41,424][__main__][INFO] - Starting iteration 504. +[2026-03-26 04:59:41,824][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 50 and human policies 1. +[2026-03-26 04:59:41,825][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:00:08,384][__main__][INFO] - Number of regex retries in iteration 504: 0 +[2026-03-26 05:00:08,385][__main__][INFO] - agents played in iteration 504 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:00:09,141][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:00:09,165][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:00:09,188][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:00:09,212][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:00:09,212][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:00:09,213][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:00:09,965][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:00:10,422][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:00:10,924][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:00:11,429][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:00:11,935][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:00:12,435][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:00:12,937][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:00:13,440][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:00:13,943][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:00:14,444][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:00:14,944][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:00:15,445][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:00:15,944][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:00:16,448][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:00:16,949][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:00:17,448][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:00:17,950][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:00:18,453][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:00:18,948][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:00:19,445][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:00:19,948][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:00:20,445][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:00:20,946][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:00:21,447][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:00:21,955][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:00:22,458][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:00:22,969][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:00:23,472][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:00:23,973][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:00:24,475][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:00:24,980][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:00:25,481][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:00:25,983][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:00:26,482][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:00:26,982][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:00:27,482][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:00:27,982][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:00:28,481][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:00:28,980][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:00:29,479][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:00:29,980][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:00:30,481][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:00:30,980][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:00:31,480][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:00:31,982][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:00:32,483][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:00:32,984][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:00:33,485][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:00:33,984][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:00:34,484][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:00:34,983][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:00:35,480][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:00:35,978][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:00:36,476][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:00:36,975][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:00:37,473][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:00:37,969][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:00:38,466][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:00:38,967][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:00:39,465][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:00:39,963][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:00:40,462][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:00:40,963][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:00:41,461][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:00:41,957][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10210 tokens. +[2026-03-26 05:00:42,748][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.09%, Current % of VRAM taken: 60.51%, Block Peak % of device VRAM: 62.28%, ΔTime: 00:00:32 +[2026-03-26 05:00:43,484][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:00:43,487][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:00:43,488][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:00:44,195][__main__][INFO] - Iteration 505 took 1m 2s (42.58% Gen, 56.28% Train). Generation: 26s, Training: 35s. Estimated remaining time: 44h 3m 28s. Estimated total time: 51h 58m 32s. Time estimates for 10 more iterations: 10m 23s, 100 more iterations: 1h 43m 57s, 500 more iterations: 8h 39m 45s. +[2026-03-26 05:00:44,197][__main__][INFO] - Starting iteration 505. +[2026-03-26 05:00:44,596][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 50 and human policies 1. +[2026-03-26 05:00:44,597][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:00:58,812][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 20 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 05:01:05,279][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 1 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 05:01:06,174][__main__][INFO] - Number of regex retries in iteration 505: 2 +[2026-03-26 05:01:06,175][__main__][INFO] - agents played in iteration 505 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:01:06,935][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:01:06,962][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:01:06,987][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:01:07,012][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:01:07,013][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:01:07,013][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:01:07,783][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:01:08,243][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:01:08,750][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:01:09,268][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:01:09,770][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:01:10,274][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:01:10,777][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:01:11,281][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:01:11,782][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:01:12,284][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:01:12,795][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:01:13,299][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:01:13,798][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:01:14,300][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:01:14,798][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:01:15,297][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:01:15,794][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:01:16,293][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:01:16,792][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:01:17,291][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:01:17,792][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:01:18,294][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:01:18,797][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:01:19,297][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:01:19,797][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:01:20,299][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:01:20,805][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:01:21,306][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:01:21,805][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:01:22,305][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:01:22,806][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:01:23,306][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:01:23,805][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:01:24,305][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:01:24,806][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:01:25,305][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:01:25,805][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:01:26,305][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:01:26,807][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:01:27,306][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:01:27,805][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:01:28,305][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:01:28,801][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:01:29,301][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:01:29,798][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:01:30,296][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:01:30,793][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:01:31,290][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:01:31,788][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:01:32,285][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:01:32,783][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:01:33,280][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:01:33,778][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:01:34,276][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:01:34,777][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:01:35,273][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:01:35,770][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:01:36,271][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:01:36,773][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:01:37,271][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:01:37,766][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:01:38,263][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:01:38,762][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:01:39,261][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:01:39,760][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10202 tokens. +[2026-03-26 05:01:40,547][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.58%, Block Peak % of device VRAM: 62.40%, ΔTime: 00:00:32 +[2026-03-26 05:01:41,298][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:01:41,302][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:01:41,304][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:01:42,032][__main__][INFO] - Iteration 506 took 57s (37.57% Gen, 61.16% Train). Generation: 21s, Training: 35s. Estimated remaining time: 39h 55m 45s. Estimated total time: 47h 51m 47s. Time estimates for 10 more iterations: 9m 34s, 100 more iterations: 1h 35m 43s, 500 more iterations: 7h 58m 37s. +[2026-03-26 05:01:42,034][__main__][INFO] - Starting iteration 506. +[2026-03-26 05:01:42,434][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 50 and human policies 1. +[2026-03-26 05:01:42,435][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:02:02,504][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 05:02:03,992][__main__][INFO] - Number of regex retries in iteration 506: 1 +[2026-03-26 05:02:03,993][__main__][INFO] - agents played in iteration 506 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:02:04,748][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:02:04,773][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:02:04,798][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:02:04,822][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:02:04,823][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:02:04,824][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:02:05,589][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:02:06,046][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:02:06,565][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:02:07,071][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:02:07,578][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:02:08,079][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:02:08,579][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:02:09,081][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:02:09,580][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:02:10,089][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:02:10,594][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:02:11,100][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:02:11,600][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:02:12,101][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:02:12,600][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:02:13,099][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:02:13,599][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:02:14,100][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:02:14,601][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:02:15,106][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:02:15,609][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:02:16,109][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:02:16,608][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:02:17,107][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:02:17,607][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:02:18,108][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:02:18,605][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:02:19,105][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:02:19,608][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:02:20,106][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:02:20,605][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:02:21,102][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:02:21,603][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:02:22,100][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:02:22,597][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:02:23,093][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:02:23,590][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:02:24,087][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:02:24,583][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:02:25,079][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:02:25,575][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:02:26,072][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:02:26,572][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:02:27,071][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:02:27,568][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:02:28,068][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:02:28,565][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:02:29,065][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:02:29,565][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:02:30,062][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:02:30,563][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:02:31,075][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:02:31,587][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:02:32,089][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:02:32,589][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:02:33,089][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:02:33,589][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:02:34,090][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:02:34,591][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:02:35,096][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:02:35,604][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:02:36,112][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:02:36,614][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:02:37,116][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:02:37,619][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10183 tokens. +[2026-03-26 05:02:38,428][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.58%, Block Peak % of device VRAM: 62.22%, ΔTime: 00:00:32 +[2026-03-26 05:02:39,139][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:02:39,141][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:02:39,143][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:02:40,018][__main__][INFO] - Iteration 507 took 57s (37.44% Gen, 61.04% Train). Generation: 21s, Training: 35s. Estimated remaining time: 40h 2m 13s. Estimated total time: 47h 59m 13s. Time estimates for 10 more iterations: 9m 35s, 100 more iterations: 1h 35m 58s, 500 more iterations: 7h 59m 52s. +[2026-03-26 05:02:40,020][__main__][INFO] - Starting iteration 507. +[2026-03-26 05:02:40,420][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 50 and human policies 1. +[2026-03-26 05:02:40,420][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:02:56,468][__main__][INFO] - Number of regex retries in iteration 507: 0 +[2026-03-26 05:02:56,469][__main__][INFO] - agents played in iteration 507 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:02:57,224][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:02:57,248][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:02:57,271][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:02:57,295][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:02:57,295][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:02:57,296][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:02:58,048][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:02:58,503][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:02:59,005][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:02:59,510][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:03:00,017][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:03:00,528][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:03:01,034][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:03:01,534][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:03:02,035][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:03:02,540][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:03:03,043][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:03:03,542][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:03:04,045][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:03:04,544][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:03:05,042][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:03:05,541][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:03:06,046][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:03:06,551][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:03:07,058][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:03:07,566][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:03:08,067][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:03:08,567][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:03:09,069][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:03:09,569][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:03:10,069][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:03:10,568][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:03:11,070][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:03:11,567][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:03:12,064][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:03:12,565][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:03:13,065][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:03:13,565][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:03:14,063][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:03:14,561][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:03:15,062][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:03:15,566][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:03:16,074][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:03:16,581][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:03:17,080][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:03:17,578][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:03:18,077][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:03:18,577][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:03:19,074][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:03:19,571][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:03:20,069][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:03:20,568][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:03:21,066][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:03:21,566][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:03:22,063][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:03:22,561][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:03:23,059][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:03:23,560][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:03:24,061][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:03:24,561][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:03:25,058][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:03:25,555][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:03:26,055][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:03:26,551][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:03:27,047][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:03:27,549][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:03:28,051][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:03:28,554][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:03:29,060][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:03:29,560][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:03:30,060][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10161 tokens. +[2026-03-26 05:03:30,867][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.02%, Current % of VRAM taken: 60.44%, Block Peak % of device VRAM: 62.03%, ΔTime: 00:00:32 +[2026-03-26 05:03:31,576][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:03:31,578][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:03:31,580][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:03:32,306][__main__][INFO] - Iteration 508 took 51s (30.93% Gen, 67.67% Train). Generation: 16s, Training: 35s. Estimated remaining time: 35h 16m 28s. Estimated total time: 43h 14m 20s. Time estimates for 10 more iterations: 8m 38s, 100 more iterations: 1h 26m 28s, 500 more iterations: 7h 12m 23s. +[2026-03-26 05:03:32,308][__main__][INFO] - Starting iteration 508. +[2026-03-26 05:03:32,708][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 50 and human policies 1. +[2026-03-26 05:03:32,709][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:03:38,366][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 05:03:46,789][__main__][INFO] - Number of regex retries in iteration 508: 1 +[2026-03-26 05:03:46,790][__main__][INFO] - agents played in iteration 508 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:03:47,538][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.44%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:03:47,562][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.44%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:03:47,586][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.44%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:03:47,609][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.44%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:03:47,610][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:03:47,610][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:03:48,362][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:03:48,817][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:03:49,320][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:03:49,822][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:03:50,319][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:03:50,822][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:03:51,321][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:03:51,820][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:03:52,319][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:03:52,825][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:03:53,327][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:03:53,834][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:03:54,347][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:03:54,849][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:03:55,352][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:03:55,855][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:03:56,359][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:03:56,858][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:03:57,358][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:03:57,858][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:03:58,358][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:03:58,857][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:03:59,356][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:03:59,855][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:04:00,356][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:04:00,854][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:04:01,352][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:04:01,849][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:04:02,347][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:04:02,844][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:04:03,345][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:04:03,841][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:04:04,338][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:04:04,835][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:04:05,339][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:04:05,838][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:04:06,339][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:04:06,837][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:04:07,335][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:04:07,836][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:04:08,333][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:04:08,830][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:04:09,331][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:04:09,830][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:04:10,334][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:04:10,835][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:04:11,344][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:04:11,844][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:04:12,345][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:04:12,845][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:04:13,345][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:04:13,846][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:04:14,345][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:04:14,847][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:04:15,346][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:04:15,845][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:04:16,347][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:04:16,846][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:04:17,345][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:04:17,844][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:04:18,343][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:04:18,842][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:04:19,341][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:04:19,839][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:04:20,338][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10183 tokens. +[2026-03-26 05:04:21,129][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 60.57%, Block Peak % of device VRAM: 62.07%, ΔTime: 00:00:32 +[2026-03-26 05:04:21,894][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:04:21,896][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:04:21,898][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:04:22,755][__main__][INFO] - Iteration 509 took 50s (28.13% Gen, 70.15% Train). Generation: 14s, Training: 35s. Estimated remaining time: 33h 43m 40s. Estimated total time: 41h 42m 22s. Time estimates for 10 more iterations: 8m 20s, 100 more iterations: 1h 23m 24s, 500 more iterations: 6h 57m 3s. +[2026-03-26 05:04:22,758][__main__][INFO] - Starting iteration 509. +[2026-03-26 05:04:23,156][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 50 and human policies 1. +[2026-03-26 05:04:23,157][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:04:37,719][__main__][INFO] - Number of regex retries in iteration 509: 0 +[2026-03-26 05:04:37,720][__main__][INFO] - agents played in iteration 509 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:04:38,472][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:04:38,496][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:04:38,519][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:04:38,543][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:04:38,543][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:04:38,544][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:04:39,309][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:04:39,761][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:04:40,263][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:04:40,760][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:04:41,260][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:04:41,757][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:04:42,254][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:04:42,750][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:04:43,247][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:04:43,748][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:04:44,245][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:04:44,749][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:04:45,247][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:04:45,748][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:04:46,249][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:04:46,745][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:04:47,245][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:04:47,764][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:04:48,262][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:04:48,761][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:04:49,263][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:04:49,761][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:04:50,265][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:04:50,768][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:04:51,270][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:04:51,771][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:04:52,270][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:04:52,776][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:04:53,280][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:04:53,782][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:04:54,281][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:04:54,780][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:04:55,279][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:04:55,782][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:04:56,281][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:04:56,781][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:04:57,284][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:04:57,782][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:04:58,282][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:04:58,784][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:04:59,286][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:04:59,783][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:05:00,284][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:05:00,785][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:05:01,281][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:05:01,782][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:05:02,279][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:05:02,776][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:05:03,273][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:05:03,771][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:05:04,268][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:05:04,766][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:05:05,264][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:05:05,762][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:05:06,260][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:05:06,759][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:05:07,256][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:05:07,758][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:05:08,253][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:05:08,749][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:05:09,249][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:05:09,746][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:05:10,246][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:05:10,743][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:05:11,240][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10190 tokens. +[2026-03-26 05:05:12,055][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 60.54%, Block Peak % of device VRAM: 62.21%, ΔTime: 00:00:32 +[2026-03-26 05:05:12,813][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:05:12,816][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:05:12,817][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:05:13,548][__main__][INFO] - Iteration 510 took 50s (28.90% Gen, 69.65% Train). Generation: 14s, Training: 35s. Estimated remaining time: 34h 0m 5s. Estimated total time: 41h 59m 38s. Time estimates for 10 more iterations: 8m 23s, 100 more iterations: 1h 23m 59s, 500 more iterations: 6h 59m 56s. +[2026-03-26 05:05:13,551][__main__][INFO] - Starting iteration 510. +[2026-03-26 05:05:13,968][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 50 and human policies 1. +[2026-03-26 05:05:13,969][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:05:29,682][__main__][INFO] - Number of regex retries in iteration 510: 0 +[2026-03-26 05:05:29,683][__main__][INFO] - agents played in iteration 510 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:05:30,459][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:05:30,485][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:05:30,511][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:05:30,536][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:05:30,537][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:05:30,538][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:05:31,299][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:05:31,764][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:05:32,268][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:05:32,767][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:05:33,265][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:05:33,764][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:05:34,263][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:05:34,769][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:05:35,275][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:05:35,775][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:05:36,280][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:05:36,779][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:05:37,277][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:05:37,776][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:05:38,282][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:05:38,791][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:05:39,298][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:05:39,798][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:05:40,299][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:05:40,799][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:05:41,298][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:05:41,797][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:05:42,301][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:05:42,812][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:05:43,313][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:05:43,812][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:05:44,313][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:05:44,812][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:05:45,317][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:05:45,816][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:05:46,315][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:05:46,813][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:05:47,322][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:05:47,831][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:05:48,332][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:05:48,833][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:05:49,333][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:05:49,832][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:05:50,331][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:05:50,835][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:05:51,333][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:05:51,832][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:05:52,329][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:05:52,826][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:05:53,324][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:05:53,825][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:05:54,323][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:05:54,820][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:05:55,319][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:05:55,816][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:05:56,314][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:05:56,811][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:05:57,310][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:05:57,806][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:05:58,304][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:05:58,800][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:05:59,300][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:05:59,799][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:06:00,299][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:06:00,799][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:06:01,296][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:06:01,795][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:06:02,293][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:06:02,790][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:06:03,288][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10049 tokens. +[2026-03-26 05:06:04,077][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 60.57%, Block Peak % of device VRAM: 62.08%, ΔTime: 00:00:32 +[2026-03-26 05:06:04,822][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:06:04,825][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:06:04,826][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:06:06,336][__main__][INFO] - Iteration 511 took 52s (30.01% Gen, 67.11% Train). Generation: 15s, Training: 35s. Estimated remaining time: 35h 37m 58s. Estimated total time: 43h 38m 24s. Time estimates for 10 more iterations: 8m 43s, 100 more iterations: 1h 27m 16s, 500 more iterations: 7h 16m 24s. +[2026-03-26 05:06:06,338][__main__][INFO] - Starting iteration 511. +[2026-03-26 05:06:06,745][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 51 and human policies 1. +[2026-03-26 05:06:06,746][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:06:21,902][__main__][INFO] - Number of regex retries in iteration 511: 0 +[2026-03-26 05:06:21,902][__main__][INFO] - agents played in iteration 511 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:06:22,659][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:06:22,682][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:06:22,706][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:06:22,729][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:06:22,730][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:06:22,730][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:06:23,495][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:06:23,954][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:06:24,461][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:06:24,977][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:06:25,477][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:06:25,982][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:06:26,483][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:06:26,981][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:06:27,481][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:06:27,986][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:06:28,500][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:06:29,004][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:06:29,512][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:06:30,015][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:06:30,518][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:06:31,022][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:06:31,529][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:06:32,034][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:06:32,536][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:06:33,035][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:06:33,535][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:06:34,033][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:06:34,532][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:06:35,034][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:06:35,532][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:06:36,029][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:06:36,530][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:06:37,030][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:06:37,528][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:06:38,029][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:06:38,529][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:06:39,026][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:06:39,535][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:06:40,043][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:06:40,544][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:06:41,046][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:06:41,548][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:06:42,050][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:06:42,552][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:06:43,051][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:06:43,550][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:06:44,049][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:06:44,548][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:06:45,046][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:06:45,545][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:06:46,043][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:06:46,545][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:06:47,043][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:06:47,542][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:06:48,043][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:06:48,544][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:06:49,043][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:06:49,544][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:06:50,044][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:06:50,544][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:06:51,045][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:06:51,545][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:06:52,041][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:06:52,538][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:06:53,033][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:06:53,530][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:06:54,027][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:06:54,525][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:06:55,022][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:06:55,520][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10167 tokens. +[2026-03-26 05:06:56,325][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 60.50%, Block Peak % of device VRAM: 62.24%, ΔTime: 00:00:32 +[2026-03-26 05:06:57,050][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:06:57,052][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:06:57,053][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:06:57,895][__main__][INFO] - Iteration 512 took 51s (29.63% Gen, 68.72% Train). Generation: 15s, Training: 35s. Estimated remaining time: 34h 36m 13s. Estimated total time: 42h 37m 31s. Time estimates for 10 more iterations: 8m 31s, 100 more iterations: 1h 25m 15s, 500 more iterations: 7h 6m 15s. +[2026-03-26 05:06:57,897][__main__][INFO] - Starting iteration 512. +[2026-03-26 05:06:58,319][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 51 and human policies 1. +[2026-03-26 05:06:58,319][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:07:13,772][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 1, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 05:07:14,721][__main__][INFO] - Number of regex retries in iteration 512: 1 +[2026-03-26 05:07:14,722][__main__][INFO] - agents played in iteration 512 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:07:15,495][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:07:15,519][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:07:15,543][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:07:15,567][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:07:15,568][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:07:15,568][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:07:16,323][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:07:16,782][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:07:17,287][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:07:17,791][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:07:18,303][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:07:18,805][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:07:19,305][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:07:19,810][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:07:20,310][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:07:20,812][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:07:21,310][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:07:21,812][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:07:22,314][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:07:22,818][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:07:23,323][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:07:23,829][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:07:24,337][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:07:24,841][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:07:25,343][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:07:25,842][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:07:26,346][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:07:26,848][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:07:27,353][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:07:27,856][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:07:28,358][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:07:28,858][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:07:29,358][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:07:29,857][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:07:30,360][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:07:30,861][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:07:31,367][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:07:31,892][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:07:32,397][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:07:32,901][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:07:33,402][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:07:33,903][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:07:34,404][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:07:34,903][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:07:35,401][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:07:35,911][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:07:36,425][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:07:36,929][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:07:37,431][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:07:37,935][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:07:38,438][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:07:38,939][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:07:39,440][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:07:39,940][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:07:40,441][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:07:40,943][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:07:41,441][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:07:41,940][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:07:42,442][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:07:42,940][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:07:43,442][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:07:43,937][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:07:44,435][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:07:44,938][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:07:45,436][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:07:45,934][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:07:46,435][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:07:46,936][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:07:47,436][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:07:47,938][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:07:48,439][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10262 tokens. +[2026-03-26 05:07:49,239][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 60.64%, Block Peak % of device VRAM: 62.13%, ΔTime: 00:00:32 +[2026-03-26 05:07:49,994][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:07:49,997][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:07:49,998][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:07:50,745][__main__][INFO] - Iteration 513 took 52s (31.29% Gen, 67.29% Train). Generation: 16s, Training: 35s. Estimated remaining time: 35h 39m 10s. Estimated total time: 43h 41m 21s. Time estimates for 10 more iterations: 8m 44s, 100 more iterations: 1h 27m 22s, 500 more iterations: 7h 16m 53s. +[2026-03-26 05:07:50,747][__main__][INFO] - Starting iteration 513. +[2026-03-26 05:07:51,163][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 51 and human policies 1. +[2026-03-26 05:07:51,164][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:08:05,719][__main__][INFO] - Number of regex retries in iteration 513: 0 +[2026-03-26 05:08:05,720][__main__][INFO] - agents played in iteration 513 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:08:06,480][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:08:06,504][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:08:06,528][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:08:06,552][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:08:06,553][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:08:06,553][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:08:07,325][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:08:07,785][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:08:08,290][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:08:08,794][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:08:09,295][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:08:09,795][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:08:10,295][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:08:10,797][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:08:11,296][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:08:11,801][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:08:12,302][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:08:12,812][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:08:13,313][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:08:13,813][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:08:14,316][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:08:14,816][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:08:15,324][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:08:15,835][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:08:16,335][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:08:16,836][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:08:17,336][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:08:17,836][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:08:18,335][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:08:18,840][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:08:19,356][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:08:19,860][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:08:20,365][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:08:20,867][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:08:21,371][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:08:21,873][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:08:22,375][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:08:22,882][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:08:23,396][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:08:23,897][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:08:24,400][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:08:24,903][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:08:25,405][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:08:25,904][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:08:26,403][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:08:26,905][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:08:27,409][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:08:27,919][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:08:28,419][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:08:28,921][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:08:29,420][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:08:29,919][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:08:30,417][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:08:30,915][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:08:31,415][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:08:31,916][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:08:32,423][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:08:32,932][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:08:33,430][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:08:33,931][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:08:34,431][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:08:34,930][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:08:35,428][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:08:35,926][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:08:36,425][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:08:36,926][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:08:37,431][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:08:37,936][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:08:38,442][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:08:38,940][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:08:39,439][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10056 tokens. +[2026-03-26 05:08:40,261][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 60.51%, Block Peak % of device VRAM: 62.24%, ΔTime: 00:00:32 +[2026-03-26 05:08:41,022][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:08:41,024][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:08:41,026][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:08:41,751][__main__][INFO] - Iteration 514 took 50s (28.77% Gen, 69.79% Train). Generation: 14s, Training: 35s. Estimated remaining time: 34h 6m 25s. Estimated total time: 42h 9m 26s. Time estimates for 10 more iterations: 8m 25s, 100 more iterations: 1h 24m 18s, 500 more iterations: 7h 1m 34s. +[2026-03-26 05:08:41,754][__main__][INFO] - Starting iteration 514. +[2026-03-26 05:08:42,171][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 51 and human policies 1. +[2026-03-26 05:08:42,171][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:08:56,192][__main__][INFO] - Number of regex retries in iteration 514: 0 +[2026-03-26 05:08:56,193][__main__][INFO] - agents played in iteration 514 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:08:56,960][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:08:56,984][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:08:57,007][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:08:57,031][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:08:57,032][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:08:57,033][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:08:57,796][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:08:58,259][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:08:58,765][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:08:59,266][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:08:59,766][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:09:00,264][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:09:00,765][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:09:01,265][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:09:01,772][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:09:02,277][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:09:02,784][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:09:03,288][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:09:03,788][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:09:04,291][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:09:04,793][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:09:05,292][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:09:05,789][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:09:06,291][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:09:06,791][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:09:07,293][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:09:07,791][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:09:08,289][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:09:08,786][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:09:09,283][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:09:09,782][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:09:10,280][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:09:10,778][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:09:11,276][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:09:11,779][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:09:12,275][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:09:12,773][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:09:13,270][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:09:13,768][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:09:14,265][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:09:14,761][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:09:15,259][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:09:15,756][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:09:16,252][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:09:16,748][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:09:17,245][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:09:17,741][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:09:18,242][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:09:18,742][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:09:19,241][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:09:19,741][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:09:20,240][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:09:20,734][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:09:21,233][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:09:21,734][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:09:22,232][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:09:22,732][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:09:23,229][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:09:23,727][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:09:24,228][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:09:24,725][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:09:25,226][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:09:25,727][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:09:26,225][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:09:26,724][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:09:27,222][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:09:27,718][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:09:28,214][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:09:28,711][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:09:29,208][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:09:29,706][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10124 tokens. +[2026-03-26 05:09:30,492][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.58%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:32 +[2026-03-26 05:09:31,245][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:09:31,247][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:09:31,249][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:09:31,970][__main__][INFO] - Iteration 515 took 49s (28.16% Gen, 70.39% Train). Generation: 14s, Training: 35s. Estimated remaining time: 33h 26m 9s. Estimated total time: 41h 30m 1s. Time estimates for 10 more iterations: 8m 18s, 100 more iterations: 1h 23m 0s, 500 more iterations: 6h 55m 0s. +[2026-03-26 05:09:31,973][__main__][INFO] - Starting iteration 515. +[2026-03-26 05:09:32,386][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 51 and human policies 1. +[2026-03-26 05:09:32,387][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:09:47,217][__main__][INFO] - Number of regex retries in iteration 515: 0 +[2026-03-26 05:09:47,218][__main__][INFO] - agents played in iteration 515 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:09:47,990][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:09:48,015][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:09:48,039][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:09:48,063][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:09:48,064][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:09:48,065][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:09:48,830][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:09:49,289][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:09:49,793][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:09:50,300][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:09:50,814][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:09:51,315][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:09:51,818][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:09:52,319][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:09:52,820][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:09:53,322][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:09:53,830][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:09:54,336][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:09:54,836][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:09:55,338][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:09:55,843][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:09:56,343][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:09:56,842][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:09:57,341][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:09:57,848][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:09:58,351][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:09:58,851][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:09:59,351][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:09:59,850][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:10:00,349][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:10:00,848][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:10:01,348][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:10:01,855][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:10:02,368][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:10:02,875][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:10:03,375][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:10:03,876][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:10:04,376][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:10:04,876][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:10:05,379][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:10:05,884][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:10:06,387][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:10:06,893][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:10:07,399][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:10:07,904][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:10:08,405][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:10:08,908][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:10:09,410][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:10:09,912][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:10:10,415][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:10:10,913][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:10:11,412][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:10:11,911][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:10:12,413][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:10:12,921][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:10:13,430][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:10:13,930][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:10:14,433][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:10:14,932][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:10:15,436][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:10:15,940][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:10:16,438][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:10:16,937][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:10:17,436][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:10:17,935][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:10:18,434][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:10:18,937][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:10:19,441][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:10:19,948][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:10:20,453][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:10:20,955][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10161 tokens. +[2026-03-26 05:10:21,769][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 60.62%, Block Peak % of device VRAM: 62.24%, ΔTime: 00:00:32 +[2026-03-26 05:10:22,523][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:10:22,525][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:10:22,527][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:10:23,256][__main__][INFO] - Iteration 516 took 50s (29.16% Gen, 69.41% Train). Generation: 14s, Training: 35s. Estimated remaining time: 34h 18m 47s. Estimated total time: 42h 23m 30s. Time estimates for 10 more iterations: 8m 28s, 100 more iterations: 1h 24m 47s, 500 more iterations: 7h 3m 55s. +[2026-03-26 05:10:23,258][__main__][INFO] - Starting iteration 516. +[2026-03-26 05:10:23,674][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 51 and human policies 1. +[2026-03-26 05:10:23,675][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:10:40,052][__main__][INFO] - Number of regex retries in iteration 516: 0 +[2026-03-26 05:10:40,053][__main__][INFO] - agents played in iteration 516 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:10:40,818][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:10:40,841][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:10:40,865][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:10:40,888][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:10:40,889][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:10:40,889][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:10:41,644][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:10:42,097][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:10:42,600][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:10:43,099][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:10:43,603][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:10:44,119][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:10:44,625][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:10:45,126][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:10:45,627][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:10:46,135][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:10:46,651][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:10:47,156][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:10:47,686][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:10:48,190][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:10:48,694][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:10:49,196][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:10:49,701][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:10:50,201][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:10:50,703][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:10:51,202][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:10:51,704][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:10:52,206][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:10:52,703][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:10:53,203][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:10:53,702][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:10:54,200][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:10:54,702][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:10:55,203][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:10:55,700][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:10:56,197][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:10:56,694][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:10:57,194][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:10:57,690][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:10:58,186][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:10:58,682][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:10:59,179][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:10:59,675][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:11:00,170][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:11:00,670][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:11:01,169][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:11:01,665][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:11:02,165][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:11:02,666][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:11:03,164][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:11:03,665][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:11:04,162][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:11:04,660][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:11:05,161][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:11:05,658][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:11:06,154][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:11:06,650][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:11:07,145][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:11:07,642][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:11:08,140][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:11:08,640][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:11:09,144][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:11:09,656][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:11:10,162][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:11:10,664][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:11:11,165][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:11:11,665][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:11:12,165][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:11:12,669][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:11:13,172][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:11:13,672][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10163 tokens. +[2026-03-26 05:11:14,472][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 60.56%, Block Peak % of device VRAM: 62.28%, ΔTime: 00:00:32 +[2026-03-26 05:11:15,218][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:11:15,220][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:11:15,222][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:11:16,082][__main__][INFO] - Iteration 517 took 52s (31.25% Gen, 67.10% Train). Generation: 16s, Training: 35s. Estimated remaining time: 35h 34m 50s. Estimated total time: 43h 40m 26s. Time estimates for 10 more iterations: 8m 44s, 100 more iterations: 1h 27m 20s, 500 more iterations: 7h 16m 44s. +[2026-03-26 05:11:16,084][__main__][INFO] - Starting iteration 517. +[2026-03-26 05:11:16,502][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 51 and human policies 1. +[2026-03-26 05:11:16,502][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:11:31,305][__main__][INFO] - Number of regex retries in iteration 517: 0 +[2026-03-26 05:11:31,306][__main__][INFO] - agents played in iteration 517 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:11:32,079][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:11:32,104][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:11:32,128][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:11:32,151][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:11:32,152][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:11:32,153][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:11:32,921][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:11:33,383][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:11:33,887][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:11:34,400][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:11:34,910][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:11:35,411][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:11:35,916][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:11:36,422][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:11:36,923][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:11:37,427][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:11:37,927][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:11:38,431][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:11:38,935][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:11:39,441][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:11:39,943][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:11:40,445][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:11:40,949][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:11:41,449][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:11:41,948][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:11:42,447][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:11:42,947][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:11:43,451][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:11:43,955][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:11:44,459][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:11:44,960][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:11:45,460][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:11:45,960][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:11:46,458][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:11:46,960][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:11:47,461][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:11:47,960][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:11:48,460][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:11:48,964][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:11:49,469][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:11:49,975][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:11:50,478][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:11:50,982][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:11:51,482][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:11:51,983][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:11:52,482][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:11:52,981][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:11:53,479][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:11:53,977][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:11:54,476][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:11:54,979][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:11:55,480][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:11:55,992][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:11:56,492][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:11:56,991][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:11:57,495][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:11:57,996][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:11:58,498][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:11:59,000][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:11:59,502][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:12:00,003][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:12:00,503][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:12:01,004][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:12:01,505][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:12:02,011][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:12:02,509][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:12:03,008][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:12:03,506][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:12:04,009][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:12:04,508][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:12:05,009][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10191 tokens. +[2026-03-26 05:12:05,828][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 60.64%, Block Peak % of device VRAM: 62.20%, ΔTime: 00:00:32 +[2026-03-26 05:12:06,570][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:12:06,573][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:12:06,575][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:12:07,314][__main__][INFO] - Iteration 518 took 50s (29.13% Gen, 69.41% Train). Generation: 14s, Training: 35s. Estimated remaining time: 34h 14m 10s. Estimated total time: 42h 20m 37s. Time estimates for 10 more iterations: 8m 28s, 100 more iterations: 1h 24m 41s, 500 more iterations: 7h 3m 26s. +[2026-03-26 05:12:07,316][__main__][INFO] - Starting iteration 518. +[2026-03-26 05:12:07,726][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 51 and human policies 1. +[2026-03-26 05:12:07,727][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:12:23,084][__main__][INFO] - Number of regex retries in iteration 518: 0 +[2026-03-26 05:12:23,085][__main__][INFO] - agents played in iteration 518 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:12:23,851][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:12:23,876][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:12:23,901][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:12:23,926][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:12:23,927][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:12:23,927][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:12:24,707][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:12:25,163][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:12:25,668][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:12:26,167][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:12:26,670][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:12:27,168][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:12:27,667][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:12:28,167][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:12:28,673][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:12:29,180][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:12:29,681][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:12:30,185][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:12:30,685][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:12:31,185][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:12:31,685][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:12:32,186][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:12:32,685][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:12:33,186][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:12:33,685][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:12:34,184][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:12:34,683][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:12:35,182][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:12:35,681][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:12:36,180][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:12:36,680][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:12:37,183][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:12:37,688][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:12:38,197][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:12:38,701][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:12:39,201][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:12:39,702][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:12:40,203][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:12:40,706][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:12:41,207][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:12:41,705][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:12:42,204][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:12:42,704][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:12:43,202][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:12:43,701][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:12:44,199][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:12:44,696][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:12:45,194][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:12:45,693][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:12:46,194][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:12:46,694][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:12:47,195][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:12:47,692][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:12:48,193][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:12:48,689][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:12:49,190][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:12:49,690][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:12:50,188][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:12:50,688][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:12:51,188][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:12:51,687][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:12:52,192][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:12:52,692][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:12:53,192][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:12:53,688][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:12:54,186][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:12:54,685][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:12:55,182][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:12:55,680][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:12:56,176][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:12:56,673][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10093 tokens. +[2026-03-26 05:12:57,498][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 60.49%, Block Peak % of device VRAM: 62.17%, ΔTime: 00:00:32 +[2026-03-26 05:12:58,250][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:12:58,253][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:12:58,254][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:12:58,984][__main__][INFO] - Iteration 519 took 51s (29.96% Gen, 68.61% Train). Generation: 15s, Training: 35s. Estimated remaining time: 34h 35m 36s. Estimated total time: 42h 42m 55s. Time estimates for 10 more iterations: 8m 32s, 100 more iterations: 1h 25m 25s, 500 more iterations: 7h 7m 9s. +[2026-03-26 05:12:58,986][__main__][INFO] - Starting iteration 519. +[2026-03-26 05:12:59,385][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 51 and human policies 1. +[2026-03-26 05:12:59,386][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:13:04,612][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 05:13:04,696][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 05:13:15,757][__main__][INFO] - Number of regex retries in iteration 519: 2 +[2026-03-26 05:13:15,758][__main__][INFO] - agents played in iteration 519 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:13:16,530][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:13:16,555][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:13:16,580][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:13:16,605][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:13:16,606][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:13:16,606][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:13:17,381][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:13:17,842][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:13:18,346][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:13:18,852][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:13:19,352][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:13:19,851][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:13:20,349][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:13:20,849][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:13:21,350][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:13:21,848][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:13:22,348][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:13:22,848][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:13:23,347][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:13:23,845][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:13:24,344][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:13:24,844][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:13:25,343][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:13:25,852][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:13:26,354][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:13:26,859][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:13:27,361][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:13:27,862][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:13:28,367][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:13:28,867][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:13:29,366][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:13:29,866][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:13:30,365][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:13:30,867][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:13:31,365][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:13:31,868][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:13:32,367][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:13:32,864][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:13:33,363][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:13:33,884][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:13:34,392][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:13:34,892][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:13:35,393][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:13:35,895][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:13:36,403][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:13:36,908][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:13:37,410][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:13:37,906][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:13:38,403][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:13:38,901][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:13:39,397][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:13:39,894][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:13:40,390][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:13:40,887][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:13:41,383][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:13:41,882][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:13:42,381][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:13:42,883][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:13:43,382][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:13:43,882][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:13:44,381][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:13:44,877][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:13:45,379][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:13:45,878][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:13:46,374][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:13:46,874][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:13:47,370][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:13:47,867][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:13:48,368][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:13:48,867][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:13:49,365][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10244 tokens. +[2026-03-26 05:13:50,175][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.50%, ΔTime: 00:00:32 +[2026-03-26 05:13:50,928][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:13:50,930][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:13:50,932][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:13:51,649][__main__][INFO] - Iteration 520 took 52s (31.33% Gen, 67.30% Train). Generation: 16s, Training: 35s. Estimated remaining time: 35h 25m 1s. Estimated total time: 43h 33m 12s. Time estimates for 10 more iterations: 8m 42s, 100 more iterations: 1h 27m 6s, 500 more iterations: 7h 15m 32s. +[2026-03-26 05:13:51,651][__main__][INFO] - Starting iteration 520. +[2026-03-26 05:13:52,066][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 51 and human policies 1. +[2026-03-26 05:13:52,066][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:14:06,773][__main__][INFO] - Number of regex retries in iteration 520: 0 +[2026-03-26 05:14:06,774][__main__][INFO] - agents played in iteration 520 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:14:07,528][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:14:07,552][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:14:07,575][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:14:07,599][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:14:07,599][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:14:07,600][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:14:08,362][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:14:08,828][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:14:09,341][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:14:09,843][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:14:10,345][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:14:10,848][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:14:11,356][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:14:11,864][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:14:12,367][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:14:12,868][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:14:13,369][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:14:13,873][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:14:14,371][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:14:14,872][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:14:15,374][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:14:15,880][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:14:16,385][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:14:16,886][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:14:17,388][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:14:17,891][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:14:18,390][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:14:18,890][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:14:19,390][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:14:19,889][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:14:20,398][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:14:20,906][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:14:21,408][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:14:21,909][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:14:22,410][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:14:22,909][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:14:23,409][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:14:23,909][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:14:24,407][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:14:24,907][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:14:25,408][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:14:25,913][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:14:26,417][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:14:26,916][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:14:27,416][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:14:27,917][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:14:28,417][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:14:28,917][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:14:29,416][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:14:29,916][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:14:30,415][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:14:30,915][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:14:31,413][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:14:31,912][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:14:32,411][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:14:32,911][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:14:33,410][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:14:33,909][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:14:34,408][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:14:34,908][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:14:35,407][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:14:35,907][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:14:36,405][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:14:36,904][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:14:37,406][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:14:37,905][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:14:38,406][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:14:38,906][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:14:39,404][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:14:39,902][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:14:40,400][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10078 tokens. +[2026-03-26 05:14:41,193][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 60.51%, Block Peak % of device VRAM: 62.13%, ΔTime: 00:00:32 +[2026-03-26 05:14:41,957][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:14:41,960][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:14:41,961][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:14:43,283][__main__][INFO] - Iteration 521 took 51s (28.72% Gen, 68.70% Train). Generation: 14s, Training: 35s. Estimated remaining time: 34h 31m 49s. Estimated total time: 42h 40m 52s. Time estimates for 10 more iterations: 8m 32s, 100 more iterations: 1h 25m 21s, 500 more iterations: 7h 6m 48s. +[2026-03-26 05:14:43,285][__main__][INFO] - Starting iteration 521. +[2026-03-26 05:14:43,685][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 52 and human policies 1. +[2026-03-26 05:14:43,685][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:14:59,994][__main__][INFO] - Number of regex retries in iteration 521: 0 +[2026-03-26 05:14:59,995][__main__][INFO] - agents played in iteration 521 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:15:00,753][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:15:00,776][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:15:00,800][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:15:00,824][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:15:00,824][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:15:00,825][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:15:01,583][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:15:02,040][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:15:02,558][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:15:03,064][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:15:03,565][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:15:04,065][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:15:04,564][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:15:05,064][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:15:05,573][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:15:06,077][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:15:06,582][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:15:07,083][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:15:07,584][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:15:08,083][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:15:08,585][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:15:09,084][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:15:09,588][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:15:10,090][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:15:10,589][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:15:11,088][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:15:11,589][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:15:12,089][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:15:12,595][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:15:13,100][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:15:13,604][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:15:14,107][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:15:14,608][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:15:15,110][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:15:15,618][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:15:16,117][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:15:16,616][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:15:17,121][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:15:17,628][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:15:18,131][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:15:18,631][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:15:19,130][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:15:19,629][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:15:20,128][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:15:20,629][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:15:21,127][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:15:21,625][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:15:22,124][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:15:22,627][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:15:23,131][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:15:23,634][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:15:24,143][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:15:24,645][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:15:25,145][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:15:25,645][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:15:26,144][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:15:26,644][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:15:27,144][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:15:27,642][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:15:28,138][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:15:28,637][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:15:29,135][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:15:29,631][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:15:30,134][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:15:30,641][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:15:31,151][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:15:31,653][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:15:32,159][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:15:32,661][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:15:33,165][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:15:33,665][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10176 tokens. +[2026-03-26 05:15:34,491][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.20%, ΔTime: 00:00:32 +[2026-03-26 05:15:35,248][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:15:35,251][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:15:35,252][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:15:35,973][__main__][INFO] - Iteration 522 took 52s (31.19% Gen, 67.43% Train). Generation: 16s, Training: 35s. Estimated remaining time: 35h 24m 31s. Estimated total time: 43h 34m 26s. Time estimates for 10 more iterations: 8m 42s, 100 more iterations: 1h 27m 8s, 500 more iterations: 7h 15m 44s. +[2026-03-26 05:15:35,975][__main__][INFO] - Starting iteration 522. +[2026-03-26 05:15:36,375][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 52 and human policies 1. +[2026-03-26 05:15:36,376][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:15:51,254][__main__][INFO] - Number of regex retries in iteration 522: 0 +[2026-03-26 05:15:51,255][__main__][INFO] - agents played in iteration 522 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:15:52,019][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:15:52,043][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:15:52,068][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:15:52,092][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:15:52,093][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:15:52,094][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:15:52,875][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:15:53,341][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:15:53,849][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:15:54,350][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:15:54,851][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:15:55,352][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:15:55,852][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:15:56,352][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:15:56,856][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:15:57,357][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:15:57,856][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:15:58,359][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:15:58,859][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:15:59,357][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:15:59,862][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:16:00,370][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:16:00,881][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:16:01,379][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:16:01,879][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:16:02,378][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:16:02,876][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:16:03,377][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:16:03,876][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:16:04,373][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:16:04,873][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:16:05,371][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:16:05,868][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:16:06,365][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:16:06,862][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:16:07,358][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:16:07,855][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:16:08,353][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:16:08,852][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:16:09,350][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:16:09,850][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:16:10,352][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:16:10,853][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:16:11,351][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:16:11,849][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:16:12,347][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:16:12,848][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:16:13,349][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:16:13,849][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:16:14,349][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:16:14,850][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:16:15,346][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:16:15,844][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:16:16,341][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:16:16,838][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:16:17,336][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:16:17,831][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:16:18,327][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:16:18,824][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:16:19,321][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:16:19,821][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:16:20,317][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:16:20,812][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:16:21,309][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:16:21,807][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:16:22,305][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:16:22,804][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:16:23,301][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:16:23,798][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:16:24,298][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:16:24,794][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10151 tokens. +[2026-03-26 05:16:25,619][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.21%, ΔTime: 00:00:32 +[2026-03-26 05:16:26,375][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:16:26,377][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:16:26,379][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:16:27,101][__main__][INFO] - Iteration 523 took 50s (29.33% Gen, 69.24% Train). Generation: 14s, Training: 35s. Estimated remaining time: 34h 5m 31s. Estimated total time: 42h 16m 18s. Time estimates for 10 more iterations: 8m 27s, 100 more iterations: 1h 24m 32s, 500 more iterations: 7h 2m 43s. +[2026-03-26 05:16:27,104][__main__][INFO] - Starting iteration 523. +[2026-03-26 05:16:27,502][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 52 and human policies 1. +[2026-03-26 05:16:27,502][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:16:36,400][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 05:16:42,883][__main__][INFO] - Number of regex retries in iteration 523: 1 +[2026-03-26 05:16:42,883][__main__][INFO] - agents played in iteration 523 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:16:43,640][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:16:43,664][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:16:43,687][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:16:43,710][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:16:43,711][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:16:43,711][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:16:44,475][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:16:44,929][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:16:45,432][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:16:45,935][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:16:46,433][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:16:46,930][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:16:47,435][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:16:47,942][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:16:48,447][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:16:48,955][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:16:49,454][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:16:49,955][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:16:50,459][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:16:50,962][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:16:51,460][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:16:51,962][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:16:52,464][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:16:52,963][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:16:53,463][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:16:53,966][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:16:54,464][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:16:54,965][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:16:55,467][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:16:55,967][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:16:56,465][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:16:56,966][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:16:57,466][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:16:57,971][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:16:58,478][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:16:58,981][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:16:59,492][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:16:59,994][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:17:00,500][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:17:00,999][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:17:01,498][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:17:01,999][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:17:02,498][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:17:02,997][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:17:03,493][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:17:03,991][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:17:04,488][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:17:04,987][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:17:05,488][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:17:05,987][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:17:06,485][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:17:06,984][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:17:07,482][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:17:07,984][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:17:08,485][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:17:08,986][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:17:09,487][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:17:09,983][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:17:10,480][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:17:10,981][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:17:11,475][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:17:11,971][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:17:12,467][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:17:12,967][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:17:13,467][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:17:13,965][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:17:14,463][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:17:14,964][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:17:15,462][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:17:15,962][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:17:16,462][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10260 tokens. +[2026-03-26 05:17:17,262][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 60.64%, Block Peak % of device VRAM: 62.41%, ΔTime: 00:00:32 +[2026-03-26 05:17:17,999][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:17:18,001][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:17:18,003][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:17:18,835][__main__][INFO] - Iteration 524 took 51s (29.96% Gen, 68.41% Train). Generation: 15s, Training: 35s. Estimated remaining time: 34h 35m 5s. Estimated total time: 42h 46m 43s. Time estimates for 10 more iterations: 8m 33s, 100 more iterations: 1h 25m 33s, 500 more iterations: 7h 7m 47s. +[2026-03-26 05:17:18,838][__main__][INFO] - Starting iteration 524. +[2026-03-26 05:17:19,254][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 52 and human policies 1. +[2026-03-26 05:17:19,255][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:17:34,105][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 05:17:34,924][__main__][INFO] - Number of regex retries in iteration 524: 1 +[2026-03-26 05:17:34,924][__main__][INFO] - agents played in iteration 524 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:17:35,694][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:17:35,718][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:17:35,741][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:17:35,764][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:17:35,765][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:17:35,766][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:17:36,525][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:17:36,978][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:17:37,480][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:17:37,977][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:17:38,474][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:17:38,970][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:17:39,466][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:17:39,962][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:17:40,458][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:17:40,955][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:17:41,457][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:17:41,953][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:17:42,458][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:17:42,956][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:17:43,460][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:17:43,962][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:17:44,465][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:17:44,973][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:17:45,472][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:17:45,972][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:17:46,471][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:17:46,971][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:17:47,473][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:17:47,971][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:17:48,471][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:17:48,973][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:17:49,470][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:17:49,973][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:17:50,476][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:17:50,976][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:17:51,473][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:17:51,975][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:17:52,477][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:17:52,978][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:17:53,474][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:17:53,972][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:17:54,469][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:17:54,967][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:17:55,464][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:17:55,964][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:17:56,462][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:17:56,959][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:17:57,456][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:17:57,958][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:17:58,456][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:17:58,960][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:17:59,460][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:17:59,958][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:18:00,462][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:18:00,989][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:18:01,491][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:18:01,992][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:18:02,491][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:18:02,989][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:18:03,490][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:18:03,988][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:18:04,486][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:18:04,984][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:18:05,481][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:18:05,979][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:18:06,480][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:18:06,982][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:18:07,479][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:18:07,975][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:18:08,475][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10140 tokens. +[2026-03-26 05:18:09,272][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 60.61%, Block Peak % of device VRAM: 62.17%, ΔTime: 00:00:32 +[2026-03-26 05:18:10,019][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:18:10,021][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:18:10,023][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:18:10,747][__main__][INFO] - Iteration 525 took 51s (30.43% Gen, 68.16% Train). Generation: 15s, Training: 35s. Estimated remaining time: 34h 42m 9s. Estimated total time: 42h 54m 39s. Time estimates for 10 more iterations: 8m 34s, 100 more iterations: 1h 25m 49s, 500 more iterations: 7h 9m 6s. +[2026-03-26 05:18:10,749][__main__][INFO] - Starting iteration 525. +[2026-03-26 05:18:11,147][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 52 and human policies 1. +[2026-03-26 05:18:11,147][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:18:28,304][__main__][INFO] - Number of regex retries in iteration 525: 0 +[2026-03-26 05:18:28,305][__main__][INFO] - agents played in iteration 525 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:18:29,069][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:18:29,093][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:18:29,117][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:18:29,141][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:18:29,142][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:18:29,142][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:18:29,916][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:18:30,375][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:18:30,897][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:18:31,397][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:18:31,900][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:18:32,400][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:18:32,906][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:18:33,406][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:18:33,909][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:18:34,408][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:18:34,910][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:18:35,420][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:18:35,919][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:18:36,420][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:18:36,924][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:18:37,424][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:18:37,925][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:18:38,427][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:18:38,928][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:18:39,432][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:18:39,937][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:18:40,451][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:18:40,953][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:18:41,456][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:18:41,959][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:18:42,459][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:18:42,961][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:18:43,462][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:18:43,963][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:18:44,465][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:18:44,969][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:18:45,471][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:18:45,971][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:18:46,468][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:18:46,966][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:18:47,465][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:18:47,964][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:18:48,467][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:18:48,972][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:18:49,480][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:18:49,980][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:18:50,480][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:18:50,981][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:18:51,481][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:18:51,983][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:18:52,482][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:18:52,984][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:18:53,482][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:18:53,980][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:18:54,478][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:18:54,981][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:18:55,486][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:18:55,987][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:18:56,487][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:18:56,985][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:18:57,483][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:18:57,983][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:18:58,483][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:18:58,983][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:18:59,484][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:18:59,986][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:19:00,484][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:19:00,983][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:19:01,484][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:19:01,986][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10134 tokens. +[2026-03-26 05:19:02,800][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.57%, Block Peak % of device VRAM: 62.14%, ΔTime: 00:00:32 +[2026-03-26 05:19:03,559][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:19:03,561][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:19:03,563][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:19:04,278][__main__][INFO] - Iteration 526 took 53s (32.29% Gen, 66.36% Train). Generation: 17s, Training: 35s. Estimated remaining time: 36h 3m 12s. Estimated total time: 44h 16m 36s. Time estimates for 10 more iterations: 8m 51s, 100 more iterations: 1h 28m 33s, 500 more iterations: 7h 22m 46s. +[2026-03-26 05:19:04,281][__main__][INFO] - Starting iteration 526. +[2026-03-26 05:19:04,680][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 52 and human policies 1. +[2026-03-26 05:19:04,681][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:19:20,316][__main__][INFO] - Number of regex retries in iteration 526: 0 +[2026-03-26 05:19:20,317][__main__][INFO] - agents played in iteration 526 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:19:21,090][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:19:21,114][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:19:21,139][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:19:21,163][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:19:21,164][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:19:21,164][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:19:21,940][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:19:22,398][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:19:22,904][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:19:23,414][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:19:23,927][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:19:24,432][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:19:24,937][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:19:25,438][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:19:25,942][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:19:26,444][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:19:26,952][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:19:27,468][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:19:27,972][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:19:28,481][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:19:28,985][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:19:29,488][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:19:29,991][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:19:30,489][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:19:30,987][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:19:31,485][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:19:31,989][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:19:32,487][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:19:32,986][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:19:33,484][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:19:33,984][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:19:34,487][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:19:34,990][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:19:35,488][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:19:35,988][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:19:36,490][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:19:36,993][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:19:37,493][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:19:37,991][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:19:38,489][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:19:38,987][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:19:39,484][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:19:39,982][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:19:40,480][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:19:40,977][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:19:41,474][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:19:41,980][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:19:42,488][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:19:42,990][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:19:43,493][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:19:43,992][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:19:44,490][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:19:44,989][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:19:45,488][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:19:45,991][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:19:46,495][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:19:47,002][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:19:47,501][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:19:48,001][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:19:48,501][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:19:48,999][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:19:49,498][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:19:49,997][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:19:50,499][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:19:51,005][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:19:51,511][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:19:52,020][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:19:52,524][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:19:53,027][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:19:53,527][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:19:54,030][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10148 tokens. +[2026-03-26 05:19:54,841][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 +[2026-03-26 05:19:55,587][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:19:55,589][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:19:55,591][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:19:56,327][__main__][INFO] - Iteration 527 took 51s (30.27% Gen, 68.30% Train). Generation: 15s, Training: 35s. Estimated remaining time: 34h 48m 6s. Estimated total time: 43h 2m 22s. Time estimates for 10 more iterations: 8m 36s, 100 more iterations: 1h 26m 4s, 500 more iterations: 7h 10m 23s. +[2026-03-26 05:19:56,329][__main__][INFO] - Starting iteration 527. +[2026-03-26 05:19:56,747][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 52 and human policies 1. +[2026-03-26 05:19:56,748][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:20:10,942][__main__][INFO] - Number of regex retries in iteration 527: 0 +[2026-03-26 05:20:10,943][__main__][INFO] - agents played in iteration 527 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:20:11,714][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:20:11,740][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:20:11,764][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:20:11,790][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:20:11,790][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:20:11,791][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:20:12,575][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:20:13,031][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:20:13,536][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:20:14,037][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:20:14,541][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:20:15,051][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:20:15,553][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:20:16,054][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:20:16,557][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:20:17,061][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:20:17,564][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:20:18,065][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:20:18,565][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:20:19,070][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:20:19,574][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:20:20,087][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:20:20,588][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:20:21,088][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:20:21,586][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:20:22,084][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:20:22,581][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:20:23,081][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:20:23,578][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:20:24,075][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:20:24,572][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:20:25,069][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:20:25,567][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:20:26,067][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:20:26,565][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:20:27,062][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:20:27,560][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:20:28,058][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:20:28,555][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:20:29,056][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:20:29,554][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:20:30,055][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:20:30,553][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:20:31,051][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:20:31,549][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:20:32,052][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:20:32,550][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:20:33,052][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:20:33,549][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:20:34,047][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:20:34,546][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:20:35,055][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:20:35,565][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:20:36,072][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:20:36,576][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:20:37,077][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:20:37,577][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:20:38,099][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:20:38,599][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:20:39,098][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:20:39,597][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:20:40,096][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:20:40,598][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:20:41,106][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:20:41,613][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:20:42,114][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:20:42,615][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:20:43,117][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:20:43,617][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:20:44,117][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:20:44,619][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10054 tokens. +[2026-03-26 05:20:45,436][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 60.57%, Block Peak % of device VRAM: 62.09%, ΔTime: 00:00:32 +[2026-03-26 05:20:46,196][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:20:46,198][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:20:46,199][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:20:46,934][__main__][INFO] - Iteration 528 took 50s (28.28% Gen, 70.25% Train). Generation: 14s, Training: 35s. Estimated remaining time: 33h 34m 16s. Estimated total time: 41h 49m 23s. Time estimates for 10 more iterations: 8m 21s, 100 more iterations: 1h 23m 38s, 500 more iterations: 6h 58m 13s. +[2026-03-26 05:20:46,937][__main__][INFO] - Starting iteration 528. +[2026-03-26 05:20:47,352][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 52 and human policies 1. +[2026-03-26 05:20:47,353][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:21:01,574][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 05:21:03,230][__main__][INFO] - Number of regex retries in iteration 528: 1 +[2026-03-26 05:21:03,231][__main__][INFO] - agents played in iteration 528 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:21:03,999][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:21:04,023][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:21:04,047][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:21:04,070][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:21:04,071][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:21:04,071][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:21:04,847][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:21:05,309][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:21:05,818][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:21:06,327][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:21:06,833][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:21:07,332][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:21:07,837][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:21:08,346][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:21:08,852][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:21:09,357][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:21:09,883][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:21:10,392][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:21:10,895][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:21:11,396][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:21:11,899][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:21:12,401][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:21:12,902][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:21:13,401][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:21:13,899][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:21:14,394][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:21:14,895][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:21:15,391][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:21:15,890][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:21:16,388][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:21:16,885][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:21:17,382][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:21:17,883][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:21:18,380][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:21:18,877][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:21:19,375][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:21:19,872][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:21:20,369][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:21:20,878][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:21:21,386][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:21:21,891][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:21:22,397][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:21:22,897][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:21:23,397][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:21:23,896][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:21:24,399][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:21:24,901][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:21:25,404][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:21:25,904][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:21:26,404][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:21:26,903][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:21:27,403][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:21:27,904][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:21:28,413][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:21:28,922][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:21:29,424][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:21:29,926][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:21:30,428][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:21:30,929][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:21:31,429][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:21:31,932][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:21:32,436][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:21:32,939][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:21:33,446][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:21:33,952][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:21:34,452][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:21:34,953][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:21:35,453][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:21:35,953][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:21:36,457][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:21:36,956][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10225 tokens. +[2026-03-26 05:21:37,764][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.57%, Block Peak % of device VRAM: 62.38%, ΔTime: 00:00:32 +[2026-03-26 05:21:38,520][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:21:38,523][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:21:38,526][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:21:39,294][__main__][INFO] - Iteration 529 took 51s (30.57% Gen, 67.95% Train). Generation: 15s, Training: 35s. Estimated remaining time: 35h 1m 8s. Estimated total time: 43h 17m 7s. Time estimates for 10 more iterations: 8m 39s, 100 more iterations: 1h 26m 34s, 500 more iterations: 7h 12m 51s. +[2026-03-26 05:21:39,297][__main__][INFO] - Starting iteration 529. +[2026-03-26 05:21:39,704][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 52 and human policies 1. +[2026-03-26 05:21:39,704][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:21:55,338][__main__][INFO] - Number of regex retries in iteration 529: 0 +[2026-03-26 05:21:55,339][__main__][INFO] - agents played in iteration 529 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:21:56,110][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:21:56,135][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:21:56,159][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:21:56,183][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:21:56,183][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:21:56,184][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:21:56,961][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:21:57,421][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:21:57,933][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:21:58,434][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:21:58,938][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:21:59,439][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:21:59,948][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:22:00,458][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:22:00,960][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:22:01,460][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:22:01,961][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:22:02,461][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:22:02,963][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:22:03,468][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:22:03,974][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:22:04,474][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:22:04,974][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:22:05,475][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:22:05,976][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:22:06,477][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:22:06,981][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:22:07,492][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:22:07,996][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:22:08,502][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:22:09,004][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:22:09,507][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:22:10,006][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:22:10,507][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:22:11,004][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:22:11,505][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:22:12,007][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:22:12,511][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:22:13,012][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:22:13,513][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:22:14,013][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:22:14,511][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:22:15,010][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:22:15,509][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:22:16,009][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:22:16,509][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:22:17,010][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:22:17,518][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:22:18,023][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:22:18,526][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:22:19,027][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:22:19,532][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:22:20,031][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:22:20,532][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:22:21,034][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:22:21,536][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:22:22,035][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:22:22,537][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:22:23,040][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:22:23,544][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:22:24,050][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:22:24,553][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:22:25,059][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:22:25,560][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:22:26,062][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:22:26,565][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:22:27,064][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:22:27,566][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:22:28,066][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:22:28,566][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:22:29,068][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10137 tokens. +[2026-03-26 05:22:29,893][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 60.56%, Block Peak % of device VRAM: 62.23%, ΔTime: 00:00:32 +[2026-03-26 05:22:30,703][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:22:30,705][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:22:30,707][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:22:31,536][__main__][INFO] - Iteration 530 took 51s (30.16% Gen, 68.23% Train). Generation: 15s, Training: 35s. Estimated remaining time: 34h 54m 46s. Estimated total time: 43h 11m 37s. Time estimates for 10 more iterations: 8m 38s, 100 more iterations: 1h 26m 23s, 500 more iterations: 7h 11m 56s. +[2026-03-26 05:22:31,538][__main__][INFO] - Starting iteration 530. +[2026-03-26 05:22:31,960][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 52 and human policies 1. +[2026-03-26 05:22:31,961][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:22:39,563][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 05:22:46,803][__main__][INFO] - Number of regex retries in iteration 530: 1 +[2026-03-26 05:22:46,804][__main__][INFO] - agents played in iteration 530 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:22:47,560][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:22:47,584][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:22:47,608][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:22:47,632][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:22:47,632][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:22:47,633][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:22:48,409][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:22:48,874][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:22:49,383][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:22:49,882][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:22:50,388][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:22:50,898][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:22:51,398][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:22:51,898][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:22:52,405][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:22:52,906][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:22:53,408][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:22:53,909][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:22:54,409][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:22:54,910][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:22:55,409][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:22:55,909][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:22:56,409][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:22:56,914][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:22:57,411][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:22:57,914][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:22:58,411][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:22:58,910][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:22:59,409][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:22:59,907][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:23:00,409][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:23:00,918][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:23:01,422][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:23:01,933][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:23:02,434][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:23:02,934][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:23:03,434][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:23:03,937][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:23:04,436][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:23:04,935][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:23:05,439][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:23:05,937][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:23:06,440][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:23:06,938][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:23:07,436][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:23:07,937][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:23:08,435][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:23:08,933][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:23:09,431][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:23:09,932][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:23:10,429][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:23:10,930][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:23:11,427][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:23:11,925][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:23:12,424][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:23:12,924][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:23:13,424][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:23:13,924][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:23:14,422][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:23:14,920][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:23:15,419][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:23:15,918][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:23:16,423][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:23:16,921][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:23:17,419][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:23:17,916][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:23:18,415][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:23:18,913][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:23:19,411][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:23:19,910][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:23:20,407][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10117 tokens. +[2026-03-26 05:23:21,220][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.05%, Current % of VRAM taken: 60.48%, Block Peak % of device VRAM: 62.04%, ΔTime: 00:00:32 +[2026-03-26 05:23:21,993][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:23:21,995][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:23:21,997][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:23:23,338][__main__][INFO] - Iteration 531 took 51s (28.89% Gen, 68.50% Train). Generation: 14s, Training: 35s. Estimated remaining time: 34h 31m 12s. Estimated total time: 42h 48m 55s. Time estimates for 10 more iterations: 8m 33s, 100 more iterations: 1h 25m 37s, 500 more iterations: 7h 8m 9s. +[2026-03-26 05:23:23,340][__main__][INFO] - Starting iteration 531. +[2026-03-26 05:23:23,741][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 53 and human policies 1. +[2026-03-26 05:23:23,742][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:23:37,613][__main__][INFO] - Number of regex retries in iteration 531: 0 +[2026-03-26 05:23:37,614][__main__][INFO] - agents played in iteration 531 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:23:38,394][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:23:38,418][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:23:38,442][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:23:38,465][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:23:38,466][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:23:38,466][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:23:39,224][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:23:39,683][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:23:40,185][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:23:40,687][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:23:41,196][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:23:41,702][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:23:42,212][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:23:42,712][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:23:43,211][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:23:43,716][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:23:44,219][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:23:44,720][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:23:45,223][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:23:45,726][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:23:46,229][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:23:46,731][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:23:47,230][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:23:47,728][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:23:48,226][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:23:48,724][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:23:49,221][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:23:49,717][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:23:50,214][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:23:50,711][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:23:51,211][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:23:51,710][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:23:52,207][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:23:52,705][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:23:53,208][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:23:53,707][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:23:54,213][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:23:54,713][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:23:55,219][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:23:55,720][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:23:56,225][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:23:56,734][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:23:57,236][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:23:57,737][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:23:58,239][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:23:58,739][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:23:59,238][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:23:59,736][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:24:00,233][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:24:00,734][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:24:01,234][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:24:01,732][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:24:02,232][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:24:02,734][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:24:03,230][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:24:03,728][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:24:04,226][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:24:04,725][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:24:05,224][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:24:05,722][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:24:06,222][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:24:06,722][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:24:07,221][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:24:07,718][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:24:08,215][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:24:08,712][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:24:09,210][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:24:09,707][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:24:10,205][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:24:10,702][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:24:11,200][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10122 tokens. +[2026-03-26 05:24:12,011][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.05%, Current % of VRAM taken: 60.48%, Block Peak % of device VRAM: 62.23%, ΔTime: 00:00:32 +[2026-03-26 05:24:12,766][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:24:12,768][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:24:12,770][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:24:13,503][__main__][INFO] - Iteration 532 took 49s (27.88% Gen, 70.65% Train). Generation: 13s, Training: 35s. Estimated remaining time: 33h 9m 35s. Estimated total time: 41h 28m 8s. Time estimates for 10 more iterations: 8m 17s, 100 more iterations: 1h 22m 56s, 500 more iterations: 6h 54m 41s. +[2026-03-26 05:24:13,506][__main__][INFO] - Starting iteration 532. +[2026-03-26 05:24:13,928][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 53 and human policies 1. +[2026-03-26 05:24:13,929][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:24:28,972][__main__][INFO] - Number of regex retries in iteration 532: 0 +[2026-03-26 05:24:28,973][__main__][INFO] - agents played in iteration 532 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:24:29,723][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:24:29,746][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:24:29,770][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:24:29,793][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:24:29,793][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:24:29,794][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:24:30,558][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:24:31,024][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:24:31,533][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:24:32,036][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:24:32,537][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:24:33,037][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:24:33,545][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:24:34,053][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:24:34,554][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:24:35,055][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:24:35,555][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:24:36,061][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:24:36,559][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:24:37,059][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:24:37,558][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:24:38,061][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:24:38,559][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:24:39,059][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:24:39,559][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:24:40,057][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:24:40,560][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:24:41,058][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:24:41,558][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:24:42,059][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:24:42,576][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:24:43,076][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:24:43,577][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:24:44,076][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:24:44,575][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:24:45,077][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:24:45,575][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:24:46,081][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:24:46,592][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:24:47,091][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:24:47,592][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:24:48,089][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:24:48,588][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:24:49,090][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:24:49,588][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:24:50,087][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:24:50,589][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:24:51,093][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:24:51,598][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:24:52,097][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:24:52,598][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:24:53,096][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:24:53,594][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:24:54,093][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:24:54,591][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:24:55,090][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:24:55,590][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:24:56,098][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:24:56,609][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:24:57,110][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:24:57,611][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:24:58,111][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:24:58,611][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:24:59,114][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:24:59,614][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:25:00,139][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:25:00,643][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:25:01,144][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:25:01,648][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:25:02,157][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:25:02,667][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10045 tokens. +[2026-03-26 05:25:03,509][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 60.65%, Block Peak % of device VRAM: 62.15%, ΔTime: 00:00:32 +[2026-03-26 05:25:04,275][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:25:04,277][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:25:04,279][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:25:05,002][__main__][INFO] - Iteration 533 took 51s (29.46% Gen, 69.13% Train). Generation: 15s, Training: 35s. Estimated remaining time: 34h 14m 17s. Estimated total time: 42h 33m 41s. Time estimates for 10 more iterations: 8m 30s, 100 more iterations: 1h 25m 7s, 500 more iterations: 7h 5m 36s. +[2026-03-26 05:25:05,004][__main__][INFO] - Starting iteration 533. +[2026-03-26 05:25:05,404][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 53 and human policies 1. +[2026-03-26 05:25:05,405][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:25:20,487][__main__][INFO] - Number of regex retries in iteration 533: 0 +[2026-03-26 05:25:20,488][__main__][INFO] - agents played in iteration 533 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:25:21,259][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:25:21,283][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:25:21,310][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:25:21,334][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:25:21,335][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:25:21,335][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:25:22,135][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:25:22,597][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:25:23,109][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:25:23,613][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:25:24,115][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:25:24,615][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:25:25,121][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:25:25,637][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:25:26,141][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:25:26,644][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:25:27,144][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:25:27,643][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:25:28,143][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:25:28,642][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:25:29,140][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:25:29,639][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:25:30,137][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:25:30,637][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:25:31,138][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:25:31,640][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:25:32,140][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:25:32,648][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:25:33,147][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:25:33,649][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:25:34,148][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:25:34,646][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:25:35,144][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:25:35,643][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:25:36,145][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:25:36,649][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:25:37,148][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:25:37,647][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:25:38,151][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:25:38,655][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:25:39,186][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:25:39,695][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:25:40,202][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:25:40,704][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:25:41,207][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:25:41,708][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:25:42,210][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:25:42,714][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:25:43,217][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:25:43,721][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:25:44,221][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:25:44,722][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:25:45,220][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:25:45,719][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:25:46,218][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:25:46,718][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:25:47,224][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:25:47,727][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:25:48,237][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:25:48,747][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:25:49,249][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:25:49,756][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:25:50,264][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:25:50,763][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:25:51,268][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:25:51,768][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:25:52,268][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:25:52,772][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:25:53,271][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:25:53,778][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:25:54,282][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10169 tokens. +[2026-03-26 05:25:55,119][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 60.56%, Block Peak % of device VRAM: 62.24%, ΔTime: 00:00:32 +[2026-03-26 05:25:55,869][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:25:55,872][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:25:55,874][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:25:56,596][__main__][INFO] - Iteration 534 took 51s (29.46% Gen, 69.12% Train). Generation: 15s, Training: 35s. Estimated remaining time: 34h 19m 22s. Estimated total time: 42h 39m 39s. Time estimates for 10 more iterations: 8m 31s, 100 more iterations: 1h 25m 19s, 500 more iterations: 7h 6m 36s. +[2026-03-26 05:25:56,599][__main__][INFO] - Starting iteration 534. +[2026-03-26 05:25:57,006][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 53 and human policies 1. +[2026-03-26 05:25:57,007][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:26:12,826][__main__][INFO] - Number of regex retries in iteration 534: 0 +[2026-03-26 05:26:12,827][__main__][INFO] - agents played in iteration 534 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:26:13,594][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:26:13,618][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:26:13,642][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:26:13,665][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:26:13,666][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:26:13,666][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:26:14,452][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:26:14,904][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:26:15,413][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:26:15,919][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:26:16,420][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:26:16,924][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:26:17,430][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:26:17,934][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:26:18,435][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:26:18,937][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:26:19,436][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:26:19,936][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:26:20,435][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:26:20,932][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:26:21,430][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:26:21,929][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:26:22,426][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:26:22,928][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:26:23,426][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:26:23,927][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:26:24,427][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:26:24,925][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:26:25,426][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:26:25,927][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:26:26,428][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:26:26,930][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:26:27,430][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:26:27,935][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:26:28,440][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:26:28,941][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:26:29,441][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:26:29,942][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:26:30,443][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:26:30,949][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:26:31,451][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:26:31,956][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:26:32,458][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:26:32,961][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:26:33,465][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:26:33,966][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:26:34,469][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:26:34,966][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:26:35,465][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:26:35,962][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:26:36,461][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:26:36,958][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:26:37,456][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:26:37,953][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:26:38,450][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:26:38,947][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:26:39,443][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:26:39,939][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:26:40,436][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:26:40,937][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:26:41,434][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:26:41,937][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:26:42,432][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:26:42,930][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:26:43,428][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:26:43,926][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:26:44,425][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:26:44,923][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:26:45,421][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:26:45,922][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:26:46,423][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10120 tokens. +[2026-03-26 05:26:47,246][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 60.62%, Block Peak % of device VRAM: 62.40%, ΔTime: 00:00:32 +[2026-03-26 05:26:47,994][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:26:47,996][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:26:47,998][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:26:48,746][__main__][INFO] - Iteration 535 took 51s (30.58% Gen, 67.97% Train). Generation: 15s, Training: 35s. Estimated remaining time: 34h 45m 53s. Estimated total time: 43h 7m 1s. Time estimates for 10 more iterations: 8m 37s, 100 more iterations: 1h 26m 14s, 500 more iterations: 7h 11m 10s. +[2026-03-26 05:26:48,748][__main__][INFO] - Starting iteration 535. +[2026-03-26 05:26:49,170][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 53 and human policies 1. +[2026-03-26 05:26:49,170][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:27:05,682][__main__][INFO] - Number of regex retries in iteration 535: 0 +[2026-03-26 05:27:05,683][__main__][INFO] - agents played in iteration 535 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:27:06,477][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:27:06,501][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:27:06,525][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:27:06,549][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:27:06,550][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:27:06,551][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:27:07,331][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:27:07,791][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:27:08,304][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:27:08,802][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:27:09,310][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:27:09,816][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:27:10,318][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:27:10,822][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:27:11,322][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:27:11,822][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:27:12,323][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:27:12,827][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:27:13,345][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:27:13,846][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:27:14,349][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:27:14,856][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:27:15,356][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:27:15,857][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:27:16,356][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:27:16,856][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:27:17,357][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:27:17,856][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:27:18,355][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:27:18,854][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:27:19,353][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:27:19,852][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:27:20,352][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:27:20,853][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:27:21,352][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:27:21,851][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:27:22,352][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:27:22,860][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:27:23,371][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:27:23,872][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:27:24,375][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:27:24,874][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:27:25,374][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:27:25,876][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:27:26,377][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:27:26,879][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:27:27,377][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:27:27,880][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:27:28,399][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:27:28,897][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:27:29,395][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:27:29,893][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:27:30,401][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:27:30,912][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:27:31,413][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:27:31,916][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:27:32,422][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:27:32,925][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:27:33,426][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:27:33,928][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:27:34,429][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:27:34,932][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:27:35,448][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:27:35,952][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:27:36,454][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:27:36,958][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:27:37,459][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:27:37,964][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:27:38,465][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:27:38,967][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:27:39,468][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10152 tokens. +[2026-03-26 05:27:40,299][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 60.60%, Block Peak % of device VRAM: 62.22%, ΔTime: 00:00:32 +[2026-03-26 05:27:41,058][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:27:41,060][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:27:41,062][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:27:41,786][__main__][INFO] - Iteration 536 took 52s (31.38% Gen, 67.24% Train). Generation: 16s, Training: 35s. Estimated remaining time: 35h 28m 51s. Estimated total time: 43h 50m 52s. Time estimates for 10 more iterations: 8m 46s, 100 more iterations: 1h 27m 41s, 500 more iterations: 7h 18m 28s. +[2026-03-26 05:27:41,789][__main__][INFO] - Starting iteration 536. +[2026-03-26 05:27:42,192][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 53 and human policies 1. +[2026-03-26 05:27:42,192][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:27:58,156][__main__][INFO] - Number of regex retries in iteration 536: 0 +[2026-03-26 05:27:58,157][__main__][INFO] - agents played in iteration 536 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:27:58,926][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:27:58,951][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:27:58,976][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:27:59,001][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:27:59,001][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:27:59,002][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:27:59,783][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:28:00,245][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:28:00,757][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:28:01,258][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:28:01,760][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:28:02,260][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:28:02,760][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:28:03,259][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:28:03,765][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:28:04,266][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:28:04,771][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:28:05,273][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:28:05,778][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:28:06,283][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:28:06,793][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:28:07,300][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:28:07,806][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:28:08,309][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:28:08,809][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:28:09,311][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:28:09,813][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:28:10,314][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:28:10,813][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:28:11,311][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:28:11,811][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:28:12,309][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:28:12,809][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:28:13,306][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:28:13,804][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:28:14,303][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:28:14,800][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:28:15,299][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:28:15,795][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:28:16,293][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:28:16,796][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:28:17,300][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:28:17,812][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:28:18,313][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:28:18,813][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:28:19,318][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:28:19,818][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:28:20,344][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:28:20,842][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:28:21,343][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:28:21,847][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:28:22,345][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:28:22,845][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:28:23,346][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:28:23,848][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:28:24,355][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:28:24,862][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:28:25,363][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:28:25,864][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:28:26,369][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:28:26,869][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:28:27,370][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:28:27,869][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:28:28,369][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:28:28,868][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:28:29,366][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:28:29,865][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:28:30,365][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:28:30,864][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:28:31,363][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:28:31,863][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10225 tokens. +[2026-03-26 05:28:32,681][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.15%, ΔTime: 00:00:32 +[2026-03-26 05:28:33,430][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:28:33,432][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:28:33,434][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:28:34,286][__main__][INFO] - Iteration 537 took 52s (30.65% Gen, 67.72% Train). Generation: 15s, Training: 35s. Estimated remaining time: 35h 1m 51s. Estimated total time: 43h 24m 45s. Time estimates for 10 more iterations: 8m 40s, 100 more iterations: 1h 26m 49s, 500 more iterations: 7h 14m 7s. +[2026-03-26 05:28:34,288][__main__][INFO] - Starting iteration 537. +[2026-03-26 05:28:34,689][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 53 and human policies 1. +[2026-03-26 05:28:34,690][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:28:49,961][__main__][INFO] - Number of regex retries in iteration 537: 0 +[2026-03-26 05:28:49,962][__main__][INFO] - agents played in iteration 537 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:28:50,718][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:28:50,742][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:28:50,765][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:28:50,789][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:28:50,789][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:28:50,790][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:28:51,552][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:28:52,007][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:28:52,515][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:28:53,020][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:28:53,525][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:28:54,027][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:28:54,528][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:28:55,034][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:28:55,533][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:28:56,033][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:28:56,534][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:28:57,047][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:28:57,554][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:28:58,057][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:28:58,558][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:28:59,060][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:28:59,559][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:29:00,060][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:29:00,560][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:29:01,060][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:29:01,565][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:29:02,080][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:29:02,581][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:29:03,085][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:29:03,586][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:29:04,090][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:29:04,591][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:29:05,094][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:29:05,596][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:29:06,099][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:29:06,597][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:29:07,097][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:29:07,599][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:29:08,096][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:29:08,595][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:29:09,094][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:29:09,596][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:29:10,102][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:29:10,613][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:29:11,114][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:29:11,615][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:29:12,115][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:29:12,614][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:29:13,114][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:29:13,614][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:29:14,112][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:29:14,611][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:29:15,111][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:29:15,610][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:29:16,116][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:29:16,620][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:29:17,135][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:29:17,638][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:29:18,142][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:29:18,645][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:29:19,149][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:29:19,651][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:29:20,152][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:29:20,651][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:29:21,150][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:29:21,651][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:29:22,149][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:29:22,651][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:29:23,150][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:29:23,652][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10214 tokens. +[2026-03-26 05:29:24,459][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 60.62%, Block Peak % of device VRAM: 62.18%, ΔTime: 00:00:32 +[2026-03-26 05:29:25,269][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:29:25,271][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:29:25,273][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:29:26,019][__main__][INFO] - Iteration 538 took 51s (29.75% Gen, 68.79% Train). Generation: 15s, Training: 35s. Estimated remaining time: 34h 22m 45s. Estimated total time: 42h 46m 31s. Time estimates for 10 more iterations: 8m 33s, 100 more iterations: 1h 25m 33s, 500 more iterations: 7h 7m 45s. +[2026-03-26 05:29:26,021][__main__][INFO] - Starting iteration 538. +[2026-03-26 05:29:26,441][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 53 and human policies 1. +[2026-03-26 05:29:26,442][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:29:42,077][__main__][INFO] - Number of regex retries in iteration 538: 0 +[2026-03-26 05:29:42,077][__main__][INFO] - agents played in iteration 538 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:29:42,862][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:29:42,890][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:29:42,917][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:29:42,943][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:29:42,944][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:29:42,944][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:29:43,720][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:29:44,180][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:29:44,689][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:29:45,203][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:29:45,720][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:29:46,222][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:29:46,727][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:29:47,228][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:29:47,729][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:29:48,228][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:29:48,733][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:29:49,238][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:29:49,743][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:29:50,246][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:29:50,747][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:29:51,247][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:29:51,746][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:29:52,245][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:29:52,744][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:29:53,253][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:29:53,759][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:29:54,265][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:29:54,767][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:29:55,268][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:29:55,767][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:29:56,271][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:29:56,770][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:29:57,273][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:29:57,772][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:29:58,275][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:29:58,788][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:29:59,289][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:29:59,791][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:30:00,294][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:30:00,797][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:30:01,303][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:30:01,808][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:30:02,329][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:30:02,834][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:30:03,336][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:30:03,844][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:30:04,352][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:30:04,859][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:30:05,360][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:30:05,861][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:30:06,364][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:30:06,866][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:30:07,366][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:30:07,865][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:30:08,365][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:30:08,866][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:30:09,370][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:30:09,870][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:30:10,373][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:30:10,877][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:30:11,378][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:30:11,884][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:30:12,391][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:30:12,891][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:30:13,391][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:30:13,891][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:30:14,390][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:30:14,890][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:30:15,391][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:30:15,890][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10066 tokens. +[2026-03-26 05:30:16,704][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 60.49%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:32 +[2026-03-26 05:30:17,459][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:30:17,461][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:30:17,463][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:30:18,185][__main__][INFO] - Iteration 539 took 51s (30.22% Gen, 68.38% Train). Generation: 15s, Training: 35s. Estimated remaining time: 34h 42m 37s. Estimated total time: 43h 7m 14s. Time estimates for 10 more iterations: 8m 37s, 100 more iterations: 1h 26m 14s, 500 more iterations: 7h 11m 12s. +[2026-03-26 05:30:18,188][__main__][INFO] - Starting iteration 539. +[2026-03-26 05:30:18,615][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 53 and human policies 1. +[2026-03-26 05:30:18,615][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:30:32,319][__main__][INFO] - Number of regex retries in iteration 539: 0 +[2026-03-26 05:30:32,320][__main__][INFO] - agents played in iteration 539 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:30:33,101][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:30:33,125][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:30:33,149][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:30:33,172][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:30:33,173][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:30:33,173][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:30:33,937][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:30:34,399][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:30:34,912][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:30:35,421][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:30:35,922][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:30:36,424][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:30:36,929][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:30:37,445][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:30:37,948][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:30:38,450][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:30:38,951][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:30:39,451][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:30:39,952][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:30:40,456][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:30:40,955][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:30:41,455][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:30:41,954][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:30:42,453][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:30:42,956][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:30:43,464][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:30:43,978][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:30:44,484][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:30:44,986][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:30:45,489][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:30:45,990][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:30:46,492][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:30:46,993][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:30:47,492][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:30:47,994][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:30:48,492][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:30:48,994][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:30:49,492][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:30:49,990][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:30:50,489][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:30:50,988][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:30:51,486][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:30:51,987][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:30:52,485][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:30:52,986][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:30:53,484][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:30:53,981][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:30:54,479][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:30:54,975][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:30:55,472][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:30:55,970][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:30:56,470][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:30:56,967][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:30:57,464][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:30:57,963][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:30:58,460][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:30:58,960][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:30:59,455][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:30:59,952][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:31:00,449][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:31:00,946][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:31:01,442][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:31:01,944][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:31:02,445][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:31:02,941][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:31:03,436][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:31:03,932][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:31:04,431][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:31:04,928][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:31:05,424][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:31:05,921][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10148 tokens. +[2026-03-26 05:31:06,719][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 60.48%, Block Peak % of device VRAM: 62.18%, ΔTime: 00:00:32 +[2026-03-26 05:31:07,478][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:31:07,480][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:31:07,481][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:31:08,245][__main__][INFO] - Iteration 540 took 49s (27.61% Gen, 70.85% Train). Generation: 13s, Training: 35s. Estimated remaining time: 32h 56m 5s. Estimated total time: 41h 21m 33s. Time estimates for 10 more iterations: 8m 16s, 100 more iterations: 1h 22m 43s, 500 more iterations: 6h 53m 35s. +[2026-03-26 05:31:08,248][__main__][INFO] - Starting iteration 540. +[2026-03-26 05:31:08,659][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 53 and human policies 1. +[2026-03-26 05:31:08,660][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:31:13,342][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 05:31:23,634][__main__][INFO] - Number of regex retries in iteration 540: 1 +[2026-03-26 05:31:23,635][__main__][INFO] - agents played in iteration 540 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:31:24,426][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:31:24,450][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:31:24,475][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:31:24,499][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:31:24,500][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:31:24,500][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:31:25,269][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:31:25,736][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:31:26,240][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:31:26,738][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:31:27,238][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:31:27,736][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:31:28,239][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:31:28,744][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:31:29,254][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:31:29,754][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:31:30,255][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:31:30,755][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:31:31,254][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:31:31,754][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:31:32,255][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:31:32,753][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:31:33,252][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:31:33,752][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:31:34,249][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:31:34,749][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:31:35,250][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:31:35,754][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:31:36,262][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:31:36,772][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:31:37,272][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:31:37,773][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:31:38,274][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:31:38,773][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:31:39,274][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:31:39,774][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:31:40,273][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:31:40,775][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:31:41,274][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:31:41,772][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:31:42,273][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:31:42,770][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:31:43,267][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:31:43,765][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:31:44,261][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:31:44,757][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:31:45,255][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:31:45,754][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:31:46,255][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:31:46,754][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:31:47,254][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:31:47,753][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:31:48,252][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:31:48,750][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:31:49,251][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:31:49,747][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:31:50,244][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:31:50,740][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:31:51,237][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:31:51,733][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:31:52,230][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:31:52,728][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:31:53,226][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:31:53,725][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:31:54,224][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:31:54,746][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:31:55,243][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:31:55,740][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:31:56,237][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:31:56,733][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:31:57,231][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10015 tokens. +[2026-03-26 05:31:58,023][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 60.56%, Block Peak % of device VRAM: 62.00%, ΔTime: 00:00:32 +[2026-03-26 05:31:58,762][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:31:58,764][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:31:58,766][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:32:00,102][__main__][INFO] - Iteration 541 took 51s (29.11% Gen, 68.29% Train). Generation: 14s, Training: 35s. Estimated remaining time: 34h 25m 49s. Estimated total time: 42h 52m 9s. Time estimates for 10 more iterations: 8m 34s, 100 more iterations: 1h 25m 44s, 500 more iterations: 7h 8m 41s. +[2026-03-26 05:32:00,104][__main__][INFO] - Starting iteration 541. +[2026-03-26 05:32:00,525][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 54 and human policies 1. +[2026-03-26 05:32:00,525][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:32:13,895][__main__][INFO] - Number of regex retries in iteration 541: 0 +[2026-03-26 05:32:13,896][__main__][INFO] - agents played in iteration 541 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:32:14,657][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:32:14,681][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:32:14,706][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:32:14,731][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:32:14,732][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:32:14,732][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:32:15,494][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:32:15,949][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:32:16,455][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:32:16,954][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:32:17,452][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:32:17,951][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:32:18,450][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:32:18,948][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:32:19,448][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:32:19,947][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:32:20,445][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:32:20,943][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:32:21,441][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:32:21,940][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:32:22,437][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:32:22,942][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:32:23,445][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:32:23,947][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:32:24,453][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:32:24,955][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:32:25,463][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:32:25,964][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:32:26,468][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:32:26,968][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:32:27,472][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:32:27,968][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:32:28,465][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:32:28,961][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:32:29,458][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:32:29,954][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:32:30,452][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:32:30,950][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:32:31,446][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:32:31,950][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:32:32,448][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:32:32,946][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:32:33,445][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:32:33,942][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:32:34,439][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:32:34,939][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:32:35,435][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:32:35,934][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:32:36,433][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:32:36,930][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:32:37,428][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:32:37,929][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:32:38,427][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:32:38,926][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:32:39,426][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:32:39,925][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:32:40,425][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:32:40,922][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:32:41,420][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:32:41,916][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:32:42,414][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:32:42,910][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:32:43,409][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:32:43,906][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:32:44,404][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:32:44,901][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:32:45,396][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:32:45,892][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:32:46,388][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:32:46,884][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:32:47,382][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9995 tokens. +[2026-03-26 05:32:48,186][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.01%, Current % of VRAM taken: 60.44%, Block Peak % of device VRAM: 62.16%, ΔTime: 00:00:32 +[2026-03-26 05:32:48,930][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:32:48,932][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:32:48,934][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:32:49,655][__main__][INFO] - Iteration 542 took 49s (27.21% Gen, 71.32% Train). Generation: 13s, Training: 35s. Estimated remaining time: 32h 29m 24s. Estimated total time: 40h 56m 33s. Time estimates for 10 more iterations: 8m 11s, 100 more iterations: 1h 21m 53s, 500 more iterations: 6h 49m 25s. +[2026-03-26 05:32:49,658][__main__][INFO] - Starting iteration 542. +[2026-03-26 05:32:50,064][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 54 and human policies 1. +[2026-03-26 05:32:50,064][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:32:54,383][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 05:33:03,984][__main__][INFO] - Number of regex retries in iteration 542: 1 +[2026-03-26 05:33:03,985][__main__][INFO] - agents played in iteration 542 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:33:04,757][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.45%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:33:04,781][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.45%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:33:04,805][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.45%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:33:04,828][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.45%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:33:04,829][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:33:04,829][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:33:05,596][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:33:06,062][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:33:06,576][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:33:07,078][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:33:07,578][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:33:08,082][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:33:08,595][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:33:09,096][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:33:09,597][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:33:10,096][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:33:10,595][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:33:11,097][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:33:11,600][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:33:12,106][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:33:12,608][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:33:13,109][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:33:13,610][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:33:14,117][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:33:14,617][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:33:15,116][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:33:15,623][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:33:16,136][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:33:16,637][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:33:17,138][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:33:17,638][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:33:18,137][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:33:18,636][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:33:19,136][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:33:19,639][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:33:20,144][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:33:20,651][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:33:21,152][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:33:21,653][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:33:22,153][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:33:22,653][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:33:23,153][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:33:23,652][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:33:24,153][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:33:24,652][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:33:25,156][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:33:25,657][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:33:26,160][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:33:26,665][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:33:27,168][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:33:27,667][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:33:28,170][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:33:28,686][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:33:29,189][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:33:29,691][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:33:30,190][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:33:30,689][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:33:31,191][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:33:31,688][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:33:32,188][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:33:32,687][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:33:33,184][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:33:33,685][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:33:34,183][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:33:34,681][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:33:35,183][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:33:35,688][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:33:36,191][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:33:36,696][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:33:37,196][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:33:37,697][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10019 tokens. +[2026-03-26 05:33:38,525][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 60.48%, Block Peak % of device VRAM: 62.43%, ΔTime: 00:00:32 +[2026-03-26 05:33:39,276][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:33:39,278][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:33:39,280][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:33:40,002][__main__][INFO] - Iteration 543 took 49s (27.87% Gen, 70.68% Train). Generation: 13s, Training: 35s. Estimated remaining time: 33h 8m 57s. Estimated total time: 41h 36m 57s. Time estimates for 10 more iterations: 8m 19s, 100 more iterations: 1h 23m 13s, 500 more iterations: 6h 56m 9s. +[2026-03-26 05:33:40,005][__main__][INFO] - Starting iteration 543. +[2026-03-26 05:33:40,414][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 54 and human policies 1. +[2026-03-26 05:33:40,414][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:33:48,271][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 05:33:53,531][__main__][INFO] - Number of regex retries in iteration 543: 1 +[2026-03-26 05:33:53,532][__main__][INFO] - agents played in iteration 543 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:33:54,299][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:33:54,323][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:33:54,347][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:33:54,371][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:33:54,371][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:33:54,372][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:33:55,147][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:33:55,610][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:33:56,116][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:33:56,616][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:33:57,126][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:33:57,626][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:33:58,127][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:33:58,627][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:33:59,127][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:33:59,627][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:34:00,130][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:34:00,631][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:34:01,132][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:34:01,637][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:34:02,148][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:34:02,653][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:34:03,163][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:34:03,667][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:34:04,170][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:34:04,670][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:34:05,170][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:34:05,670][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:34:06,169][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:34:06,669][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:34:07,172][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:34:07,673][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:34:08,171][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:34:08,670][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:34:09,169][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:34:09,668][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:34:10,167][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:34:10,665][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:34:11,163][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:34:11,661][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:34:12,160][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:34:12,658][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:34:13,157][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:34:13,655][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:34:14,153][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:34:14,651][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:34:15,149][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:34:15,649][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:34:16,152][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:34:16,653][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:34:17,159][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:34:17,659][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:34:18,169][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:34:18,672][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:34:19,172][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:34:19,671][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:34:20,173][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:34:20,672][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:34:21,169][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:34:21,666][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:34:22,164][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:34:22,663][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:34:23,160][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:34:23,656][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:34:24,152][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:34:24,651][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:34:25,147][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:34:25,646][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:34:26,143][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:34:26,642][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:34:27,137][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10121 tokens. +[2026-03-26 05:34:27,937][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.03%, Current % of VRAM taken: 60.46%, Block Peak % of device VRAM: 62.33%, ΔTime: 00:00:32 +[2026-03-26 05:34:28,688][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:34:28,690][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:34:28,692][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:34:29,420][__main__][INFO] - Iteration 544 took 49s (26.77% Gen, 71.74% Train). Generation: 13s, Training: 35s. Estimated remaining time: 32h 21m 30s. Estimated total time: 40h 50m 19s. Time estimates for 10 more iterations: 8m 10s, 100 more iterations: 1h 21m 40s, 500 more iterations: 6h 48m 23s. +[2026-03-26 05:34:29,422][__main__][INFO] - Starting iteration 544. +[2026-03-26 05:34:29,827][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 54 and human policies 1. +[2026-03-26 05:34:29,828][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:34:40,229][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 05:34:45,967][__main__][INFO] - Number of regex retries in iteration 544: 1 +[2026-03-26 05:34:45,968][__main__][INFO] - agents played in iteration 544 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:34:46,743][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:34:46,767][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:34:46,790][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:34:46,814][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:34:46,815][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:34:46,815][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:34:47,586][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:34:48,039][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:34:48,545][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:34:49,046][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:34:49,546][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:34:50,044][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:34:50,552][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:34:51,052][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:34:51,553][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:34:52,051][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:34:52,550][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:34:53,051][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:34:53,548][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:34:54,067][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:34:54,568][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:34:55,066][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:34:55,567][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:34:56,064][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:34:56,560][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:34:57,057][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:34:57,555][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:34:58,056][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:34:58,553][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:34:59,050][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:34:59,551][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:35:00,050][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:35:00,549][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:35:01,051][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:35:01,550][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:35:02,052][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:35:02,556][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:35:03,060][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:35:03,569][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:35:04,069][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:35:04,569][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:35:05,073][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:35:05,572][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:35:06,072][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:35:06,572][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:35:07,071][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:35:07,571][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:35:08,073][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:35:08,576][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:35:09,074][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:35:09,573][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:35:10,073][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:35:10,573][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:35:11,072][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:35:11,571][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:35:12,071][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:35:12,573][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:35:13,073][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:35:13,571][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:35:14,069][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:35:14,570][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:35:15,071][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:35:15,569][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:35:16,066][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:35:16,564][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:35:17,061][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:35:17,558][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:35:18,056][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:35:18,554][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:35:19,051][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:35:19,548][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10130 tokens. +[2026-03-26 05:35:20,364][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.09%, Current % of VRAM taken: 60.51%, Block Peak % of device VRAM: 62.12%, ΔTime: 00:00:32 +[2026-03-26 05:35:21,114][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:35:21,116][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:35:21,118][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:35:21,835][__main__][INFO] - Iteration 545 took 52s (31.03% Gen, 67.59% Train). Generation: 16s, Training: 35s. Estimated remaining time: 34h 50m 43s. Estimated total time: 43h 20m 25s. Time estimates for 10 more iterations: 8m 40s, 100 more iterations: 1h 26m 40s, 500 more iterations: 7h 13m 24s. +[2026-03-26 05:35:21,837][__main__][INFO] - Starting iteration 545. +[2026-03-26 05:35:22,245][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 54 and human policies 1. +[2026-03-26 05:35:22,246][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:35:29,334][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 05:35:36,241][__main__][INFO] - Number of regex retries in iteration 545: 1 +[2026-03-26 05:35:36,242][__main__][INFO] - agents played in iteration 545 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:35:37,020][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:35:37,044][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:35:37,068][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:35:37,092][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:35:37,092][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:35:37,093][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:35:37,852][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:35:38,305][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:35:38,811][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:35:39,310][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:35:39,808][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:35:40,307][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:35:40,805][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:35:41,302][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:35:41,799][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:35:42,299][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:35:42,797][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:35:43,300][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:35:43,800][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:35:44,298][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:35:44,796][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:35:45,294][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:35:45,794][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:35:46,291][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:35:46,787][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:35:47,284][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:35:47,781][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:35:48,278][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:35:48,774][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:35:49,271][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:35:49,768][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:35:50,271][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:35:50,771][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:35:51,271][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:35:51,777][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:35:52,273][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:35:52,770][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:35:53,270][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:35:53,770][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:35:54,266][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:35:54,762][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:35:55,257][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:35:55,753][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:35:56,249][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:35:56,744][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:35:57,239][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:35:57,734][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:35:58,233][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:35:58,730][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:35:59,227][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:35:59,723][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:36:00,218][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:36:00,714][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:36:01,211][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:36:01,708][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:36:02,208][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:36:02,711][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:36:03,211][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:36:03,710][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:36:04,212][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:36:04,718][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:36:05,217][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:36:05,721][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:36:06,233][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:36:06,737][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:36:07,239][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:36:07,739][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:36:08,238][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:36:08,738][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:36:09,236][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:36:09,736][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10071 tokens. +[2026-03-26 05:36:10,545][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.35%, ΔTime: 00:00:32 +[2026-03-26 05:36:11,301][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:36:11,303][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:36:11,305][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:36:12,027][__main__][INFO] - Iteration 546 took 49s (28.11% Gen, 70.43% Train). Generation: 13s, Training: 35s. Estimated remaining time: 32h 58m 36s. Estimated total time: 41h 29m 8s. Time estimates for 10 more iterations: 8m 17s, 100 more iterations: 1h 22m 58s, 500 more iterations: 6h 54m 51s. +[2026-03-26 05:36:12,029][__main__][INFO] - Starting iteration 546. +[2026-03-26 05:36:12,454][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 54 and human policies 1. +[2026-03-26 05:36:12,455][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:36:26,438][__main__][INFO] - Number of regex retries in iteration 546: 0 +[2026-03-26 05:36:26,438][__main__][INFO] - agents played in iteration 546 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:36:27,203][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:36:27,227][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:36:27,250][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:36:27,274][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:36:27,274][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:36:27,275][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:36:28,041][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:36:28,493][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:36:28,995][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:36:29,492][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:36:29,990][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:36:30,488][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:36:30,987][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:36:31,485][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:36:31,985][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:36:32,488][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:36:32,987][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:36:33,487][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:36:33,986][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:36:34,485][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:36:34,985][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:36:35,482][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:36:35,982][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:36:36,479][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:36:36,975][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:36:37,472][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:36:37,970][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:36:38,469][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:36:38,973][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:36:39,472][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:36:39,973][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:36:40,471][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:36:40,973][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:36:41,471][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:36:41,971][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:36:42,469][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:36:42,969][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:36:43,473][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:36:43,972][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:36:44,470][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:36:44,967][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:36:45,467][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:36:45,966][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:36:46,464][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:36:46,962][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:36:47,459][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:36:47,957][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:36:48,456][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:36:48,955][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:36:49,453][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:36:49,952][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:36:50,452][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:36:50,951][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:36:51,454][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:36:51,953][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:36:52,451][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:36:52,950][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:36:53,450][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:36:53,948][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:36:54,446][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:36:54,944][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:36:55,445][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:36:55,946][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:36:56,443][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:36:56,941][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:36:57,438][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:36:57,936][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:36:58,435][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:36:58,933][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:36:59,432][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:36:59,933][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9964 tokens. +[2026-03-26 05:37:00,744][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 61.94%, ΔTime: 00:00:32 +[2026-03-26 05:37:01,514][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:37:01,516][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:37:01,518][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:37:02,215][__main__][INFO] - Iteration 547 took 49s (28.10% Gen, 70.49% Train). Generation: 13s, Training: 35s. Estimated remaining time: 32h 56m 43s. Estimated total time: 41h 28m 5s. Time estimates for 10 more iterations: 8m 17s, 100 more iterations: 1h 22m 56s, 500 more iterations: 6h 54m 40s. +[2026-03-26 05:37:02,218][__main__][INFO] - Starting iteration 547. +[2026-03-26 05:37:02,637][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 54 and human policies 1. +[2026-03-26 05:37:02,638][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:37:17,454][__main__][INFO] - Number of regex retries in iteration 547: 0 +[2026-03-26 05:37:17,455][__main__][INFO] - agents played in iteration 547 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:37:18,212][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:37:18,236][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:37:18,259][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:37:18,283][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:37:18,283][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:37:18,284][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:37:19,045][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:37:19,505][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:37:20,015][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:37:20,514][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:37:21,013][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:37:21,511][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:37:22,013][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:37:22,512][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:37:23,012][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:37:23,514][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:37:24,011][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:37:24,508][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:37:25,009][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:37:25,509][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:37:26,008][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:37:26,510][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:37:27,009][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:37:27,509][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:37:28,007][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:37:28,505][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:37:29,003][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:37:29,502][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:37:29,999][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:37:30,497][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:37:30,995][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:37:31,493][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:37:31,993][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:37:32,497][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:37:32,996][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:37:33,494][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:37:33,993][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:37:34,495][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:37:34,992][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:37:35,490][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:37:35,988][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:37:36,490][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:37:36,991][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:37:37,489][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:37:37,992][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:37:38,489][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:37:38,988][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:37:39,487][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:37:39,985][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:37:40,484][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:37:40,983][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:37:41,483][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:37:41,984][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:37:42,482][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:37:42,981][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:37:43,478][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:37:43,974][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:37:44,471][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:37:44,969][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:37:45,465][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:37:45,965][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:37:46,462][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:37:46,959][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:37:47,458][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:37:47,955][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:37:48,453][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:37:48,950][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:37:49,449][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:37:49,952][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:37:50,449][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:37:50,948][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10046 tokens. +[2026-03-26 05:37:51,748][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.58%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:32 +[2026-03-26 05:37:52,477][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:37:52,479][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:37:52,481][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:37:53,208][__main__][INFO] - Iteration 548 took 50s (29.30% Gen, 69.26% Train). Generation: 14s, Training: 35s. Estimated remaining time: 33h 36m 19s. Estimated total time: 42h 8m 32s. Time estimates for 10 more iterations: 8m 25s, 100 more iterations: 1h 24m 17s, 500 more iterations: 7h 1m 25s. +[2026-03-26 05:37:53,210][__main__][INFO] - Starting iteration 548. +[2026-03-26 05:37:53,620][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 54 and human policies 1. +[2026-03-26 05:37:53,621][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:38:07,485][__main__][INFO] - Number of regex retries in iteration 548: 0 +[2026-03-26 05:38:07,486][__main__][INFO] - agents played in iteration 548 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:38:08,243][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:38:08,267][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:38:08,290][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:38:08,314][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:38:08,314][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:38:08,315][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:38:09,073][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:38:09,530][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:38:10,038][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:38:10,534][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:38:11,034][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:38:11,531][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:38:12,028][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:38:12,526][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:38:13,023][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:38:13,518][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:38:14,013][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:38:14,509][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:38:15,005][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:38:15,500][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:38:15,995][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:38:16,490][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:38:16,986][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:38:17,483][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:38:17,980][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:38:18,479][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:38:18,974][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:38:19,471][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:38:19,972][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:38:20,468][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:38:20,967][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:38:21,465][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:38:21,961][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:38:22,456][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:38:22,956][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:38:23,451][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:38:23,947][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:38:24,447][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:38:24,942][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:38:25,444][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:38:25,940][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:38:26,442][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:38:26,941][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:38:27,439][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:38:27,940][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:38:28,437][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:38:28,934][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:38:29,443][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:38:29,941][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:38:30,440][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:38:30,939][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:38:31,437][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:38:31,941][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:38:32,456][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:38:32,955][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:38:33,454][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:38:33,953][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:38:34,452][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:38:34,952][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:38:35,451][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:38:35,950][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:38:36,449][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:38:36,950][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:38:37,449][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:38:37,947][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:38:38,445][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:38:38,943][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:38:39,441][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:38:39,938][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:38:40,435][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:38:40,933][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10067 tokens. +[2026-03-26 05:38:41,722][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 60.52%, Block Peak % of device VRAM: 62.11%, ΔTime: 00:00:32 +[2026-03-26 05:38:42,472][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:38:42,474][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:38:42,475][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:38:43,200][__main__][INFO] - Iteration 549 took 49s (27.96% Gen, 70.57% Train). Generation: 13s, Training: 34s. Estimated remaining time: 32h 45m 58s. Estimated total time: 41h 19m 1s. Time estimates for 10 more iterations: 8m 15s, 100 more iterations: 1h 22m 38s, 500 more iterations: 6h 53m 10s. +[2026-03-26 05:38:43,202][__main__][INFO] - Starting iteration 549. +[2026-03-26 05:38:43,618][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 54 and human policies 1. +[2026-03-26 05:38:43,618][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:38:59,039][__main__][INFO] - Number of regex retries in iteration 549: 0 +[2026-03-26 05:38:59,041][__main__][INFO] - agents played in iteration 549 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:38:59,836][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:38:59,860][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:38:59,884][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:38:59,907][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:38:59,907][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:38:59,908][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:39:00,668][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:39:01,129][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:39:01,631][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:39:02,134][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:39:02,633][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:39:03,131][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:39:03,627][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:39:04,129][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:39:04,627][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:39:05,126][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:39:05,625][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:39:06,125][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:39:06,626][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:39:07,134][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:39:07,632][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:39:08,130][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:39:08,632][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:39:09,134][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:39:09,634][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:39:10,133][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:39:10,634][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:39:11,134][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:39:11,664][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:39:12,164][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:39:12,672][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:39:13,169][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:39:13,671][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:39:14,171][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:39:14,668][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:39:15,169][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:39:15,667][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:39:16,164][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:39:16,666][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:39:17,166][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:39:17,664][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:39:18,161][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:39:18,659][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:39:19,156][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:39:19,654][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:39:20,153][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:39:20,653][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:39:21,150][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:39:21,646][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:39:22,144][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:39:22,644][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:39:23,140][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:39:23,639][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:39:24,134][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:39:24,634][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:39:25,134][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:39:25,631][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:39:26,128][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:39:26,628][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:39:27,130][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:39:27,630][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:39:28,130][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:39:28,630][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:39:29,133][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:39:29,636][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:39:30,138][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:39:30,640][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:39:31,142][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:39:31,643][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:39:32,145][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:39:32,650][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10273 tokens. +[2026-03-26 05:39:33,452][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.33%, ΔTime: 00:00:32 +[2026-03-26 05:39:34,245][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:39:34,247][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:39:34,248][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:39:34,962][__main__][INFO] - Iteration 550 took 51s (30.04% Gen, 68.57% Train). Generation: 15s, Training: 35s. Estimated remaining time: 34h 13m 21s. Estimated total time: 42h 47m 16s. Time estimates for 10 more iterations: 8m 33s, 100 more iterations: 1h 25m 34s, 500 more iterations: 7h 7m 52s. +[2026-03-26 05:39:34,965][__main__][INFO] - Starting iteration 550. +[2026-03-26 05:39:35,377][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 54 and human policies 1. +[2026-03-26 05:39:35,377][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:39:48,935][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 05:39:49,882][__main__][INFO] - Number of regex retries in iteration 550: 1 +[2026-03-26 05:39:49,883][__main__][INFO] - agents played in iteration 550 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:39:50,641][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:39:50,664][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:39:50,688][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:39:50,712][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:39:50,713][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:39:50,713][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:39:51,482][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:39:51,939][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:39:52,440][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:39:52,937][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:39:53,434][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:39:53,933][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:39:54,436][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:39:54,933][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:39:55,433][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:39:55,932][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:39:56,430][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:39:56,931][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:39:57,431][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:39:57,929][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:39:58,428][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:39:58,927][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:39:59,425][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:39:59,932][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:40:00,432][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:40:00,934][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:40:01,435][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:40:01,937][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:40:02,437][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:40:02,936][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:40:03,436][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:40:03,933][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:40:04,430][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:40:04,928][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:40:05,426][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:40:05,924][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:40:06,420][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:40:06,917][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:40:07,413][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:40:07,912][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:40:08,411][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:40:08,910][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:40:09,409][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:40:09,908][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:40:10,406][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:40:10,904][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:40:11,402][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:40:11,900][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:40:12,397][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:40:12,895][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:40:13,394][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:40:13,892][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:40:14,393][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:40:14,890][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:40:15,388][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:40:15,886][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:40:16,384][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:40:16,882][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:40:17,379][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:40:17,877][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:40:18,375][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:40:18,872][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:40:19,369][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:40:19,865][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:40:20,362][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:40:20,861][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:40:21,357][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:40:21,853][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:40:22,351][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:40:22,849][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:40:23,346][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10035 tokens. +[2026-03-26 05:40:24,155][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.01%, Current % of VRAM taken: 60.44%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:32 +[2026-03-26 05:40:24,939][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:40:24,941][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:40:24,942][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:40:26,262][__main__][INFO] - Iteration 551 took 50s (28.50% Gen, 68.90% Train). Generation: 14s, Training: 35s. Estimated remaining time: 33h 49m 32s. Estimated total time: 42h 24m 17s. Time estimates for 10 more iterations: 8m 28s, 100 more iterations: 1h 24m 48s, 500 more iterations: 7h 4m 2s. +[2026-03-26 05:40:26,264][__main__][INFO] - Starting iteration 551. +[2026-03-26 05:40:26,684][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 55 and human policies 1. +[2026-03-26 05:40:26,685][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:40:40,741][__main__][INFO] - Number of regex retries in iteration 551: 0 +[2026-03-26 05:40:40,742][__main__][INFO] - agents played in iteration 551 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:40:41,523][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.44%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:40:41,547][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.44%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:40:41,570][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.44%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:40:41,593][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.44%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:40:41,594][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:40:41,594][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:40:42,354][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:40:42,806][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:40:43,307][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:40:43,803][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:40:44,302][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:40:44,799][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:40:45,294][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:40:45,791][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:40:46,288][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:40:46,785][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:40:47,283][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:40:47,779][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:40:48,274][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:40:48,770][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:40:49,266][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:40:49,762][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:40:50,259][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:40:50,760][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:40:51,257][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:40:51,754][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:40:52,253][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:40:52,752][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:40:53,250][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:40:53,748][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:40:54,251][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:40:54,748][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:40:55,247][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:40:55,748][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:40:56,246][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:40:56,743][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:40:57,242][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:40:57,743][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:40:58,240][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:40:58,736][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:40:59,235][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:40:59,733][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:41:00,231][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:41:00,729][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:41:01,230][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:41:01,730][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:41:02,231][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:41:02,735][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:41:03,260][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:41:03,761][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:41:04,260][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:41:04,758][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:41:05,257][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:41:05,756][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:41:06,255][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:41:06,752][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:41:07,251][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:41:07,750][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:41:08,249][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:41:08,747][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:41:09,248][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:41:09,746][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:41:10,246][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:41:10,743][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:41:11,244][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:41:11,744][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:41:12,241][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:41:12,743][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:41:13,240][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:41:13,737][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:41:14,236][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10073 tokens. +[2026-03-26 05:41:15,035][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 +[2026-03-26 05:41:15,789][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:41:15,792][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:41:15,793][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:41:16,457][__main__][INFO] - Iteration 552 took 49s (28.24% Gen, 70.42% Train). Generation: 14s, Training: 35s. Estimated remaining time: 32h 53m 2s. Estimated total time: 41h 28m 38s. Time estimates for 10 more iterations: 8m 17s, 100 more iterations: 1h 22m 57s, 500 more iterations: 6h 54m 46s. +[2026-03-26 05:41:16,459][__main__][INFO] - Starting iteration 552. +[2026-03-26 05:41:16,861][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 55 and human policies 1. +[2026-03-26 05:41:16,861][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:41:30,498][__main__][INFO] - Number of regex retries in iteration 552: 0 +[2026-03-26 05:41:30,499][__main__][INFO] - agents played in iteration 552 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:41:31,277][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:41:31,301][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:41:31,325][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:41:31,348][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:41:31,349][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:41:31,349][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:41:32,118][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:41:32,574][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:41:33,076][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:41:33,573][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:41:34,072][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:41:34,571][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:41:35,069][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:41:35,568][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:41:36,067][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:41:36,565][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:41:37,061][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:41:37,558][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:41:38,054][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:41:38,554][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:41:39,052][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:41:39,552][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:41:40,050][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:41:40,548][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:41:41,046][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:41:41,544][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:41:42,043][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:41:42,544][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:41:43,042][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:41:43,540][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:41:44,039][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:41:44,537][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:41:45,037][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:41:45,536][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:41:46,035][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:41:46,533][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:41:47,032][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:41:47,532][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:41:48,033][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:41:48,531][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:41:49,029][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:41:49,528][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:41:50,032][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:41:50,530][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:41:51,029][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:41:51,529][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:41:52,027][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:41:52,526][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:41:53,024][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:41:53,521][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:41:54,019][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:41:54,520][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:41:55,016][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:41:55,514][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:41:56,012][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:41:56,510][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:41:57,007][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:41:57,506][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:41:58,005][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:41:58,502][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:41:59,001][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:41:59,497][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:41:59,994][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:42:00,491][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:42:00,988][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:42:01,485][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:42:01,983][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:42:02,479][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:42:02,975][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:42:03,474][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:42:03,971][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10056 tokens. +[2026-03-26 05:42:04,763][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 60.50%, Block Peak % of device VRAM: 62.15%, ΔTime: 00:00:32 +[2026-03-26 05:42:05,510][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:42:05,512][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:42:05,514][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:42:06,189][__main__][INFO] - Iteration 553 took 49s (27.65% Gen, 70.98% Train). Generation: 13s, Training: 35s. Estimated remaining time: 32h 30m 1s. Estimated total time: 41h 6m 27s. Time estimates for 10 more iterations: 8m 13s, 100 more iterations: 1h 22m 12s, 500 more iterations: 6h 51m 4s. +[2026-03-26 05:42:06,191][__main__][INFO] - Starting iteration 553. +[2026-03-26 05:42:06,597][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 55 and human policies 1. +[2026-03-26 05:42:06,598][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:42:11,408][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 05:42:20,967][__main__][INFO] - Number of regex retries in iteration 553: 1 +[2026-03-26 05:42:20,967][__main__][INFO] - agents played in iteration 553 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:42:21,762][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:42:21,785][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:42:21,809][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:42:21,832][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:42:21,833][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:42:21,833][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:42:22,595][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:42:23,056][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:42:23,561][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:42:24,060][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:42:24,558][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:42:25,057][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:42:25,556][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:42:26,055][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:42:26,553][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:42:27,052][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:42:27,555][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:42:28,056][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:42:28,554][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:42:29,053][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:42:29,553][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:42:30,053][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:42:30,554][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:42:31,051][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:42:31,551][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:42:32,049][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:42:32,547][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:42:33,045][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:42:33,542][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:42:34,042][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:42:34,538][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:42:35,035][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:42:35,532][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:42:36,030][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:42:36,528][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:42:37,028][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:42:37,526][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:42:38,024][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:42:38,522][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:42:39,017][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:42:39,513][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:42:40,009][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:42:40,506][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:42:41,001][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:42:41,496][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:42:41,992][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:42:42,488][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:42:42,986][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:42:43,483][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:42:43,980][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:42:44,477][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:42:44,974][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:42:45,471][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:42:45,968][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:42:46,465][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:42:46,961][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:42:47,455][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:42:47,952][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:42:48,448][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:42:48,948][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:42:49,444][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:42:49,941][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:42:50,438][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:42:50,934][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:42:51,430][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:42:51,927][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:42:52,425][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:42:52,920][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:42:53,416][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:42:53,912][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:42:54,411][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10011 tokens. +[2026-03-26 05:42:55,203][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 60.62%, Block Peak % of device VRAM: 62.41%, ΔTime: 00:00:32 +[2026-03-26 05:42:55,954][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:42:55,956][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:42:55,958][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:42:56,617][__main__][INFO] - Iteration 554 took 50s (28.73% Gen, 69.95% Train). Generation: 14s, Training: 34s. Estimated remaining time: 33h 3m 45s. Estimated total time: 41h 41m 1s. Time estimates for 10 more iterations: 8m 20s, 100 more iterations: 1h 23m 22s, 500 more iterations: 6h 56m 50s. +[2026-03-26 05:42:56,620][__main__][INFO] - Starting iteration 554. +[2026-03-26 05:42:57,040][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 55 and human policies 1. +[2026-03-26 05:42:57,040][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:43:11,226][__main__][INFO] - Number of regex retries in iteration 554: 0 +[2026-03-26 05:43:11,226][__main__][INFO] - agents played in iteration 554 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:43:12,033][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:43:12,057][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:43:12,081][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:43:12,105][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:43:12,106][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:43:12,106][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:43:12,884][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:43:13,347][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:43:13,866][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:43:14,370][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:43:14,877][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:43:15,378][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:43:15,882][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:43:16,396][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:43:16,899][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:43:17,400][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:43:17,899][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:43:18,404][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:43:18,910][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:43:19,414][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:43:19,925][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:43:20,428][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:43:20,932][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:43:21,432][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:43:21,933][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:43:22,440][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:43:22,943][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:43:23,450][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:43:23,953][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:43:24,454][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:43:24,955][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:43:25,455][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:43:25,955][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:43:26,456][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:43:26,963][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:43:27,472][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:43:27,977][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:43:28,477][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:43:28,981][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:43:29,483][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:43:29,985][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:43:30,491][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:43:30,994][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:43:31,499][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:43:32,007][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:43:32,519][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:43:33,023][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:43:33,548][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:43:34,054][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:43:34,559][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:43:35,062][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:43:35,565][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:43:36,068][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:43:36,574][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:43:37,079][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:43:37,581][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:43:38,086][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:43:38,597][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:43:39,097][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:43:39,598][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:43:40,100][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:43:40,601][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:43:41,102][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:43:41,601][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:43:42,105][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:43:42,604][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:43:43,104][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:43:43,605][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:43:44,105][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:43:44,604][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:43:45,103][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10129 tokens. +[2026-03-26 05:43:45,936][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.58%, Block Peak % of device VRAM: 62.40%, ΔTime: 00:00:33 +[2026-03-26 05:43:46,693][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:43:46,696][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:43:46,698][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:43:47,566][__main__][INFO] - Iteration 555 took 50s (28.08% Gen, 70.20% Train). Generation: 14s, Training: 35s. Estimated remaining time: 33h 28m 14s. Estimated total time: 42h 6m 22s. Time estimates for 10 more iterations: 8m 25s, 100 more iterations: 1h 24m 12s, 500 more iterations: 7h 1m 3s. +[2026-03-26 05:43:47,569][__main__][INFO] - Starting iteration 555. +[2026-03-26 05:43:47,977][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 55 and human policies 1. +[2026-03-26 05:43:47,977][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:44:01,317][__main__][INFO] - Number of regex retries in iteration 555: 0 +[2026-03-26 05:44:01,318][__main__][INFO] - agents played in iteration 555 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:44:02,121][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:44:02,145][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:44:02,169][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:44:02,192][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:44:02,193][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:44:02,193][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:44:02,968][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:44:03,432][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:44:03,935][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:44:04,434][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:44:04,933][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:44:05,435][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:44:05,936][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:44:06,440][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:44:06,941][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:44:07,444][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:44:07,955][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:44:08,457][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:44:08,960][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:44:09,461][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:44:09,964][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:44:10,463][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:44:10,964][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:44:11,465][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:44:11,964][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:44:12,462][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:44:12,960][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:44:13,458][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:44:13,956][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:44:14,453][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:44:14,952][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:44:15,451][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:44:15,951][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:44:16,453][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:44:16,951][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:44:17,450][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:44:17,949][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:44:18,447][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:44:18,947][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:44:19,448][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:44:19,947][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:44:20,450][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:44:20,947][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:44:21,450][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:44:21,950][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:44:22,454][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:44:22,956][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:44:23,459][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:44:23,960][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:44:24,461][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:44:24,961][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:44:25,465][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:44:25,969][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:44:26,468][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:44:26,968][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:44:27,468][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:44:27,967][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:44:28,467][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:44:28,968][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:44:29,469][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:44:29,969][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:44:30,468][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:44:30,969][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:44:31,465][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:44:31,966][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:44:32,467][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:44:32,964][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:44:33,461][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:44:33,957][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:44:34,453][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:44:34,949][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10094 tokens. +[2026-03-26 05:44:35,773][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.09%, Current % of VRAM taken: 60.52%, Block Peak % of device VRAM: 62.27%, ΔTime: 00:00:32 +[2026-03-26 05:44:36,536][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:44:36,538][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:44:36,540][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:44:37,266][__main__][INFO] - Iteration 556 took 49s (27.07% Gen, 71.46% Train). Generation: 13s, Training: 35s. Estimated remaining time: 32h 25m 31s. Estimated total time: 41h 4m 28s. Time estimates for 10 more iterations: 8m 12s, 100 more iterations: 1h 22m 8s, 500 more iterations: 6h 50m 44s. +[2026-03-26 05:44:37,268][__main__][INFO] - Starting iteration 556. +[2026-03-26 05:44:37,693][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 55 and human policies 1. +[2026-03-26 05:44:37,694][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:44:42,225][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 05:44:51,858][__main__][INFO] - Number of regex retries in iteration 556: 1 +[2026-03-26 05:44:51,859][__main__][INFO] - agents played in iteration 556 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:44:52,660][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:44:52,684][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:44:52,708][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:44:52,731][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:44:52,732][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:44:52,732][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:44:53,495][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:44:53,951][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:44:54,454][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:44:54,952][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:44:55,456][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:44:55,955][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:44:56,463][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:44:56,965][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:44:57,479][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:44:57,980][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:44:58,481][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:44:58,980][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:44:59,481][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:44:59,981][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:45:00,481][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:45:00,982][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:45:01,487][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:45:01,987][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:45:02,488][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:45:02,988][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:45:03,490][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:45:03,988][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:45:04,485][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:45:04,982][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:45:05,480][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:45:05,977][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:45:06,479][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:45:06,977][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:45:07,476][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:45:07,974][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:45:08,472][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:45:08,969][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:45:09,466][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:45:09,963][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:45:10,467][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:45:10,969][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:45:11,468][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:45:11,973][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:45:12,472][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:45:12,972][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:45:13,474][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:45:13,973][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:45:14,477][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:45:14,975][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:45:15,474][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:45:15,973][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:45:16,472][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:45:16,979][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:45:17,478][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:45:17,978][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:45:18,477][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:45:18,977][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:45:19,476][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:45:19,976][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:45:20,475][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:45:20,974][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:45:21,475][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:45:21,972][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:45:22,470][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:45:22,966][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:45:23,463][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:45:23,960][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:45:24,457][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:45:24,954][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:45:25,452][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10020 tokens. +[2026-03-26 05:45:26,266][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 60.48%, Block Peak % of device VRAM: 62.21%, ΔTime: 00:00:32 +[2026-03-26 05:45:26,993][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:45:26,995][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:45:26,997][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:45:27,725][__main__][INFO] - Iteration 557 took 50s (28.31% Gen, 70.23% Train). Generation: 14s, Training: 35s. Estimated remaining time: 33h 1m 51s. Estimated total time: 41h 41m 38s. Time estimates for 10 more iterations: 8m 20s, 100 more iterations: 1h 23m 23s, 500 more iterations: 6h 56m 56s. +[2026-03-26 05:45:27,727][__main__][INFO] - Starting iteration 557. +[2026-03-26 05:45:28,157][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 55 and human policies 1. +[2026-03-26 05:45:28,158][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:45:32,879][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 05:45:43,066][__main__][INFO] - Number of regex retries in iteration 557: 1 +[2026-03-26 05:45:43,067][__main__][INFO] - agents played in iteration 557 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:45:43,832][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:45:43,856][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:45:43,880][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:45:43,903][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:45:43,904][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:45:43,904][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:45:44,665][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:45:45,120][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:45:45,627][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:45:46,126][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:45:46,629][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:45:47,127][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:45:47,626][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:45:48,127][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:45:48,625][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:45:49,123][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:45:49,621][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:45:50,118][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:45:50,615][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:45:51,113][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:45:51,610][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:45:52,109][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:45:52,607][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:45:53,105][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:45:53,608][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:45:54,107][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:45:54,605][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:45:55,112][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:45:55,611][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:45:56,127][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:45:56,632][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:45:57,133][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:45:57,635][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:45:58,142][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:45:58,647][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:45:59,147][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:45:59,650][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:46:00,149][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:46:00,649][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:46:01,149][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:46:01,650][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:46:02,150][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:46:02,649][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:46:03,152][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:46:03,651][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:46:04,149][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:46:04,648][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:46:05,147][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:46:05,647][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:46:06,146][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:46:06,644][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:46:07,143][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:46:07,641][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:46:08,138][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:46:08,638][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:46:09,135][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:46:09,633][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:46:10,131][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:46:10,633][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:46:11,131][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:46:11,631][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:46:12,132][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:46:12,630][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:46:13,128][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:46:13,626][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:46:14,126][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:46:14,622][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:46:15,118][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:46:15,617][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:46:16,113][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:46:16,610][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10053 tokens. +[2026-03-26 05:46:17,433][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.04%, Current % of VRAM taken: 60.47%, Block Peak % of device VRAM: 62.14%, ΔTime: 00:00:32 +[2026-03-26 05:46:18,221][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:46:18,224][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:46:18,225][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:46:18,951][__main__][INFO] - Iteration 558 took 50s (29.35% Gen, 69.22% Train). Generation: 14s, Training: 35s. Estimated remaining time: 33h 39m 5s. Estimated total time: 42h 19m 44s. Time estimates for 10 more iterations: 8m 27s, 100 more iterations: 1h 24m 39s, 500 more iterations: 7h 3m 17s. +[2026-03-26 05:46:18,953][__main__][INFO] - Starting iteration 558. +[2026-03-26 05:46:19,376][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 55 and human policies 1. +[2026-03-26 05:46:19,377][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:46:24,204][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 05:46:29,751][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 05:46:33,842][__main__][INFO] - Number of regex retries in iteration 558: 2 +[2026-03-26 05:46:33,843][__main__][INFO] - agents played in iteration 558 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:46:34,630][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:46:34,658][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:46:34,683][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:46:34,707][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:46:34,708][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:46:34,708][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:46:35,491][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:46:35,952][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:46:36,462][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:46:36,963][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:46:37,462][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:46:37,960][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:46:38,472][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:46:38,979][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:46:39,479][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:46:39,986][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:46:40,491][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:46:40,994][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:46:41,503][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:46:42,016][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:46:42,517][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:46:43,022][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:46:43,525][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:46:44,024][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:46:44,523][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:46:45,027][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:46:45,528][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:46:46,029][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:46:46,529][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:46:47,030][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:46:47,532][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:46:48,032][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:46:48,531][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:46:49,033][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:46:49,548][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:46:50,050][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:46:50,551][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:46:51,053][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:46:51,554][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:46:52,055][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:46:52,554][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:46:53,053][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:46:53,551][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:46:54,057][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:46:54,566][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:46:55,066][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:46:55,565][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:46:56,067][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:46:56,566][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:46:57,065][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:46:57,564][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:46:58,063][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:46:58,561][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:46:59,061][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:46:59,559][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:47:00,058][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:47:00,555][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:47:01,054][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:47:01,553][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:47:02,052][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:47:02,556][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:47:03,055][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:47:03,553][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:47:04,052][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:47:04,549][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:47:05,046][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:47:05,544][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:47:06,041][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:47:06,538][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:47:07,040][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:47:07,541][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10048 tokens. +[2026-03-26 05:47:08,368][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 60.65%, Block Peak % of device VRAM: 62.13%, ΔTime: 00:00:32 +[2026-03-26 05:47:09,127][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:47:09,130][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:47:09,131][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:47:09,862][__main__][INFO] - Iteration 559 took 50s (28.65% Gen, 69.90% Train). Generation: 14s, Training: 35s. Estimated remaining time: 33h 22m 51s. Estimated total time: 42h 4m 20s. Time estimates for 10 more iterations: 8m 24s, 100 more iterations: 1h 24m 8s, 500 more iterations: 7h 0m 43s. +[2026-03-26 05:47:09,865][__main__][INFO] - Starting iteration 559. +[2026-03-26 05:47:10,267][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 55 and human policies 1. +[2026-03-26 05:47:10,267][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:47:25,881][__main__][INFO] - Number of regex retries in iteration 559: 0 +[2026-03-26 05:47:25,882][__main__][INFO] - agents played in iteration 559 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:47:26,680][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:47:26,704][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:47:26,728][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:47:26,752][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:47:26,753][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:47:26,753][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:47:27,529][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:47:27,993][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:47:28,500][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:47:28,999][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:47:29,498][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:47:30,005][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:47:30,513][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:47:31,015][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:47:31,515][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:47:32,020][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:47:32,525][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:47:33,038][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:47:33,539][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:47:34,041][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:47:34,570][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:47:35,073][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:47:35,572][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:47:36,077][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:47:36,595][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:47:37,097][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:47:37,599][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:47:38,099][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:47:38,599][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:47:39,100][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:47:39,600][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:47:40,109][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:47:40,615][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:47:41,124][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:47:41,650][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:47:42,156][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:47:42,658][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:47:43,161][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:47:43,662][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:47:44,164][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:47:44,662][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:47:45,164][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:47:45,668][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:47:46,173][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:47:46,674][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:47:47,175][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:47:47,675][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:47:48,175][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:47:48,675][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:47:49,178][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:47:49,676][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:47:50,176][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:47:50,676][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:47:51,175][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:47:51,680][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:47:52,184][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:47:52,695][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:47:53,195][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:47:53,697][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:47:54,204][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:47:54,703][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:47:55,206][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:47:55,704][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:47:56,204][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:47:56,703][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:47:57,207][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:47:57,706][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:47:58,207][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:47:58,706][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:47:59,208][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:47:59,715][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10104 tokens. +[2026-03-26 05:48:00,536][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 60.65%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:33 +[2026-03-26 05:48:01,360][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:48:01,363][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:48:01,364][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:48:02,096][__main__][INFO] - Iteration 560 took 51s (30.13% Gen, 68.46% Train). Generation: 15s, Training: 35s. Estimated remaining time: 34h 29m 8s. Estimated total time: 43h 11m 29s. Time estimates for 10 more iterations: 8m 38s, 100 more iterations: 1h 26m 22s, 500 more iterations: 7h 11m 54s. +[2026-03-26 05:48:02,099][__main__][INFO] - Starting iteration 560. +[2026-03-26 05:48:02,514][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 55 and human policies 1. +[2026-03-26 05:48:02,515][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:48:07,467][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 05:48:17,372][__main__][INFO] - Number of regex retries in iteration 560: 1 +[2026-03-26 05:48:17,373][__main__][INFO] - agents played in iteration 560 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:48:18,160][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:48:18,184][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:48:18,207][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:48:18,230][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:48:18,231][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:48:18,232][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:48:19,003][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:48:19,464][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:48:19,973][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:48:20,480][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:48:20,980][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:48:21,488][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:48:21,992][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:48:22,492][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:48:22,999][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:48:23,499][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:48:23,999][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:48:24,499][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:48:24,999][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:48:25,505][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:48:26,005][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:48:26,504][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:48:27,003][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:48:27,504][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:48:28,013][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:48:28,516][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:48:29,017][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:48:29,520][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:48:30,017][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:48:30,518][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:48:31,026][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:48:31,529][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:48:32,035][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:48:32,546][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:48:33,047][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:48:33,549][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:48:34,050][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:48:34,550][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:48:35,049][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:48:35,552][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:48:36,056][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:48:36,557][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:48:37,058][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:48:37,560][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:48:38,061][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:48:38,558][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:48:39,060][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:48:39,557][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:48:40,056][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:48:40,554][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:48:41,052][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:48:41,553][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:48:42,051][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:48:42,549][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:48:43,046][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:48:43,545][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:48:44,044][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:48:44,543][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:48:45,043][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:48:45,541][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:48:46,046][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:48:46,549][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:48:47,058][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:48:47,559][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:48:48,071][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:48:48,573][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:48:49,074][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:48:49,575][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:48:50,076][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:48:50,576][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:48:51,076][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10079 tokens. +[2026-03-26 05:48:51,908][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.05%, Current % of VRAM taken: 60.48%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 +[2026-03-26 05:48:52,627][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:48:52,630][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:48:52,632][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:48:53,901][__main__][INFO] - Iteration 561 took 51s (28.91% Gen, 68.61% Train). Generation: 14s, Training: 35s. Estimated remaining time: 34h 6m 9s. Estimated total time: 42h 49m 23s. Time estimates for 10 more iterations: 8m 33s, 100 more iterations: 1h 25m 38s, 500 more iterations: 7h 8m 13s. +[2026-03-26 05:48:53,903][__main__][INFO] - Starting iteration 561. +[2026-03-26 05:48:54,315][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 56 and human policies 1. +[2026-03-26 05:48:54,316][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:49:04,527][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 05:49:08,780][__main__][INFO] - Number of regex retries in iteration 561: 1 +[2026-03-26 05:49:08,781][__main__][INFO] - agents played in iteration 561 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:49:09,558][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:49:09,582][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:49:09,606][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:49:09,629][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:49:09,629][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:49:09,630][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:49:10,423][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:49:10,881][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:49:11,387][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:49:11,887][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:49:12,388][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:49:12,899][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:49:13,405][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:49:13,910][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:49:14,411][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:49:14,915][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:49:15,417][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:49:15,921][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:49:16,427][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:49:16,943][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:49:17,448][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:49:17,951][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:49:18,456][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:49:18,955][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:49:19,453][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:49:19,953][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:49:20,452][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:49:20,952][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:49:21,450][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:49:21,948][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:49:22,446][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:49:22,943][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:49:23,444][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:49:23,958][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:49:24,459][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:49:24,959][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:49:25,462][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:49:25,962][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:49:26,461][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:49:26,963][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:49:27,466][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:49:27,969][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:49:28,483][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:49:28,985][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:49:29,486][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:49:29,988][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:49:30,488][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:49:30,992][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:49:31,493][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:49:31,999][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:49:32,499][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:49:32,998][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:49:33,497][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:49:34,006][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:49:34,517][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:49:35,021][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:49:35,521][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:49:36,022][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:49:36,527][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:49:37,030][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:49:37,536][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:49:38,036][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:49:38,539][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:49:39,038][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:49:39,538][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:49:40,044][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:49:40,559][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:49:41,071][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:49:41,576][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:49:42,076][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:49:42,578][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10108 tokens. +[2026-03-26 05:49:43,406][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.14%, ΔTime: 00:00:32 +[2026-03-26 05:49:44,299][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:49:44,301][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:49:44,303][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:49:45,061][__main__][INFO] - Iteration 562 took 50s (28.50% Gen, 70.00% Train). Generation: 14s, Training: 35s. Estimated remaining time: 33h 33m 14s. Estimated total time: 42h 17m 19s. Time estimates for 10 more iterations: 8m 27s, 100 more iterations: 1h 24m 34s, 500 more iterations: 7h 2m 53s. +[2026-03-26 05:49:45,063][__main__][INFO] - Starting iteration 562. +[2026-03-26 05:49:45,474][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 56 and human policies 1. +[2026-03-26 05:49:45,475][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:49:59,214][__main__][INFO] - Number of regex retries in iteration 562: 0 +[2026-03-26 05:49:59,215][__main__][INFO] - agents played in iteration 562 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:49:59,985][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:50:00,009][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:50:00,033][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:50:00,057][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:50:00,058][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:50:00,058][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:50:00,841][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:50:01,301][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:50:01,813][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:50:02,336][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:50:02,841][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:50:03,346][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:50:03,851][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:50:04,356][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:50:04,862][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:50:05,370][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:50:05,887][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:50:06,393][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:50:06,895][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:50:07,397][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:50:07,901][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:50:08,401][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:50:08,902][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:50:09,403][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:50:09,906][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:50:10,405][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:50:10,907][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:50:11,410][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:50:11,911][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:50:12,411][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:50:12,914][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:50:13,412][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:50:13,911][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:50:14,409][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:50:14,908][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:50:15,406][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:50:15,904][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:50:16,400][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:50:16,897][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:50:17,395][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:50:17,895][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:50:18,393][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:50:18,891][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:50:19,390][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:50:19,889][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:50:20,389][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:50:20,889][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:50:21,389][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:50:21,890][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:50:22,389][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:50:22,889][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:50:23,389][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:50:23,886][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:50:24,388][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:50:24,888][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:50:25,384][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:50:25,879][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:50:26,376][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:50:26,873][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:50:27,370][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:50:27,867][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:50:28,363][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:50:28,860][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:50:29,356][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:50:29,851][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:50:30,352][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:50:30,851][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:50:31,349][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:50:31,847][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:50:32,343][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:50:32,839][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10122 tokens. +[2026-03-26 05:50:33,646][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 60.51%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:32 +[2026-03-26 05:50:34,421][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:50:34,423][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:50:34,425][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:50:35,158][__main__][INFO] - Iteration 563 took 49s (27.65% Gen, 70.87% Train). Generation: 13s, Training: 35s. Estimated remaining time: 32h 39m 19s. Estimated total time: 41h 24m 13s. Time estimates for 10 more iterations: 8m 16s, 100 more iterations: 1h 22m 48s, 500 more iterations: 6h 54m 2s. +[2026-03-26 05:50:35,160][__main__][INFO] - Starting iteration 563. +[2026-03-26 05:50:35,569][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 56 and human policies 1. +[2026-03-26 05:50:35,569][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:50:40,153][mllm.models.large_language_model_local][WARNING] - Response NewProposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 05:50:49,823][__main__][INFO] - Number of regex retries in iteration 563: 1 +[2026-03-26 05:50:49,824][__main__][INFO] - agents played in iteration 563 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:50:50,586][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:50:50,612][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:50:50,638][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:50:50,668][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:50:50,668][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:50:50,669][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:50:51,461][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:50:51,927][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:50:52,431][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:50:52,932][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:50:53,432][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:50:53,932][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:50:54,437][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:50:54,940][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:50:55,443][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:50:55,954][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:50:56,455][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:50:56,957][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:50:57,457][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:50:57,957][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:50:58,456][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:50:58,956][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:50:59,458][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:50:59,958][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:51:00,458][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:51:00,958][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:51:01,457][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:51:01,959][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:51:02,459][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:51:02,958][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:51:03,461][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:51:03,957][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:51:04,453][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:51:04,950][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:51:05,448][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:51:05,944][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:51:06,444][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:51:06,942][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:51:07,440][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:51:07,939][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:51:08,437][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:51:08,936][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:51:09,435][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:51:09,941][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:51:10,439][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:51:10,937][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:51:11,435][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:51:11,933][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:51:12,432][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:51:12,930][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:51:13,429][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:51:13,928][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:51:14,426][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:51:14,923][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:51:15,426][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:51:15,924][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:51:16,421][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:51:16,917][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:51:17,417][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:51:17,914][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:51:18,411][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:51:18,909][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:51:19,407][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:51:19,905][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:51:20,403][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:51:20,901][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:51:21,399][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:51:21,899][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:51:22,396][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:51:22,896][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:51:23,393][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10013 tokens. +[2026-03-26 05:51:24,222][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 60.52%, Block Peak % of device VRAM: 62.04%, ΔTime: 00:00:32 +[2026-03-26 05:51:25,018][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:51:25,020][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:51:25,021][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:51:25,877][__main__][INFO] - Iteration 564 took 50s (28.33% Gen, 69.96% Train). Generation: 14s, Training: 35s. Estimated remaining time: 33h 9m 42s. Estimated total time: 41h 55m 27s. Time estimates for 10 more iterations: 8m 23s, 100 more iterations: 1h 23m 50s, 500 more iterations: 6h 59m 14s. +[2026-03-26 05:51:25,881][__main__][INFO] - Starting iteration 564. +[2026-03-26 05:51:26,292][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 56 and human policies 1. +[2026-03-26 05:51:26,293][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:51:41,242][__main__][INFO] - Number of regex retries in iteration 564: 0 +[2026-03-26 05:51:41,243][__main__][INFO] - agents played in iteration 564 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:51:42,016][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:51:42,040][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:51:42,064][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:51:42,087][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:51:42,087][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:51:42,088][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:51:42,859][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:51:43,317][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:51:43,825][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:51:44,324][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:51:44,829][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:51:45,340][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:51:45,843][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:51:46,344][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:51:46,845][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:51:47,350][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:51:47,853][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:51:48,357][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:51:48,857][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:51:49,357][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:51:49,858][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:51:50,359][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:51:50,859][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:51:51,358][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:51:51,857][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:51:52,356][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:51:52,856][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:51:53,361][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:51:53,867][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:51:54,368][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:51:54,876][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:51:55,378][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:51:55,879][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:51:56,383][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:51:56,890][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:51:57,393][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:51:57,893][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:51:58,393][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:51:58,894][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:51:59,396][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:51:59,895][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:52:00,394][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:52:00,892][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:52:01,391][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:52:01,892][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:52:02,392][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:52:02,894][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:52:03,396][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:52:03,895][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:52:04,393][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:52:04,894][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:52:05,392][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:52:05,891][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:52:06,390][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:52:06,891][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:52:07,394][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:52:07,896][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:52:08,395][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:52:08,897][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:52:09,397][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:52:09,896][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:52:10,397][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:52:10,898][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:52:11,397][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:52:11,894][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:52:12,392][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:52:12,890][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:52:13,391][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:52:13,889][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:52:14,390][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:52:14,890][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10184 tokens. +[2026-03-26 05:52:15,704][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 60.60%, Block Peak % of device VRAM: 62.28%, ΔTime: 00:00:32 +[2026-03-26 05:52:16,445][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:52:16,447][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:52:16,449][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:52:17,335][__main__][INFO] - Iteration 565 took 51s (29.29% Gen, 68.97% Train). Generation: 14s, Training: 35s. Estimated remaining time: 33h 45m 33s. Estimated total time: 42h 32m 10s. Time estimates for 10 more iterations: 8m 30s, 100 more iterations: 1h 25m 4s, 500 more iterations: 7h 5m 21s. +[2026-03-26 05:52:17,337][__main__][INFO] - Starting iteration 565. +[2026-03-26 05:52:17,741][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 56 and human policies 1. +[2026-03-26 05:52:17,742][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:52:28,472][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 05:52:38,797][__main__][INFO] - Number of regex retries in iteration 565: 1 +[2026-03-26 05:52:38,797][__main__][INFO] - agents played in iteration 565 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:52:39,553][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:52:39,577][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:52:39,601][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:52:39,624][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:52:39,625][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:52:39,625][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:52:40,396][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:52:40,849][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:52:41,352][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:52:41,854][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:52:42,372][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:52:42,873][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:52:43,375][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:52:43,875][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:52:44,374][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:52:44,880][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:52:45,389][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:52:45,895][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:52:46,397][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:52:46,896][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:52:47,399][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:52:47,903][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:52:48,415][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:52:48,916][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:52:49,420][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:52:49,921][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:52:50,421][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:52:50,920][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:52:51,420][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:52:51,920][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:52:52,421][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:52:52,920][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:52:53,420][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:52:53,918][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:52:54,418][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:52:54,922][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:52:55,433][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:52:55,943][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:52:56,463][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:52:56,968][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:52:57,470][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:52:57,972][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:52:58,472][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:52:58,972][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:52:59,474][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:52:59,989][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:53:00,494][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:53:00,994][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:53:01,496][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:53:02,002][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:53:02,502][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:53:03,002][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:53:03,501][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:53:04,004][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:53:04,504][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:53:05,002][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:53:05,503][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:53:06,003][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:53:06,504][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:53:07,001][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:53:07,498][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:53:07,998][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:53:08,494][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:53:08,995][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:53:09,492][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:53:09,990][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:53:10,492][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:53:10,990][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:53:11,487][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:53:11,984][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:53:12,480][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10191 tokens. +[2026-03-26 05:53:13,298][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 60.53%, Block Peak % of device VRAM: 62.13%, ΔTime: 00:00:32 +[2026-03-26 05:53:14,006][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:53:14,008][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:53:14,010][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:53:14,841][__main__][INFO] - Iteration 566 took 57s (36.87% Gen, 61.67% Train). Generation: 21s, Training: 35s. Estimated remaining time: 38h 47m 25s. Estimated total time: 47h 34m 59s. Time estimates for 10 more iterations: 9m 30s, 100 more iterations: 1h 35m 9s, 500 more iterations: 7h 55m 49s. +[2026-03-26 05:53:14,843][__main__][INFO] - Starting iteration 566. +[2026-03-26 05:53:15,252][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 56 and human policies 1. +[2026-03-26 05:53:15,253][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:53:34,061][__main__][INFO] - Number of regex retries in iteration 566: 0 +[2026-03-26 05:53:34,062][__main__][INFO] - agents played in iteration 566 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:53:34,838][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:53:34,865][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:53:34,890][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:53:34,914][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:53:34,914][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:53:34,915][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:53:35,718][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:53:36,186][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:53:36,692][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:53:37,193][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:53:37,691][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:53:38,194][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:53:38,693][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:53:39,191][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:53:39,692][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:53:40,198][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:53:40,697][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:53:41,198][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:53:41,702][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:53:42,200][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:53:42,723][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:53:43,224][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:53:43,724][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:53:44,223][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:53:44,721][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:53:45,223][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:53:45,724][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:53:46,228][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:53:46,730][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:53:47,235][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:53:47,746][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:53:48,248][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:53:48,752][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:53:49,252][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:53:49,752][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:53:50,258][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:53:50,758][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:53:51,258][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:53:51,757][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:53:52,258][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:53:52,757][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:53:53,258][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:53:53,757][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:53:54,256][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:53:54,756][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:53:55,255][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:53:55,757][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:53:56,254][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:53:56,752][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:53:57,252][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:53:57,749][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:53:58,245][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:53:58,742][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:53:59,239][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:53:59,737][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:54:00,237][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:54:00,737][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:54:01,237][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:54:01,738][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:54:02,238][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:54:02,737][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:54:03,237][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:54:03,741][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:54:04,243][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:54:04,740][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:54:05,238][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:54:05,738][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:54:06,237][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:54:06,737][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:54:07,237][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:54:07,735][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10148 tokens. +[2026-03-26 05:54:08,659][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.58%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 +[2026-03-26 05:54:09,460][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:54:09,462][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:54:09,464][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:54:10,191][__main__][INFO] - Iteration 567 took 54s (34.24% Gen, 64.44% Train). Generation: 18s, Training: 35s. Estimated remaining time: 36h 58m 30s. Estimated total time: 45h 46m 59s. Time estimates for 10 more iterations: 9m 9s, 100 more iterations: 1h 31m 33s, 500 more iterations: 7h 37m 49s. +[2026-03-26 05:54:10,196][__main__][INFO] - Starting iteration 567. +[2026-03-26 05:54:10,606][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 56 and human policies 1. +[2026-03-26 05:54:10,606][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:55:12,546][__main__][INFO] - Number of regex retries in iteration 567: 0 +[2026-03-26 05:55:12,546][__main__][INFO] - agents played in iteration 567 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:55:13,334][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:55:13,358][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:55:13,382][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:55:13,405][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:55:13,406][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:55:13,406][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:55:14,214][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:55:14,677][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:55:15,185][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:55:15,687][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:55:16,193][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:55:16,697][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:55:17,197][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:55:17,698][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:55:18,201][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:55:18,704][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:55:19,203][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:55:19,712][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:55:20,225][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:55:20,728][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:55:21,231][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:55:21,732][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:55:22,234][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:55:22,737][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:55:23,243][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:55:23,752][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:55:24,257][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:55:24,759][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:55:25,260][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:55:25,760][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:55:26,261][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:55:26,762][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:55:27,263][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:55:27,769][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:55:28,269][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:55:28,773][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:55:29,273][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:55:29,778][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:55:30,285][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:55:30,797][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:55:31,297][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:55:31,802][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:55:32,304][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:55:32,803][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:55:33,303][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:55:33,803][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:55:34,302][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:55:34,806][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:55:35,315][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:55:35,822][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:55:36,326][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:55:36,835][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:55:37,338][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:55:37,839][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:55:38,340][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:55:38,840][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:55:39,342][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:55:39,841][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:55:40,346][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:55:40,851][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:55:41,361][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:55:41,862][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:55:42,362][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:55:42,868][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:55:43,372][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:55:43,872][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:55:44,373][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:55:44,874][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:55:45,374][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:55:45,875][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:55:46,377][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10222 tokens. +[2026-03-26 05:55:47,210][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 60.63%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:33 +[2026-03-26 05:55:48,006][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:55:48,008][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:55:48,010][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:55:48,752][__main__][INFO] - Iteration 568 took 1m 38s (63.11% Gen, 36.13% Train). Generation: 1m 1s, Training: 35s. Estimated remaining time: 72h 57m 14s. Estimated total time: 81h 47m 22s. Time estimates for 10 more iterations: 16m 21s, 100 more iterations: 2h 43m 34s, 500 more iterations: 13h 37m 53s. +[2026-03-26 05:55:48,756][__main__][INFO] - Starting iteration 568. +[2026-03-26 05:55:49,169][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 56 and human policies 1. +[2026-03-26 05:55:49,170][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 05:58:26,291][__main__][INFO] - Number of regex retries in iteration 568: 0 +[2026-03-26 05:58:26,292][__main__][INFO] - agents played in iteration 568 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 05:58:27,054][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:58:27,077][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:58:27,101][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:58:27,125][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 05:58:27,126][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 05:58:27,126][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 05:58:27,939][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 05:58:28,404][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 05:58:28,912][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 05:58:29,413][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 05:58:29,919][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 05:58:30,426][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 05:58:30,935][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 05:58:31,439][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 05:58:31,942][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 05:58:32,442][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 05:58:32,944][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 05:58:33,446][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 05:58:33,959][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 05:58:34,465][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 05:58:34,967][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 05:58:35,470][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 05:58:35,970][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 05:58:36,474][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 05:58:36,976][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 05:58:37,483][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 05:58:37,991][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 05:58:38,494][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 05:58:38,998][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 05:58:39,498][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 05:58:39,998][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 05:58:40,498][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 05:58:40,998][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 05:58:41,499][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 05:58:41,999][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 05:58:42,498][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 05:58:42,999][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 05:58:43,497][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 05:58:43,996][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 05:58:44,495][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 05:58:44,994][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 05:58:45,495][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 05:58:45,993][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 05:58:46,490][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 05:58:46,988][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 05:58:47,486][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 05:58:47,983][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 05:58:48,479][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 05:58:48,975][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 05:58:49,471][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 05:58:49,969][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 05:58:50,464][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 05:58:50,961][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 05:58:51,457][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 05:58:51,953][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 05:58:52,450][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 05:58:52,952][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 05:58:53,450][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 05:58:53,949][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 05:58:54,447][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 05:58:54,945][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 05:58:55,447][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 05:58:55,943][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 05:58:56,442][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 05:58:56,940][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 05:58:57,436][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 05:58:57,933][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 05:58:58,430][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 05:58:58,929][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 05:58:59,429][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 05:58:59,926][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10058 tokens. +[2026-03-26 05:59:00,746][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 60.51%, Block Peak % of device VRAM: 62.16%, ΔTime: 00:00:32 +[2026-03-26 05:59:01,512][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 05:59:01,514][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 05:59:01,516][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 05:59:02,319][__main__][INFO] - Iteration 569 took 3m 13s (81.35% Gen, 18.24% Train). Generation: 2m 37s, Training: 35s. Estimated remaining time: 152h 4m 9s. Estimated total time: 160h 57m 31s. Time estimates for 10 more iterations: 32m 11s, 100 more iterations: 5h 21m 55s, 500 more iterations: 26h 49m 35s. +[2026-03-26 05:59:02,321][__main__][INFO] - Starting iteration 569. +[2026-03-26 05:59:02,725][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 56 and human policies 1. +[2026-03-26 05:59:02,725][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:00:12,493][__main__][INFO] - Number of regex retries in iteration 569: 0 +[2026-03-26 06:00:12,494][__main__][INFO] - agents played in iteration 569 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:00:13,250][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:00:13,274][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:00:13,298][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:00:13,321][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:00:13,322][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:00:13,323][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:00:14,128][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:00:14,594][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:00:15,102][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:00:15,602][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:00:16,105][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:00:16,617][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:00:17,121][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:00:17,623][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:00:18,123][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:00:18,627][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:00:19,126][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:00:19,626][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:00:20,127][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:00:20,632][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:00:21,132][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:00:21,631][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:00:22,138][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:00:22,638][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:00:23,139][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:00:23,638][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:00:24,138][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:00:24,637][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:00:25,136][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:00:25,636][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:00:26,136][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:00:26,635][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:00:27,133][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:00:27,631][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:00:28,131][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:00:28,627][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:00:29,124][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:00:29,621][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:00:30,118][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:00:30,617][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:00:31,116][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:00:31,616][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:00:32,117][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:00:32,616][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:00:33,115][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:00:33,614][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:00:34,119][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:00:34,616][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:00:35,115][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:00:35,613][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:00:36,114][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:00:36,610][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:00:37,111][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:00:37,612][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:00:38,112][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:00:38,612][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:00:39,108][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:00:39,604][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:00:40,102][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:00:40,598][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:00:41,093][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:00:41,593][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:00:42,088][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:00:42,587][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:00:43,089][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:00:43,589][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:00:44,089][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:00:44,587][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:00:45,088][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:00:45,587][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:00:46,090][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10157 tokens. +[2026-03-26 06:00:46,928][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.26%, Current % of VRAM taken: 60.68%, Block Peak % of device VRAM: 62.22%, ΔTime: 00:00:32 +[2026-03-26 06:00:47,715][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:00:47,717][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:00:47,719][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:00:48,449][__main__][INFO] - Iteration 570 took 1m 45s (65.99% Gen, 33.32% Train). Generation: 1m 9s, Training: 35s. Estimated remaining time: 79h 11m 9s. Estimated total time: 88h 6m 17s. Time estimates for 10 more iterations: 17m 37s, 100 more iterations: 2h 56m 12s, 500 more iterations: 14h 41m 2s. +[2026-03-26 06:00:48,451][__main__][INFO] - Starting iteration 570. +[2026-03-26 06:00:48,863][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 56 and human policies 1. +[2026-03-26 06:00:48,864][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:01:05,350][__main__][INFO] - Number of regex retries in iteration 570: 0 +[2026-03-26 06:01:05,351][__main__][INFO] - agents played in iteration 570 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:01:06,120][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:01:06,144][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:01:06,168][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:01:06,193][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:01:06,193][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:01:06,194][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:01:06,984][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:01:07,443][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:01:07,966][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:01:08,468][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:01:08,969][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:01:09,474][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:01:09,977][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:01:10,477][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:01:10,975][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:01:11,475][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:01:11,978][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:01:12,481][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:01:12,989][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:01:13,495][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:01:13,998][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:01:14,498][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:01:14,999][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:01:15,500][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:01:16,005][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:01:16,503][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:01:17,004][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:01:17,508][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:01:18,013][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:01:18,524][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:01:19,031][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:01:19,536][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:01:20,040][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:01:20,543][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:01:21,048][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:01:21,550][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:01:22,056][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:01:22,559][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:01:23,060][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:01:23,560][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:01:24,063][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:01:24,568][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:01:25,073][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:01:25,575][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:01:26,076][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:01:26,581][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:01:27,081][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:01:27,581][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:01:28,085][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:01:28,585][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:01:29,088][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:01:29,592][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:01:30,091][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:01:30,591][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:01:31,094][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:01:31,596][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:01:32,099][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:01:32,610][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:01:33,113][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:01:33,613][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:01:34,115][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:01:34,616][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:01:35,117][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:01:35,638][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:01:36,143][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:01:36,645][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:01:37,149][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:01:37,649][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:01:38,152][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:01:38,652][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:01:39,155][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10245 tokens. +[2026-03-26 06:01:39,980][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.25%, Current % of VRAM taken: 60.67%, Block Peak % of device VRAM: 62.20%, ΔTime: 00:00:33 +[2026-03-26 06:01:40,735][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:01:40,738][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:01:40,739][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:01:42,033][__main__][INFO] - Iteration 571 took 53s (31.01% Gen, 66.56% Train). Generation: 16s, Training: 35s. Estimated remaining time: 35h 22m 28s. Estimated total time: 44h 18m 29s. Time estimates for 10 more iterations: 8m 51s, 100 more iterations: 1h 28m 36s, 500 more iterations: 7h 23m 4s. +[2026-03-26 06:01:42,035][__main__][INFO] - Starting iteration 571. +[2026-03-26 06:01:42,446][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 57 and human policies 1. +[2026-03-26 06:01:42,447][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:01:57,683][__main__][INFO] - Number of regex retries in iteration 571: 0 +[2026-03-26 06:01:57,684][__main__][INFO] - agents played in iteration 571 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:01:58,456][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:01:58,481][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:01:58,504][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:01:58,531][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:01:58,532][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:01:58,532][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:01:59,325][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:01:59,792][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:02:00,304][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:02:00,807][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:02:01,312][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:02:01,831][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:02:02,338][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:02:02,844][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:02:03,347][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:02:03,847][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:02:04,380][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:02:04,880][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:02:05,391][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:02:05,900][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:02:06,403][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:02:06,905][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:02:07,406][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:02:07,911][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:02:08,412][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:02:08,916][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:02:09,432][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:02:09,934][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:02:10,441][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:02:10,946][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:02:11,446][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:02:11,945][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:02:12,444][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:02:12,945][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:02:13,450][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:02:13,958][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:02:14,462][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:02:14,961][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:02:15,463][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:02:15,963][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:02:16,463][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:02:16,963][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:02:17,464][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:02:17,964][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:02:18,466][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:02:18,969][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:02:19,476][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:02:19,984][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:02:20,483][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:02:20,989][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:02:21,493][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:02:21,996][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:02:22,498][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:02:23,000][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:02:23,502][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:02:23,999][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:02:24,502][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:02:25,003][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:02:25,503][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:02:26,007][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:02:26,507][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:02:27,005][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:02:27,504][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:02:28,005][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:02:28,503][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:02:29,001][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:02:29,500][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:02:29,997][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:02:30,496][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:02:30,996][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:02:31,497][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10129 tokens. +[2026-03-26 06:02:32,344][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.35%, ΔTime: 00:00:33 +[2026-03-26 06:02:33,139][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:02:33,141][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:02:33,143][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:02:33,930][__main__][INFO] - Iteration 572 took 51s (29.60% Gen, 68.87% Train). Generation: 15s, Training: 35s. Estimated remaining time: 33h 57m 18s. Estimated total time: 42h 54m 12s. Time estimates for 10 more iterations: 8m 34s, 100 more iterations: 1h 25m 48s, 500 more iterations: 7h 9m 2s. +[2026-03-26 06:02:33,932][__main__][INFO] - Starting iteration 572. +[2026-03-26 06:02:34,344][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 57 and human policies 1. +[2026-03-26 06:02:34,345][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:02:54,200][__main__][INFO] - Number of regex retries in iteration 572: 0 +[2026-03-26 06:02:54,201][__main__][INFO] - agents played in iteration 572 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:02:54,966][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:02:54,990][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:02:55,015][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:02:55,039][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:02:55,040][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:02:55,040][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:02:55,841][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:02:56,303][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:02:56,819][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:02:57,323][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:02:57,828][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:02:58,332][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:02:58,844][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:02:59,348][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:02:59,851][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:03:00,353][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:03:00,857][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:03:01,359][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:03:01,867][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:03:02,382][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:03:02,884][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:03:03,388][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:03:03,890][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:03:04,394][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:03:04,892][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:03:05,398][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:03:05,899][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:03:06,401][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:03:06,905][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:03:07,403][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:03:07,901][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:03:08,407][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:03:08,917][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:03:09,423][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:03:09,927][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:03:10,434][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:03:10,939][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:03:11,441][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:03:11,943][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:03:12,441][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:03:12,948][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:03:13,460][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:03:13,964][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:03:14,465][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:03:14,965][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:03:15,465][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:03:15,987][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:03:16,486][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:03:16,988][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:03:17,487][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:03:17,986][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:03:18,488][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:03:18,991][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:03:19,496][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:03:19,998][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:03:20,498][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:03:21,001][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:03:21,502][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:03:22,002][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:03:22,499][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:03:22,998][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:03:23,499][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:03:23,999][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:03:24,498][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:03:25,000][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:03:25,502][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:03:26,007][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:03:26,509][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:03:27,013][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:03:27,518][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:03:28,018][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10136 tokens. +[2026-03-26 06:03:29,033][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.04%, Current % of VRAM taken: 60.47%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:33 +[2026-03-26 06:03:29,805][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:03:29,807][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:03:29,809][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:03:30,532][__main__][INFO] - Iteration 573 took 56s (35.34% Gen, 63.37% Train). Generation: 19s, Training: 35s. Estimated remaining time: 37h 51m 37s. Estimated total time: 46h 49m 27s. Time estimates for 10 more iterations: 9m 21s, 100 more iterations: 1h 33m 38s, 500 more iterations: 7h 48m 14s. +[2026-03-26 06:03:30,534][__main__][INFO] - Starting iteration 573. +[2026-03-26 06:03:30,937][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 57 and human policies 1. +[2026-03-26 06:03:30,937][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:03:44,958][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 06:03:53,172][__main__][INFO] - Number of regex retries in iteration 573: 1 +[2026-03-26 06:03:53,173][__main__][INFO] - agents played in iteration 573 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:03:53,944][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:03:53,971][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:03:53,997][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:03:54,023][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:03:54,023][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:03:54,024][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:03:54,841][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:03:55,386][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:03:55,894][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:03:56,398][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:03:56,906][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:03:57,407][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:03:57,908][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:03:58,407][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:03:58,905][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:03:59,406][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:03:59,909][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:04:00,428][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:04:00,932][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:04:01,436][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:04:01,941][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:04:02,442][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:04:02,942][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:04:03,443][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:04:03,943][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:04:04,446][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:04:04,947][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:04:05,448][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:04:05,948][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:04:06,447][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:04:06,947][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:04:07,470][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:04:07,982][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:04:08,491][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:04:08,993][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:04:09,494][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:04:09,996][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:04:10,497][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:04:10,998][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:04:11,495][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:04:11,996][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:04:12,498][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:04:13,002][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:04:13,513][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:04:14,019][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:04:14,517][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:04:15,018][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:04:15,518][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:04:16,017][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:04:16,515][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:04:17,013][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:04:17,512][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:04:18,011][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:04:18,509][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:04:19,007][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:04:19,508][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:04:20,010][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:04:20,513][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:04:21,014][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:04:21,515][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:04:22,018][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:04:22,519][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:04:23,019][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:04:23,516][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:04:24,013][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:04:24,510][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:04:25,011][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:04:25,506][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:04:26,002][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:04:26,498][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:04:26,995][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10065 tokens. +[2026-03-26 06:04:27,840][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 60.50%, Block Peak % of device VRAM: 62.34%, ΔTime: 00:00:33 +[2026-03-26 06:04:28,572][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:04:28,575][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:04:28,576][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:04:29,293][__main__][INFO] - Iteration 574 took 58s (38.10% Gen, 60.67% Train). Generation: 22s, Training: 35s. Estimated remaining time: 39h 39m 2s. Estimated total time: 48h 37m 51s. Time estimates for 10 more iterations: 9m 43s, 100 more iterations: 1h 37m 15s, 500 more iterations: 8h 6m 18s. +[2026-03-26 06:04:29,296][__main__][INFO] - Starting iteration 574. +[2026-03-26 06:04:29,705][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 57 and human policies 1. +[2026-03-26 06:04:29,706][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:04:39,987][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 06:04:49,893][__main__][INFO] - Number of regex retries in iteration 574: 1 +[2026-03-26 06:04:49,894][__main__][INFO] - agents played in iteration 574 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:04:50,660][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:04:50,684][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:04:50,708][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:04:50,731][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:04:50,732][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:04:50,732][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:04:51,513][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:04:51,974][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:04:52,479][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:04:52,983][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:04:53,487][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:04:53,990][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:04:54,490][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:04:54,989][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:04:55,487][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:04:55,985][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:04:56,492][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:04:56,990][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:04:57,489][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:04:57,987][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:04:58,486][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:04:58,984][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:04:59,487][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:04:59,991][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:05:00,496][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:05:01,006][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:05:01,506][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:05:02,008][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:05:02,510][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:05:03,011][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:05:03,511][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:05:04,008][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:05:04,510][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:05:05,009][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:05:05,512][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:05:06,015][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:05:06,516][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:05:07,014][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:05:07,512][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:05:08,014][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:05:08,513][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:05:09,012][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:05:09,512][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:05:10,012][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:05:10,516][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:05:11,016][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:05:11,517][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:05:12,014][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:05:12,510][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:05:13,007][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:05:13,501][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:05:13,996][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:05:14,492][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:05:14,987][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:05:15,482][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:05:15,980][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:05:16,481][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:05:16,978][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:05:17,477][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:05:17,977][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:05:18,474][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:05:18,973][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:05:19,472][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:05:19,968][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:05:20,464][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:05:20,963][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:05:21,459][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:05:21,955][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:05:22,452][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:05:22,950][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:05:23,447][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10071 tokens. +[2026-03-26 06:05:24,273][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 60.50%, Block Peak % of device VRAM: 62.17%, ΔTime: 00:00:32 +[2026-03-26 06:05:25,058][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:05:25,061][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:05:25,063][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:05:25,907][__main__][INFO] - Iteration 575 took 56s (35.92% Gen, 62.58% Train). Generation: 20s, Training: 35s. Estimated remaining time: 37h 50m 22s. Estimated total time: 46h 50m 8s. Time estimates for 10 more iterations: 9m 22s, 100 more iterations: 1h 33m 40s, 500 more iterations: 7h 48m 21s. +[2026-03-26 06:05:25,909][__main__][INFO] - Starting iteration 575. +[2026-03-26 06:05:26,329][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 57 and human policies 1. +[2026-03-26 06:05:26,329][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:06:20,343][__main__][INFO] - Number of regex retries in iteration 575: 0 +[2026-03-26 06:06:20,344][__main__][INFO] - agents played in iteration 575 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:06:21,110][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:06:21,134][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:06:21,158][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:06:21,181][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:06:21,181][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:06:21,182][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:06:21,960][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:06:22,417][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:06:22,918][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:06:23,416][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:06:23,920][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:06:24,417][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:06:24,918][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:06:25,415][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:06:25,912][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:06:26,408][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:06:26,902][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:06:27,401][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:06:27,896][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:06:28,393][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:06:28,889][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:06:29,386][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:06:29,887][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:06:30,384][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:06:30,889][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:06:31,391][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:06:31,902][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:06:32,406][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:06:32,908][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:06:33,407][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:06:33,909][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:06:34,413][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:06:34,913][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:06:35,412][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:06:35,912][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:06:36,410][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:06:36,909][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:06:37,407][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:06:37,905][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:06:38,402][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:06:38,902][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:06:39,399][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:06:39,899][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:06:40,396][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:06:40,893][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:06:41,394][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:06:41,891][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:06:42,390][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:06:42,888][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:06:43,386][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:06:43,885][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:06:44,383][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:06:44,883][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:06:45,384][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:06:45,884][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:06:46,382][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:06:46,882][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:06:47,379][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:06:47,876][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:06:48,373][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:06:48,869][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:06:49,365][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:06:49,864][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:06:50,360][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:06:50,858][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:06:51,357][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:06:51,854][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:06:52,352][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:06:52,850][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:06:53,348][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:06:53,845][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10139 tokens. +[2026-03-26 06:06:54,659][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.20%, ΔTime: 00:00:32 +[2026-03-26 06:06:55,393][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:06:55,395][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:06:55,397][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:06:56,110][__main__][INFO] - Iteration 576 took 1m 29s (60.16% Gen, 39.04% Train). Generation: 54s, Training: 35s. Estimated remaining time: 65h 47m 50s. Estimated total time: 74h 49m 6s. Time estimates for 10 more iterations: 14m 57s, 100 more iterations: 2h 29m 38s, 500 more iterations: 12h 28m 11s. +[2026-03-26 06:06:56,112][__main__][INFO] - Starting iteration 576. +[2026-03-26 06:06:56,516][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 57 and human policies 1. +[2026-03-26 06:06:56,517][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:07:14,062][__main__][INFO] - Number of regex retries in iteration 576: 0 +[2026-03-26 06:07:14,063][__main__][INFO] - agents played in iteration 576 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:07:14,848][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:07:14,872][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:07:14,896][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:07:14,919][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:07:14,920][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:07:14,921][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:07:15,698][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:07:16,157][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:07:16,661][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:07:17,177][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:07:17,687][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:07:18,191][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:07:18,692][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:07:19,196][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:07:19,695][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:07:20,198][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:07:20,706][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:07:21,208][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:07:21,707][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:07:22,210][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:07:22,709][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:07:23,209][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:07:23,717][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:07:24,226][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:07:24,731][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:07:25,234][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:07:25,736][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:07:26,237][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:07:26,737][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:07:27,237][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:07:27,738][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:07:28,240][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:07:28,743][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:07:29,242][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:07:29,744][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:07:30,243][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:07:30,747][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:07:31,251][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:07:31,758][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:07:32,263][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:07:32,763][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:07:33,261][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:07:33,761][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:07:34,260][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:07:34,759][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:07:35,257][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:07:35,759][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:07:36,268][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:07:36,778][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:07:37,281][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:07:37,783][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:07:38,283][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:07:38,784][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:07:39,284][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:07:39,783][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:07:40,283][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:07:40,789][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:07:41,289][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:07:41,788][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:07:42,288][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:07:42,789][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:07:43,288][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:07:43,813][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:07:44,312][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:07:44,814][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:07:45,318][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:07:45,826][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:07:46,328][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:07:46,828][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:07:47,327][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:07:47,827][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9997 tokens. +[2026-03-26 06:07:48,650][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 60.49%, Block Peak % of device VRAM: 62.12%, ΔTime: 00:00:32 +[2026-03-26 06:07:49,413][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:07:49,415][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:07:49,416][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:07:50,224][__main__][INFO] - Iteration 577 took 53s (32.67% Gen, 65.82% Train). Generation: 17s, Training: 35s. Estimated remaining time: 35h 43m 13s. Estimated total time: 44h 45m 23s. Time estimates for 10 more iterations: 8m 57s, 100 more iterations: 1h 29m 30s, 500 more iterations: 7h 27m 33s. +[2026-03-26 06:07:50,226][__main__][INFO] - Starting iteration 577. +[2026-03-26 06:07:50,637][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 57 and human policies 1. +[2026-03-26 06:07:50,637][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:08:08,478][__main__][INFO] - Number of regex retries in iteration 577: 0 +[2026-03-26 06:08:08,479][__main__][INFO] - agents played in iteration 577 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:08:09,255][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:08:09,279][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:08:09,303][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:08:09,326][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:08:09,326][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:08:09,327][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:08:10,108][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:08:10,578][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:08:11,090][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:08:11,596][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:08:12,096][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:08:12,597][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:08:13,096][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:08:13,663][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:08:14,161][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:08:14,662][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:08:15,163][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:08:15,662][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:08:16,168][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:08:16,680][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:08:17,184][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:08:17,686][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:08:18,186][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:08:18,692][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:08:19,196][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:08:19,706][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:08:20,268][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:08:20,780][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:08:21,286][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:08:21,882][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:08:22,460][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:08:22,962][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:08:23,464][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:08:23,968][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:08:24,482][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:08:24,989][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:08:25,493][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:08:25,997][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:08:26,501][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:08:27,000][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:08:27,499][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:08:27,999][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:08:28,504][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:08:29,012][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:08:29,514][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:08:30,014][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:08:30,513][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:08:31,014][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:08:31,513][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:08:32,016][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:08:32,517][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:08:33,017][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:08:33,520][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:08:34,026][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:08:34,532][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:08:35,036][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:08:35,539][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:08:36,039][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:08:36,542][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:08:37,044][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:08:37,547][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:08:38,045][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:08:38,546][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:08:39,049][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:08:39,548][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:08:40,055][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:08:40,559][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:08:41,067][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:08:41,574][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:08:42,075][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:08:42,577][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10140 tokens. +[2026-03-26 06:08:43,928][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 60.57%, Block Peak % of device VRAM: 62.22%, ΔTime: 00:00:33 +[2026-03-26 06:08:44,627][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:08:44,630][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:08:44,632][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:08:45,313][__main__][INFO] - Iteration 578 took 54s (32.63% Gen, 66.12% Train). Generation: 17s, Training: 36s. Estimated remaining time: 36h 30m 47s. Estimated total time: 45h 33m 52s. Time estimates for 10 more iterations: 9m 6s, 100 more iterations: 1h 31m 7s, 500 more iterations: 7h 35m 38s. +[2026-03-26 06:08:45,318][__main__][INFO] - Starting iteration 578. +[2026-03-26 06:08:45,728][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 57 and human policies 1. +[2026-03-26 06:08:45,729][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:10:05,464][__main__][INFO] - Number of regex retries in iteration 578: 0 +[2026-03-26 06:10:05,465][__main__][INFO] - agents played in iteration 578 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:10:06,465][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:10:06,489][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:10:06,513][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:10:06,536][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:10:06,537][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:10:06,537][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:10:07,332][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:10:07,793][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:10:08,303][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:10:08,802][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:10:09,305][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:10:09,816][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:10:10,318][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:10:10,823][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:10:11,323][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:10:11,829][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:10:12,329][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:10:12,830][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:10:13,330][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:10:13,828][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:10:14,326][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:10:14,824][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:10:15,326][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:10:15,825][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:10:16,324][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:10:16,824][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:10:17,323][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:10:17,822][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:10:18,320][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:10:18,824][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:10:19,329][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:10:19,829][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:10:20,333][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:10:20,834][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:10:21,343][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:10:21,849][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:10:22,353][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:10:22,856][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:10:23,358][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:10:23,860][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:10:24,362][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:10:24,863][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:10:25,364][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:10:25,865][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:10:26,366][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:10:26,866][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:10:27,371][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:10:27,873][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:10:28,377][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:10:28,879][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:10:29,403][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:10:29,906][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:10:30,405][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:10:30,904][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:10:31,402][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:10:31,906][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:10:32,403][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:10:32,903][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:10:33,401][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:10:33,907][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:10:34,407][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:10:34,911][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:10:35,412][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:10:35,922][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:10:36,426][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:10:36,931][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:10:37,433][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:10:37,933][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:10:38,434][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:10:38,935][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:10:39,436][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10037 tokens. +[2026-03-26 06:10:40,291][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 60.57%, Block Peak % of device VRAM: 62.12%, ΔTime: 00:00:32 +[2026-03-26 06:10:41,060][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:10:41,062][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:10:41,064][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:10:41,828][__main__][INFO] - Iteration 579 took 1m 56s (68.68% Gen, 30.66% Train). Generation: 1m 19s, Training: 35s. Estimated remaining time: 87h 40m 1s. Estimated total time: 96h 45m 2s. Time estimates for 10 more iterations: 19m 21s, 100 more iterations: 3h 13m 30s, 500 more iterations: 16h 7m 30s. +[2026-03-26 06:10:41,831][__main__][INFO] - Starting iteration 579. +[2026-03-26 06:10:42,236][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 57 and human policies 1. +[2026-03-26 06:10:42,236][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:13:13,445][__main__][INFO] - Number of regex retries in iteration 579: 0 +[2026-03-26 06:13:13,447][__main__][INFO] - agents played in iteration 579 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:13:14,259][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:13:14,283][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:13:14,306][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:13:14,330][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:13:14,330][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:13:14,331][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:13:15,112][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:13:15,569][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:13:16,071][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:13:16,567][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:13:17,070][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:13:17,566][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:13:18,063][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:13:18,559][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:13:19,056][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:13:19,557][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:13:20,058][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:13:20,557][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:13:21,058][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:13:21,558][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:13:22,058][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:13:22,558][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:13:23,059][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:13:23,555][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:13:24,053][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:13:24,553][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:13:25,053][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:13:25,549][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:13:26,044][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:13:26,544][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:13:27,040][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:13:27,548][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:13:28,052][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:13:28,554][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:13:29,060][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:13:29,582][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:13:30,092][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:13:30,597][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:13:31,101][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:13:31,605][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:13:32,104][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:13:32,602][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:13:33,103][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:13:33,604][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:13:34,102][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:13:34,604][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:13:35,102][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:13:35,601][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:13:36,102][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:13:36,605][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:13:37,104][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:13:37,604][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:13:38,102][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:13:38,601][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:13:39,101][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:13:39,599][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:13:40,096][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:13:40,595][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:13:41,093][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:13:41,594][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:13:42,091][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:13:42,588][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:13:43,084][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:13:43,584][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:13:44,082][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:13:44,582][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:13:45,080][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:13:45,584][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:13:46,086][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:13:46,584][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:13:47,082][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10151 tokens. +[2026-03-26 06:13:47,899][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 60.60%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 +[2026-03-26 06:13:48,670][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:13:48,672][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:13:48,673][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:13:49,401][__main__][INFO] - Iteration 580 took 3m 7s (80.79% Gen, 18.82% Train). Generation: 2m 31s, Training: 35s. Estimated remaining time: 146h 50m 9s. Estimated total time: 155h 58m 18s. Time estimates for 10 more iterations: 31m 11s, 100 more iterations: 5h 11m 56s, 500 more iterations: 25h 59m 43s. +[2026-03-26 06:13:49,404][__main__][INFO] - Starting iteration 580. +[2026-03-26 06:13:49,802][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 57 and human policies 1. +[2026-03-26 06:13:49,803][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:14:59,877][__main__][INFO] - Number of regex retries in iteration 580: 0 +[2026-03-26 06:14:59,878][__main__][INFO] - agents played in iteration 580 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:15:00,651][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:15:00,675][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:15:00,699][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:15:00,722][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:15:00,723][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:15:00,723][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:15:01,500][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:15:01,957][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:15:02,460][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:15:02,976][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:15:03,482][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:15:03,986][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:15:04,488][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:15:04,988][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:15:05,493][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:15:06,009][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:15:06,525][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:15:07,032][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:15:07,539][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:15:08,042][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:15:08,545][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:15:09,055][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:15:09,567][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:15:10,073][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:15:10,575][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:15:11,078][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:15:11,581][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:15:12,082][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:15:12,590][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:15:13,098][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:15:13,602][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:15:14,105][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:15:14,607][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:15:15,107][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:15:15,609][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:15:16,110][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:15:16,610][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:15:17,111][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:15:17,611][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:15:18,112][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:15:18,612][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:15:19,113][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:15:19,614][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:15:20,115][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:15:20,616][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:15:21,116][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:15:21,618][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:15:22,116][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:15:22,615][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:15:23,118][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:15:23,617][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:15:24,120][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:15:24,618][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:15:25,116][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:15:25,615][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:15:26,117][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:15:26,636][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:15:27,142][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:15:27,642][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:15:28,144][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:15:28,644][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:15:29,144][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:15:29,644][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:15:30,146][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:15:30,647][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:15:31,147][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:15:31,647][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:15:32,147][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:15:32,646][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:15:33,145][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:15:33,645][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10120 tokens. +[2026-03-26 06:15:34,464][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 60.62%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:32 +[2026-03-26 06:15:35,207][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:15:35,209][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:15:35,211][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:15:36,517][__main__][INFO] - Iteration 581 took 1m 46s (65.67% Gen, 33.11% Train). Generation: 1m 10s, Training: 35s. Estimated remaining time: 79h 45m 50s. Estimated total time: 88h 55m 46s. Time estimates for 10 more iterations: 17m 47s, 100 more iterations: 2h 57m 51s, 500 more iterations: 14h 49m 17s. +[2026-03-26 06:15:36,519][__main__][INFO] - Starting iteration 581. +[2026-03-26 06:15:36,933][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 58 and human policies 1. +[2026-03-26 06:15:36,934][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:16:06,086][__main__][INFO] - Number of regex retries in iteration 581: 0 +[2026-03-26 06:16:06,087][__main__][INFO] - agents played in iteration 581 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:16:06,871][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:16:06,897][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:16:06,923][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:16:06,949][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:16:06,950][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:16:06,950][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:16:07,748][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:16:08,217][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:16:08,729][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:16:09,229][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:16:09,736][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:16:10,248][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:16:10,756][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:16:11,264][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:16:11,764][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:16:12,268][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:16:12,781][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:16:13,287][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:16:13,789][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:16:14,290][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:16:14,791][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:16:15,293][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:16:15,794][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:16:16,298][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:16:16,809][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:16:17,320][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:16:17,822][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:16:18,322][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:16:18,825][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:16:19,323][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:16:19,823][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:16:20,323][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:16:20,831][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:16:21,340][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:16:21,845][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:16:22,348][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:16:22,850][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:16:23,351][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:16:23,851][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:16:24,351][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:16:24,850][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:16:25,351][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:16:25,854][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:16:26,354][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:16:26,854][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:16:27,353][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:16:27,852][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:16:28,350][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:16:28,850][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:16:29,348][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:16:29,851][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:16:30,352][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:16:30,850][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:16:31,348][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:16:31,847][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:16:32,348][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:16:32,850][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:16:33,347][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:16:33,846][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:16:34,344][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:16:34,843][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:16:35,340][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:16:35,838][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:16:36,337][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:16:36,837][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:16:37,336][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:16:37,834][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:16:38,331][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:16:38,827][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:16:39,322][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:16:39,819][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10048 tokens. +[2026-03-26 06:16:40,641][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 13.99%, Current % of VRAM taken: 60.42%, Block Peak % of device VRAM: 62.02%, ΔTime: 00:00:32 +[2026-03-26 06:16:41,346][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:16:41,348][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:16:41,349][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:16:42,084][__main__][INFO] - Iteration 582 took 1m 5s (44.75% Gen, 54.12% Train). Generation: 29s, Training: 35s. Estimated remaining time: 45h 6m 32s. Estimated total time: 54h 17m 33s. Time estimates for 10 more iterations: 10m 51s, 100 more iterations: 1h 48m 35s, 500 more iterations: 9h 2m 55s. +[2026-03-26 06:16:42,086][__main__][INFO] - Starting iteration 582. +[2026-03-26 06:16:42,514][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 58 and human policies 1. +[2026-03-26 06:16:42,515][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:17:12,385][__main__][INFO] - Number of regex retries in iteration 582: 0 +[2026-03-26 06:17:12,386][__main__][INFO] - agents played in iteration 582 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:17:13,163][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.45%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:17:13,187][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.45%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:17:13,211][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.45%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:17:13,235][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.45%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:17:13,236][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:17:13,237][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:17:14,038][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:17:14,510][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:17:15,017][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:17:15,518][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:17:16,020][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:17:16,525][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:17:17,035][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:17:17,538][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:17:18,043][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:17:18,542][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:17:19,041][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:17:19,546][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:17:20,057][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:17:20,558][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:17:21,059][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:17:21,558][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:17:22,060][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:17:22,559][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:17:23,067][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:17:23,579][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:17:24,083][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:17:24,586][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:17:25,088][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:17:25,587][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:17:26,088][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:17:26,587][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:17:27,088][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:17:27,598][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:17:28,106][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:17:28,608][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:17:29,112][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:17:29,610][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:17:30,109][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:17:30,609][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:17:31,110][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:17:31,609][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:17:32,110][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:17:32,610][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:17:33,110][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:17:33,610][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:17:34,110][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:17:34,609][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:17:35,115][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:17:35,617][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:17:36,117][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:17:36,618][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:17:37,119][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:17:37,618][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:17:38,116][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:17:38,614][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:17:39,113][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:17:39,613][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:17:40,111][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:17:40,608][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:17:41,105][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:17:41,602][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:17:42,099][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:17:42,597][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:17:43,095][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:17:43,592][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:17:44,090][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:17:44,587][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:17:45,084][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:17:45,582][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:17:46,110][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10112 tokens. +[2026-03-26 06:17:46,929][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.25%, Current % of VRAM taken: 60.67%, Block Peak % of device VRAM: 62.16%, ΔTime: 00:00:32 +[2026-03-26 06:17:47,684][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:17:47,686][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:17:47,688][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:17:48,459][__main__][INFO] - Iteration 583 took 1m 5s (45.30% Gen, 53.53% Train). Generation: 29s, Training: 35s. Estimated remaining time: 45h 45m 8s. Estimated total time: 54h 57m 16s. Time estimates for 10 more iterations: 10m 59s, 100 more iterations: 1h 49m 54s, 500 more iterations: 9h 9m 32s. +[2026-03-26 06:17:48,461][__main__][INFO] - Starting iteration 583. +[2026-03-26 06:17:48,862][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 58 and human policies 1. +[2026-03-26 06:17:48,863][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:18:06,199][__main__][INFO] - Number of regex retries in iteration 583: 0 +[2026-03-26 06:18:06,200][__main__][INFO] - agents played in iteration 583 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:18:06,979][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:18:07,003][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:18:07,026][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:18:07,049][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:18:07,050][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:18:07,050][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:18:07,836][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:18:08,296][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:18:08,798][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:18:09,299][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:18:09,796][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:18:10,295][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:18:10,794][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:18:11,296][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:18:11,799][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:18:12,307][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:18:12,814][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:18:13,317][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:18:13,818][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:18:14,320][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:18:14,821][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:18:15,323][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:18:15,824][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:18:16,325][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:18:16,827][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:18:17,326][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:18:17,825][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:18:18,324][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:18:18,823][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:18:19,326][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:18:19,831][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:18:20,343][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:18:20,847][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:18:21,352][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:18:21,857][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:18:22,357][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:18:22,860][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:18:23,363][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:18:23,869][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:18:24,373][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:18:24,873][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:18:25,377][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:18:25,879][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:18:26,382][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:18:26,880][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:18:27,383][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:18:27,881][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:18:28,380][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:18:28,877][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:18:29,374][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:18:29,870][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:18:30,366][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:18:30,862][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:18:31,360][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:18:31,857][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:18:32,360][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:18:32,859][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:18:33,361][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:18:33,860][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:18:34,386][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:18:34,885][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:18:35,388][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:18:35,889][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:18:36,387][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:18:36,885][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:18:37,383][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:18:37,882][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:18:38,380][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:18:38,879][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:18:39,376][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:18:39,875][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10110 tokens. +[2026-03-26 06:18:40,699][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.57%, Block Peak % of device VRAM: 62.18%, ΔTime: 00:00:32 +[2026-03-26 06:18:41,443][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:18:41,446][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:18:41,447][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:18:42,212][__main__][INFO] - Iteration 584 took 53s (32.50% Gen, 66.07% Train). Generation: 17s, Training: 35s. Estimated remaining time: 35h 14m 27s. Estimated total time: 44h 27m 29s. Time estimates for 10 more iterations: 8m 53s, 100 more iterations: 1h 28m 54s, 500 more iterations: 7h 24m 34s. +[2026-03-26 06:18:42,214][__main__][INFO] - Starting iteration 584. +[2026-03-26 06:18:42,635][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 58 and human policies 1. +[2026-03-26 06:18:42,636][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:18:56,961][__main__][INFO] - Number of regex retries in iteration 584: 0 +[2026-03-26 06:18:56,962][__main__][INFO] - agents played in iteration 584 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:18:57,728][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:18:57,751][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:18:57,775][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:18:57,799][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:18:57,800][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:18:57,800][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:18:58,590][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:18:59,054][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:18:59,571][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:19:00,075][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:19:00,578][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:19:01,084][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:19:01,588][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:19:02,099][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:19:02,604][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:19:03,110][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:19:03,610][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:19:04,116][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:19:04,617][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:19:05,119][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:19:05,626][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:19:06,137][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:19:06,642][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:19:07,145][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:19:07,646][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:19:08,146][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:19:08,645][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:19:09,144][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:19:09,643][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:19:10,142][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:19:10,645][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:19:11,143][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:19:11,641][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:19:12,141][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:19:12,643][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:19:13,147][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:19:13,657][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:19:14,158][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:19:14,658][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:19:15,160][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:19:15,661][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:19:16,165][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:19:16,665][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:19:17,166][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:19:17,673][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:19:18,182][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:19:18,688][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:19:19,188][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:19:19,692][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:19:20,191][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:19:20,691][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:19:21,191][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:19:21,689][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:19:22,187][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:19:22,686][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:19:23,183][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:19:23,684][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:19:24,182][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:19:24,680][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:19:25,178][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:19:25,677][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:19:26,176][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:19:26,675][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:19:27,174][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:19:27,671][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:19:28,170][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:19:28,668][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:19:29,165][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:19:29,662][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:19:30,163][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:19:30,661][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10040 tokens. +[2026-03-26 06:19:31,492][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.09%, Current % of VRAM taken: 60.51%, Block Peak % of device VRAM: 62.02%, ΔTime: 00:00:32 +[2026-03-26 06:19:32,235][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:19:32,237][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:19:32,239][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:19:32,969][__main__][INFO] - Iteration 585 took 50s (28.46% Gen, 70.08% Train). Generation: 14s, Training: 35s. Estimated remaining time: 32h 42m 50s. Estimated total time: 41h 56m 42s. Time estimates for 10 more iterations: 8m 23s, 100 more iterations: 1h 23m 53s, 500 more iterations: 6h 59m 27s. +[2026-03-26 06:19:32,971][__main__][INFO] - Starting iteration 585. +[2026-03-26 06:19:33,381][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 58 and human policies 1. +[2026-03-26 06:19:33,381][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:19:55,046][__main__][INFO] - Number of regex retries in iteration 585: 0 +[2026-03-26 06:19:55,046][__main__][INFO] - agents played in iteration 585 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:19:55,820][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:19:55,844][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:19:55,868][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:19:55,892][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:19:55,892][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:19:55,893][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:19:56,678][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:19:57,132][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:19:57,644][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:19:58,152][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:19:58,662][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:19:59,165][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:19:59,667][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:20:00,169][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:20:00,671][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:20:01,176][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:20:01,678][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:20:02,181][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:20:02,682][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:20:03,188][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:20:03,692][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:20:04,194][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:20:04,694][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:20:05,197][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:20:05,700][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:20:06,203][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:20:06,710][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:20:07,218][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:20:07,722][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:20:08,222][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:20:08,723][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:20:09,226][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:20:09,728][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:20:10,230][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:20:10,732][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:20:11,235][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:20:11,737][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:20:12,238][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:20:12,739][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:20:13,237][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:20:13,736][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:20:14,238][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:20:14,737][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:20:15,235][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:20:15,736][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:20:16,237][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:20:16,742][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:20:17,244][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:20:17,748][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:20:18,259][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:20:18,759][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:20:19,258][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:20:19,757][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:20:20,259][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:20:20,760][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:20:21,263][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:20:21,762][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:20:22,262][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:20:22,762][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:20:23,260][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:20:23,760][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:20:24,259][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:20:24,759][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:20:25,257][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:20:25,756][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:20:26,255][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:20:26,757][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:20:27,264][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:20:27,768][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:20:28,268][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:20:28,776][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10122 tokens. +[2026-03-26 06:20:29,620][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 60.49%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:32 +[2026-03-26 06:20:30,387][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:20:30,389][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:20:30,391][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:20:31,176][__main__][INFO] - Iteration 586 took 57s (37.49% Gen, 61.15% Train). Generation: 21s, Training: 35s. Estimated remaining time: 38h 54m 58s. Estimated total time: 48h 9m 49s. Time estimates for 10 more iterations: 9m 37s, 100 more iterations: 1h 36m 19s, 500 more iterations: 8h 1m 38s. +[2026-03-26 06:20:31,178][__main__][INFO] - Starting iteration 586. +[2026-03-26 06:20:31,591][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 58 and human policies 1. +[2026-03-26 06:20:31,592][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:20:35,925][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 06:20:45,998][__main__][INFO] - Number of regex retries in iteration 586: 1 +[2026-03-26 06:20:45,999][__main__][INFO] - agents played in iteration 586 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:20:46,758][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:20:46,782][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:20:46,807][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:20:46,831][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:20:46,832][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:20:46,833][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:20:47,621][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:20:48,081][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:20:48,585][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:20:49,082][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:20:49,580][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:20:50,077][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:20:50,575][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:20:51,077][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:20:51,592][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:20:52,093][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:20:52,594][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:20:53,093][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:20:53,594][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:20:54,099][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:20:54,610][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:20:55,111][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:20:55,611][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:20:56,115][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:20:56,616][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:20:57,123][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:20:57,624][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:20:58,124][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:20:58,623][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:20:59,122][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:20:59,622][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:21:00,121][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:21:00,649][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:21:01,164][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:21:01,666][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:21:02,170][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:21:02,671][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:21:03,171][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:21:03,671][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:21:04,170][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:21:04,670][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:21:05,177][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:21:05,690][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:21:06,190][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:21:06,693][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:21:07,194][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:21:07,693][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:21:08,193][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:21:08,691][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:21:09,189][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:21:09,687][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:21:10,187][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:21:10,685][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:21:11,183][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:21:11,680][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:21:12,179][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:21:12,678][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:21:13,181][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:21:13,694][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:21:14,202][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:21:14,703][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:21:15,208][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:21:15,707][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:21:16,206][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:21:16,704][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:21:17,205][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:21:17,704][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:21:18,202][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:21:18,702][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:21:19,201][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:21:19,700][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9960 tokens. +[2026-03-26 06:21:20,539][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 60.51%, Block Peak % of device VRAM: 62.05%, ΔTime: 00:00:32 +[2026-03-26 06:21:21,383][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:21:21,386][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:21:21,387][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:21:22,147][__main__][INFO] - Iteration 587 took 50s (28.50% Gen, 70.00% Train). Generation: 14s, Training: 35s. Estimated remaining time: 32h 52m 8s. Estimated total time: 42h 7m 49s. Time estimates for 10 more iterations: 8m 25s, 100 more iterations: 1h 24m 15s, 500 more iterations: 7h 1m 18s. +[2026-03-26 06:21:22,149][__main__][INFO] - Starting iteration 587. +[2026-03-26 06:21:22,562][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 58 and human policies 1. +[2026-03-26 06:21:22,562][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:21:37,958][__main__][INFO] - Number of regex retries in iteration 587: 0 +[2026-03-26 06:21:37,959][__main__][INFO] - agents played in iteration 587 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:21:38,739][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:21:38,764][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:21:38,787][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:21:38,811][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:21:38,811][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:21:38,812][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:21:39,613][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:21:40,069][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:21:40,581][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:21:41,090][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:21:41,598][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:21:42,101][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:21:42,603][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:21:43,106][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:21:43,607][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:21:44,108][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:21:44,616][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:21:45,118][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:21:45,625][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:21:46,125][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:21:46,625][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:21:47,126][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:21:47,629][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:21:48,142][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:21:48,650][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:21:49,152][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:21:49,660][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:21:50,166][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:21:50,670][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:21:51,173][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:21:51,677][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:21:52,179][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:21:52,679][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:21:53,179][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:21:53,677][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:21:54,176][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:21:54,674][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:21:55,171][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:21:55,669][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:21:56,167][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:21:56,667][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:21:57,169][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:21:57,670][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:21:58,171][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:21:58,672][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:21:59,173][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:21:59,671][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:22:00,173][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:22:00,671][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:22:01,168][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:22:01,667][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:22:02,166][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:22:02,667][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:22:03,164][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:22:03,664][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:22:04,160][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:22:04,658][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:22:05,156][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:22:05,654][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:22:06,151][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:22:06,648][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:22:07,146][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:22:07,643][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:22:08,143][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:22:08,640][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:22:09,138][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:22:09,636][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:22:10,134][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:22:10,631][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:22:11,129][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:22:11,631][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10229 tokens. +[2026-03-26 06:22:12,470][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 60.60%, Block Peak % of device VRAM: 62.17%, ΔTime: 00:00:32 +[2026-03-26 06:22:13,217][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:22:13,220][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:22:13,221][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:22:13,947][__main__][INFO] - Iteration 588 took 51s (29.96% Gen, 68.62% Train). Generation: 15s, Training: 35s. Estimated remaining time: 33h 32m 44s. Estimated total time: 42h 49m 18s. Time estimates for 10 more iterations: 8m 33s, 100 more iterations: 1h 25m 38s, 500 more iterations: 7h 8m 13s. +[2026-03-26 06:22:13,949][__main__][INFO] - Starting iteration 588. +[2026-03-26 06:22:14,365][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 58 and human policies 1. +[2026-03-26 06:22:14,366][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:22:19,133][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 06:22:30,149][__main__][INFO] - Number of regex retries in iteration 588: 1 +[2026-03-26 06:22:30,150][__main__][INFO] - agents played in iteration 588 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:22:30,938][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:22:30,962][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:22:30,985][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:22:31,009][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:22:31,009][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:22:31,010][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:22:31,798][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:22:32,255][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:22:32,764][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:22:33,263][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:22:33,768][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:22:34,279][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:22:34,779][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:22:35,279][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:22:35,781][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:22:36,283][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:22:36,784][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:22:37,289][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:22:37,787][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:22:38,287][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:22:38,785][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:22:39,285][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:22:39,787][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:22:40,287][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:22:40,786][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:22:41,286][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:22:41,784][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:22:42,304][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:22:42,804][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:22:43,308][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:22:43,807][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:22:44,315][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:22:44,818][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:22:45,322][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:22:45,825][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:22:46,326][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:22:46,831][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:22:47,330][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:22:47,834][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:22:48,335][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:22:48,836][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:22:49,338][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:22:49,840][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:22:50,341][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:22:50,842][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:22:51,345][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:22:51,846][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:22:52,347][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:22:52,847][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:22:53,347][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:22:53,847][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:22:54,347][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:22:54,847][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:22:55,356][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:22:55,857][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:22:56,355][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:22:56,855][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:22:57,353][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:22:57,852][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:22:58,351][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:22:58,851][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:22:59,349][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:22:59,851][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:23:00,347][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:23:00,844][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:23:01,342][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:23:01,841][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:23:02,339][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:23:02,840][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:23:03,341][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:23:03,839][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10085 tokens. +[2026-03-26 06:23:04,685][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.04%, Current % of VRAM taken: 60.47%, Block Peak % of device VRAM: 62.10%, ΔTime: 00:00:32 +[2026-03-26 06:23:05,439][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:23:05,442][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:23:05,444][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:23:06,180][__main__][INFO] - Iteration 589 took 51s (30.46% Gen, 68.11% Train). Generation: 15s, Training: 35s. Estimated remaining time: 33h 53m 20s. Estimated total time: 43h 10m 46s. Time estimates for 10 more iterations: 8m 38s, 100 more iterations: 1h 26m 21s, 500 more iterations: 7h 11m 47s. +[2026-03-26 06:23:06,183][__main__][INFO] - Starting iteration 589. +[2026-03-26 06:23:06,598][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 58 and human policies 1. +[2026-03-26 06:23:06,598][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:23:10,956][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 06:23:21,383][__main__][INFO] - Number of regex retries in iteration 589: 1 +[2026-03-26 06:23:21,384][__main__][INFO] - agents played in iteration 589 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:23:22,159][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:23:22,184][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:23:22,209][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:23:22,233][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:23:22,234][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:23:22,234][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:23:23,026][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:23:23,485][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:23:23,994][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:23:24,506][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:23:25,011][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:23:25,515][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:23:26,016][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:23:26,517][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:23:27,020][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:23:27,520][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:23:28,028][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:23:28,533][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:23:29,043][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:23:29,545][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:23:30,047][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:23:30,547][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:23:31,052][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:23:31,555][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:23:32,060][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:23:32,562][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:23:33,067][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:23:33,573][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:23:34,085][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:23:34,591][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:23:35,096][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:23:35,598][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:23:36,099][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:23:36,600][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:23:37,101][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:23:37,601][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:23:38,104][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:23:38,604][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:23:39,105][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:23:39,609][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:23:40,114][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:23:40,621][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:23:41,125][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:23:41,631][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:23:42,134][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:23:42,637][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:23:43,141][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:23:43,641][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:23:44,142][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:23:44,642][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:23:45,143][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:23:45,644][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:23:46,144][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:23:46,644][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:23:47,149][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:23:47,650][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:23:48,148][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:23:48,645][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:23:49,144][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:23:49,641][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:23:50,141][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:23:50,640][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:23:51,142][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:23:51,642][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:23:52,143][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:23:52,644][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:23:53,143][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:23:53,644][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:23:54,145][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:23:54,644][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:23:55,142][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10127 tokens. +[2026-03-26 06:23:55,971][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 60.56%, Block Peak % of device VRAM: 62.16%, ΔTime: 00:00:32 +[2026-03-26 06:23:56,722][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:23:56,724][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:23:56,726][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:23:57,466][__main__][INFO] - Iteration 590 took 50s (29.07% Gen, 69.48% Train). Generation: 14s, Training: 35s. Estimated remaining time: 33h 5m 9s. Estimated total time: 42h 23m 27s. Time estimates for 10 more iterations: 8m 28s, 100 more iterations: 1h 24m 46s, 500 more iterations: 7h 3m 54s. +[2026-03-26 06:23:57,468][__main__][INFO] - Starting iteration 590. +[2026-03-26 06:23:57,881][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 58 and human policies 1. +[2026-03-26 06:23:57,882][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:24:12,802][__main__][INFO] - Number of regex retries in iteration 590: 0 +[2026-03-26 06:24:12,803][__main__][INFO] - agents played in iteration 590 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:24:13,566][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:24:13,590][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:24:13,613][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:24:13,637][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:24:13,638][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:24:13,638][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:24:14,436][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:24:14,888][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:24:15,398][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:24:15,903][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:24:16,409][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:24:16,916][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:24:17,420][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:24:17,927][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:24:18,430][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:24:18,933][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:24:19,435][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:24:19,939][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:24:20,445][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:24:20,946][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:24:21,449][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:24:21,950][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:24:22,451][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:24:22,951][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:24:23,451][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:24:23,952][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:24:24,454][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:24:24,959][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:24:25,462][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:24:25,965][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:24:26,468][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:24:26,967][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:24:27,469][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:24:27,974][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:24:28,473][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:24:28,978][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:24:29,480][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:24:29,986][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:24:30,488][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:24:30,997][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:24:31,498][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:24:32,004][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:24:32,509][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:24:33,012][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:24:33,513][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:24:34,014][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:24:34,517][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:24:35,017][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:24:35,518][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:24:36,017][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:24:36,518][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:24:37,019][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:24:37,524][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:24:38,027][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:24:38,526][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:24:39,025][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:24:39,524][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:24:40,024][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:24:40,523][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:24:41,022][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:24:41,524][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:24:42,027][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:24:42,526][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:24:43,024][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:24:43,522][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:24:44,020][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:24:44,518][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:24:45,019][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:24:45,517][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:24:46,015][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:24:46,518][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10100 tokens. +[2026-03-26 06:24:47,368][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.57%, Block Peak % of device VRAM: 62.07%, ΔTime: 00:00:32 +[2026-03-26 06:24:48,123][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:24:48,126][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:24:48,127][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:24:49,473][__main__][INFO] - Iteration 591 took 51s (28.92% Gen, 68.47% Train). Generation: 14s, Training: 35s. Estimated remaining time: 33h 40m 27s. Estimated total time: 42h 59m 36s. Time estimates for 10 more iterations: 8m 35s, 100 more iterations: 1h 25m 59s, 500 more iterations: 7h 9m 56s. +[2026-03-26 06:24:49,475][__main__][INFO] - Starting iteration 591. +[2026-03-26 06:24:49,886][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 59 and human policies 1. +[2026-03-26 06:24:49,886][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:25:05,309][__main__][INFO] - Number of regex retries in iteration 591: 0 +[2026-03-26 06:25:05,310][__main__][INFO] - agents played in iteration 591 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:25:06,095][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:25:06,118][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:25:06,142][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:25:06,165][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:25:06,166][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:25:06,166][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:25:06,965][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:25:07,434][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:25:07,951][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:25:08,453][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:25:08,956][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:25:09,465][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:25:09,980][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:25:10,484][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:25:10,986][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:25:11,489][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:25:11,993][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:25:12,495][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:25:12,997][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:25:13,499][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:25:14,000][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:25:14,501][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:25:15,003][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:25:15,503][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:25:16,003][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:25:16,504][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:25:17,005][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:25:17,507][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:25:18,008][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:25:18,510][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:25:19,008][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:25:19,506][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:25:20,005][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:25:20,503][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:25:21,002][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:25:21,503][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:25:22,002][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:25:22,504][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:25:23,001][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:25:23,500][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:25:24,000][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:25:24,499][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:25:24,996][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:25:25,494][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:25:25,991][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:25:26,488][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:25:26,987][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:25:27,485][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:25:27,985][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:25:28,482][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:25:28,981][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:25:29,482][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:25:29,982][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:25:30,480][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:25:30,978][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:25:31,479][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:25:31,983][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:25:32,480][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:25:32,982][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:25:33,479][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:25:33,978][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:25:34,479][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:25:34,977][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:25:35,474][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:25:35,976][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:25:36,477][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:25:36,976][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:25:37,473][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:25:37,970][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:25:38,466][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:25:38,963][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10096 tokens. +[2026-03-26 06:25:39,784][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 60.52%, Block Peak % of device VRAM: 62.10%, ΔTime: 00:00:32 +[2026-03-26 06:25:40,533][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:25:40,535][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:25:40,537][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:25:41,306][__main__][INFO] - Iteration 592 took 51s (29.99% Gen, 68.51% Train). Generation: 15s, Training: 35s. Estimated remaining time: 33h 31m 0s. Estimated total time: 42h 51m 1s. Time estimates for 10 more iterations: 8m 34s, 100 more iterations: 1h 25m 42s, 500 more iterations: 7h 8m 30s. +[2026-03-26 06:25:41,308][__main__][INFO] - Starting iteration 592. +[2026-03-26 06:25:41,714][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 59 and human policies 1. +[2026-03-26 06:25:41,714][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:25:56,762][__main__][INFO] - Number of regex retries in iteration 592: 0 +[2026-03-26 06:25:56,763][__main__][INFO] - agents played in iteration 592 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:25:57,538][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:25:57,565][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:25:57,591][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:25:57,617][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:25:57,618][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:25:57,619][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:25:58,430][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:25:58,898][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:25:59,407][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:25:59,913][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:26:00,419][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:26:00,924][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:26:01,428][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:26:01,934][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:26:02,441][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:26:02,946][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:26:03,455][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:26:03,965][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:26:04,471][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:26:04,971][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:26:05,472][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:26:05,973][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:26:06,473][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:26:06,974][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:26:07,476][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:26:07,980][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:26:08,479][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:26:08,979][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:26:09,483][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:26:09,985][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:26:10,484][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:26:10,987][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:26:11,487][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:26:11,986][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:26:12,485][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:26:12,984][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:26:13,484][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:26:13,983][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:26:14,482][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:26:14,985][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:26:15,486][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:26:15,986][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:26:16,486][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:26:16,986][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:26:17,486][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:26:17,986][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:26:18,488][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:26:18,984][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:26:19,481][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:26:19,979][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:26:20,477][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:26:20,974][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:26:21,474][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:26:21,974][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:26:22,471][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:26:22,968][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:26:23,469][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:26:23,966][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:26:24,465][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:26:24,964][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:26:25,462][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:26:25,963][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:26:26,460][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:26:26,962][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:26:27,459][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:26:27,955][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:26:28,451][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:26:28,948][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:26:29,444][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:26:29,941][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:26:30,438][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10131 tokens. +[2026-03-26 06:26:31,268][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 60.57%, Block Peak % of device VRAM: 62.43%, ΔTime: 00:00:32 +[2026-03-26 06:26:31,991][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:26:31,993][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:26:31,995][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:26:32,724][__main__][INFO] - Iteration 593 took 51s (29.50% Gen, 69.07% Train). Generation: 15s, Training: 35s. Estimated remaining time: 33h 9m 40s. Estimated total time: 42h 30m 32s. Time estimates for 10 more iterations: 8m 30s, 100 more iterations: 1h 25m 1s, 500 more iterations: 7h 5m 5s. +[2026-03-26 06:26:32,726][__main__][INFO] - Starting iteration 593. +[2026-03-26 06:26:33,133][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 59 and human policies 1. +[2026-03-26 06:26:33,134][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:26:48,225][__main__][INFO] - Number of regex retries in iteration 593: 0 +[2026-03-26 06:26:48,226][__main__][INFO] - agents played in iteration 593 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:26:49,011][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:26:49,037][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:26:49,062][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:26:49,087][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:26:49,088][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:26:49,088][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:26:49,881][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:26:50,343][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:26:50,850][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:26:51,349][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:26:51,851][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:26:52,355][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:26:52,863][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:26:53,370][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:26:53,877][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:26:54,379][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:26:54,886][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:26:55,387][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:26:55,889][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:26:56,389][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:26:56,890][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:26:57,392][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:26:57,894][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:26:58,398][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:26:58,903][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:26:59,410][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:26:59,918][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:27:00,428][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:27:00,933][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:27:01,435][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:27:01,940][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:27:02,442][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:27:02,943][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:27:03,444][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:27:03,947][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:27:04,447][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:27:04,947][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:27:05,447][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:27:05,948][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:27:06,449][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:27:06,954][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:27:07,456][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:27:07,963][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:27:08,465][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:27:08,968][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:27:09,467][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:27:09,973][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:27:10,472][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:27:10,973][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:27:11,473][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:27:11,976][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:27:12,477][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:27:12,979][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:27:13,478][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:27:13,981][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:27:14,484][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:27:14,986][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:27:15,486][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:27:15,988][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:27:16,487][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:27:16,985][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:27:17,487][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:27:17,985][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:27:18,487][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:27:18,988][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:27:19,491][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:27:19,997][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:27:20,502][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:27:21,004][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:27:21,504][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:27:22,005][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10089 tokens. +[2026-03-26 06:27:22,848][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 60.49%, Block Peak % of device VRAM: 62.13%, ΔTime: 00:00:32 +[2026-03-26 06:27:23,600][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:27:23,603][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:27:23,604][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:27:24,466][__main__][INFO] - Iteration 594 took 51s (29.40% Gen, 68.92% Train). Generation: 15s, Training: 35s. Estimated remaining time: 33h 24m 55s. Estimated total time: 42h 46m 39s. Time estimates for 10 more iterations: 8m 33s, 100 more iterations: 1h 25m 33s, 500 more iterations: 7h 7m 46s. +[2026-03-26 06:27:24,468][__main__][INFO] - Starting iteration 594. +[2026-03-26 06:27:24,878][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 59 and human policies 1. +[2026-03-26 06:27:24,879][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:27:31,399][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 06:27:41,152][__main__][INFO] - Number of regex retries in iteration 594: 1 +[2026-03-26 06:27:41,153][__main__][INFO] - agents played in iteration 594 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:27:41,921][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:27:41,945][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:27:41,969][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:27:41,993][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:27:41,993][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:27:41,994][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:27:42,795][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:27:43,253][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:27:43,759][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:27:44,262][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:27:44,762][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:27:45,264][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:27:45,767][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:27:46,272][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:27:46,775][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:27:47,277][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:27:47,774][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:27:48,273][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:27:48,775][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:27:49,276][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:27:49,780][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:27:50,282][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:27:50,786][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:27:51,290][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:27:51,801][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:27:52,301][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:27:52,812][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:27:53,312][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:27:53,813][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:27:54,316][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:27:54,816][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:27:55,315][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:27:55,817][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:27:56,318][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:27:56,818][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:27:57,319][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:27:57,818][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:27:58,317][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:27:58,815][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:27:59,314][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:27:59,813][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:28:00,312][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:28:00,811][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:28:01,313][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:28:01,811][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:28:02,313][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:28:02,810][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:28:03,312][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:28:03,809][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:28:04,305][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:28:04,801][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:28:05,298][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:28:05,798][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:28:06,297][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:28:06,796][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:28:07,294][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:28:07,794][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:28:08,294][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:28:08,804][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:28:09,304][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:28:09,802][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:28:10,307][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:28:10,805][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:28:11,304][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:28:11,803][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:28:12,301][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:28:12,799][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:28:13,297][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:28:13,796][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:28:14,293][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:28:14,792][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10129 tokens. +[2026-03-26 06:28:15,635][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 +[2026-03-26 06:28:16,352][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:28:16,355][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:28:16,357][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:28:17,116][__main__][INFO] - Iteration 595 took 52s (31.15% Gen, 67.39% Train). Generation: 16s, Training: 35s. Estimated remaining time: 34h 9m 19s. Estimated total time: 43h 31m 56s. Time estimates for 10 more iterations: 8m 42s, 100 more iterations: 1h 27m 3s, 500 more iterations: 7h 15m 19s. +[2026-03-26 06:28:17,119][__main__][INFO] - Starting iteration 595. +[2026-03-26 06:28:17,521][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 59 and human policies 1. +[2026-03-26 06:28:17,521][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:28:32,099][__main__][INFO] - Number of regex retries in iteration 595: 0 +[2026-03-26 06:28:32,100][__main__][INFO] - agents played in iteration 595 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:28:32,867][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:28:32,891][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:28:32,914][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:28:32,937][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:28:32,938][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:28:32,938][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:28:33,732][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:28:34,186][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:28:34,688][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:28:35,191][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:28:35,701][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:28:36,215][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:28:36,722][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:28:37,224][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:28:37,725][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:28:38,228][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:28:38,736][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:28:39,246][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:28:39,761][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:28:40,263][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:28:40,767][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:28:41,271][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:28:41,794][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:28:42,297][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:28:42,798][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:28:43,299][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:28:43,799][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:28:44,299][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:28:44,799][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:28:45,302][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:28:45,802][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:28:46,302][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:28:46,806][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:28:47,317][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:28:47,830][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:28:48,335][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:28:48,841][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:28:49,347][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:28:49,848][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:28:50,348][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:28:50,849][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:28:51,353][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:28:51,861][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:28:52,366][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:28:52,869][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:28:53,371][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:28:53,878][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:28:54,378][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:28:54,877][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:28:55,376][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:28:55,875][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:28:56,373][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:28:56,872][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:28:57,376][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:28:57,874][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:28:58,376][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:28:58,878][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:28:59,377][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:28:59,876][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:29:00,376][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:29:00,876][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:29:01,374][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:29:01,875][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:29:02,376][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:29:02,873][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:29:03,371][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:29:03,868][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:29:04,366][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:29:04,864][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:29:05,362][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:29:05,861][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10110 tokens. +[2026-03-26 06:29:06,687][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.20%, ΔTime: 00:00:32 +[2026-03-26 06:29:07,454][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:29:07,456][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:29:07,457][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:29:08,187][__main__][INFO] - Iteration 596 took 50s (28.77% Gen, 69.78% Train). Generation: 14s, Training: 35s. Estimated remaining time: 32h 49m 52s. Estimated total time: 42h 13m 20s. Time estimates for 10 more iterations: 8m 26s, 100 more iterations: 1h 24m 26s, 500 more iterations: 7h 2m 13s. +[2026-03-26 06:29:08,189][__main__][INFO] - Starting iteration 596. +[2026-03-26 06:29:08,596][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 59 and human policies 1. +[2026-03-26 06:29:08,596][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:29:23,059][__main__][INFO] - Number of regex retries in iteration 596: 0 +[2026-03-26 06:29:23,060][__main__][INFO] - agents played in iteration 596 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:29:23,891][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:29:23,915][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:29:23,938][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:29:23,962][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:29:23,963][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:29:23,963][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:29:24,752][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:29:25,213][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:29:25,716][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:29:26,217][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:29:26,722][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:29:27,222][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:29:27,723][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:29:28,223][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:29:28,737][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:29:29,252][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:29:29,759][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:29:30,261][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:29:30,764][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:29:31,267][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:29:31,775][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:29:32,292][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:29:32,796][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:29:33,299][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:29:33,799][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:29:34,300][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:29:34,801][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:29:35,303][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:29:35,804][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:29:36,307][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:29:36,806][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:29:37,308][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:29:37,811][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:29:38,320][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:29:38,835][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:29:39,342][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:29:39,847][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:29:40,351][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:29:40,856][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:29:41,356][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:29:41,859][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:29:42,369][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:29:42,882][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:29:43,385][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:29:43,887][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:29:44,386][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:29:44,887][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:29:45,386][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:29:45,887][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:29:46,392][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:29:46,895][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:29:47,407][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:29:47,909][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:29:48,415][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:29:48,919][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:29:49,423][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:29:49,923][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:29:50,425][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:29:50,926][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:29:51,427][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:29:51,933][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:29:52,441][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:29:52,951][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:29:53,453][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:29:53,955][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:29:54,457][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:29:54,958][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:29:55,461][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:29:55,961][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:29:56,461][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:29:56,961][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10232 tokens. +[2026-03-26 06:29:57,812][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:33 +[2026-03-26 06:29:58,567][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:29:58,570][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:29:58,572][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:29:59,335][__main__][INFO] - Iteration 597 took 50s (28.51% Gen, 69.99% Train). Generation: 14s, Training: 35s. Estimated remaining time: 32h 52m 41s. Estimated total time: 42h 17m 0s. Time estimates for 10 more iterations: 8m 27s, 100 more iterations: 1h 24m 34s, 500 more iterations: 7h 2m 50s. +[2026-03-26 06:29:59,337][__main__][INFO] - Starting iteration 597. +[2026-03-26 06:29:59,747][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 59 and human policies 1. +[2026-03-26 06:29:59,748][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:30:15,427][__main__][INFO] - Number of regex retries in iteration 597: 0 +[2026-03-26 06:30:15,428][__main__][INFO] - agents played in iteration 597 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:30:16,205][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:30:16,229][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:30:16,252][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:30:16,275][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:30:16,276][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:30:16,277][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:30:17,067][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:30:17,524][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:30:18,028][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:30:18,524][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:30:19,021][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:30:19,524][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:30:20,030][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:30:20,533][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:30:21,037][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:30:21,542][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:30:22,044][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:30:22,544][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:30:23,045][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:30:23,547][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:30:24,047][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:30:24,547][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:30:25,046][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:30:25,548][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:30:26,047][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:30:26,552][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:30:27,055][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:30:27,555][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:30:28,059][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:30:28,559][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:30:29,065][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:30:29,571][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:30:30,073][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:30:30,578][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:30:31,079][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:30:31,580][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:30:32,085][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:30:32,609][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:30:33,113][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:30:33,611][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:30:34,110][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:30:34,608][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:30:35,107][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:30:35,605][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:30:36,103][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:30:36,601][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:30:37,098][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:30:37,598][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:30:38,097][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:30:38,597][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:30:39,096][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:30:39,595][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:30:40,093][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:30:40,595][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:30:41,093][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:30:41,590][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:30:42,086][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:30:42,584][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:30:43,081][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:30:43,578][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:30:44,079][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:30:44,575][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:30:45,071][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:30:45,568][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:30:46,067][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:30:46,564][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:30:47,062][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:30:47,559][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:30:48,057][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:30:48,560][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:30:49,059][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10051 tokens. +[2026-03-26 06:30:49,887][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 60.52%, Block Peak % of device VRAM: 62.15%, ΔTime: 00:00:32 +[2026-03-26 06:30:50,676][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:30:50,678][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:30:50,680][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:30:51,440][__main__][INFO] - Iteration 598 took 51s (30.33% Gen, 68.19% Train). Generation: 15s, Training: 35s. Estimated remaining time: 33h 39m 30s. Estimated total time: 43h 4m 41s. Time estimates for 10 more iterations: 8m 36s, 100 more iterations: 1h 26m 9s, 500 more iterations: 7h 10m 46s. +[2026-03-26 06:30:51,442][__main__][INFO] - Starting iteration 598. +[2026-03-26 06:30:51,858][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 59 and human policies 1. +[2026-03-26 06:30:51,859][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:31:06,122][__main__][INFO] - Number of regex retries in iteration 598: 0 +[2026-03-26 06:31:06,122][__main__][INFO] - agents played in iteration 598 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:31:06,900][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:31:06,925][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:31:06,949][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:31:06,974][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:31:06,974][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:31:06,975][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:31:07,771][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:31:08,226][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:31:08,731][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:31:09,233][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:31:09,745][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:31:10,254][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:31:10,758][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:31:11,260][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:31:11,768][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:31:12,269][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:31:12,774][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:31:13,274][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:31:13,781][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:31:14,288][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:31:14,796][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:31:15,299][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:31:15,806][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:31:16,309][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:31:16,812][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:31:17,321][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:31:17,824][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:31:18,325][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:31:18,827][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:31:19,328][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:31:19,831][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:31:20,330][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:31:20,827][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:31:21,325][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:31:21,824][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:31:22,322][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:31:22,820][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:31:23,318][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:31:23,817][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:31:24,317][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:31:24,819][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:31:25,319][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:31:25,820][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:31:26,323][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:31:26,825][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:31:27,324][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:31:27,823][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:31:28,325][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:31:28,823][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:31:29,325][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:31:29,824][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:31:30,323][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:31:30,825][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:31:31,326][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:31:31,830][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:31:32,329][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:31:32,826][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:31:33,325][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:31:33,825][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:31:34,325][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:31:34,824][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:31:35,322][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:31:35,822][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:31:36,321][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:31:36,822][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:31:37,320][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:31:37,819][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:31:38,318][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:31:38,817][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:31:39,319][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:31:39,818][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10105 tokens. +[2026-03-26 06:31:40,653][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 60.49%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:32 +[2026-03-26 06:31:41,405][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:31:41,408][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:31:41,409][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:31:42,253][__main__][INFO] - Iteration 599 took 50s (28.30% Gen, 70.02% Train). Generation: 14s, Training: 35s. Estimated remaining time: 32h 33m 45s. Estimated total time: 41h 59m 47s. Time estimates for 10 more iterations: 8m 23s, 100 more iterations: 1h 23m 59s, 500 more iterations: 6h 59m 57s. +[2026-03-26 06:31:42,255][__main__][INFO] - Starting iteration 599. +[2026-03-26 06:31:42,660][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 59 and human policies 1. +[2026-03-26 06:31:42,661][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:31:56,462][__main__][INFO] - Number of regex retries in iteration 599: 0 +[2026-03-26 06:31:56,462][__main__][INFO] - agents played in iteration 599 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:31:57,234][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:31:57,258][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:31:57,282][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:31:57,306][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:31:57,307][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:31:57,307][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:31:58,100][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:31:58,556][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:31:59,061][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:31:59,566][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:32:00,077][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:32:00,587][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:32:01,090][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:32:01,591][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:32:02,103][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:32:02,603][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:32:03,106][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:32:03,605][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:32:04,104][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:32:04,602][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:32:05,101][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:32:05,601][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:32:06,100][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:32:06,599][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:32:07,103][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:32:07,603][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:32:08,104][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:32:08,603][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:32:09,104][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:32:09,603][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:32:10,103][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:32:10,605][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:32:11,112][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:32:11,648][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:32:12,154][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:32:12,654][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:32:13,158][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:32:13,660][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:32:14,166][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:32:14,666][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:32:15,166][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:32:15,665][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:32:16,166][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:32:16,665][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:32:17,165][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:32:17,665][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:32:18,164][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:32:18,662][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:32:19,164][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:32:19,662][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:32:20,164][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:32:20,665][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:32:21,161][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:32:21,659][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:32:22,155][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:32:22,653][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:32:23,151][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:32:23,655][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:32:24,153][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:32:24,654][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:32:25,152][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:32:25,650][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:32:26,148][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:32:26,651][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:32:27,153][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:32:27,651][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:32:28,153][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:32:28,651][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:32:29,150][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:32:29,650][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:32:30,147][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10158 tokens. +[2026-03-26 06:32:30,989][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.58%, Block Peak % of device VRAM: 62.18%, ΔTime: 00:00:32 +[2026-03-26 06:32:31,737][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:32:31,739][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:32:31,741][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:32:32,475][__main__][INFO] - Iteration 600 took 49s (27.71% Gen, 70.82% Train). Generation: 13s, Training: 35s. Estimated remaining time: 32h 3m 55s. Estimated total time: 41h 30m 47s. Time estimates for 10 more iterations: 8m 18s, 100 more iterations: 1h 23m 1s, 500 more iterations: 6h 55m 7s. +[2026-03-26 06:32:32,478][__main__][INFO] - Starting iteration 600. +[2026-03-26 06:32:32,891][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 59 and human policies 1. +[2026-03-26 06:32:32,892][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:32:48,535][__main__][INFO] - Number of regex retries in iteration 600: 0 +[2026-03-26 06:32:48,536][__main__][INFO] - agents played in iteration 600 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:32:49,300][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:32:49,325][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:32:49,348][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:32:49,373][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:32:49,373][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:32:49,374][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:32:50,177][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:32:50,635][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:32:51,142][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:32:51,647][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:32:52,150][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:32:52,653][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:32:53,154][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:32:53,654][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:32:54,152][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:32:54,648][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:32:55,146][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:32:55,647][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:32:56,144][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:32:56,645][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:32:57,146][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:32:57,645][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:32:58,146][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:32:58,648][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:32:59,152][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:32:59,661][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:33:00,167][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:33:00,671][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:33:01,177][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:33:01,680][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:33:02,185][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:33:02,686][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:33:03,187][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:33:03,687][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:33:04,187][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:33:04,687][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:33:05,187][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:33:05,686][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:33:06,191][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:33:06,693][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:33:07,193][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:33:07,692][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:33:08,192][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:33:08,691][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:33:09,189][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:33:09,687][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:33:10,185][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:33:10,683][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:33:11,182][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:33:11,683][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:33:12,181][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:33:12,682][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:33:13,180][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:33:13,678][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:33:14,176][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:33:14,674][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:33:15,175][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:33:15,674][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:33:16,174][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:33:16,673][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:33:17,171][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:33:17,670][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:33:18,168][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:33:18,666][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:33:19,163][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:33:19,663][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:33:20,160][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:33:20,659][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:33:21,157][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:33:21,654][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:33:22,152][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10137 tokens. +[2026-03-26 06:33:22,994][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.09%, Current % of VRAM taken: 60.51%, Block Peak % of device VRAM: 62.23%, ΔTime: 00:00:32 +[2026-03-26 06:33:23,749][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:33:23,751][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:33:23,753][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:33:25,080][__main__][INFO] - Iteration 601 took 52s (29.98% Gen, 67.48% Train). Generation: 15s, Training: 35s. Estimated remaining time: 34h 1m 45s. Estimated total time: 43h 29m 29s. Time estimates for 10 more iterations: 8m 41s, 100 more iterations: 1h 26m 58s, 500 more iterations: 7h 14m 54s. +[2026-03-26 06:33:25,082][__main__][INFO] - Starting iteration 601. +[2026-03-26 06:33:25,487][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 60 and human policies 1. +[2026-03-26 06:33:25,488][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:33:30,847][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 06:33:32,028][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 06:33:41,934][__main__][INFO] - Number of regex retries in iteration 601: 2 +[2026-03-26 06:33:41,935][__main__][INFO] - agents played in iteration 601 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:33:42,715][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:33:42,739][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:33:42,763][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:33:42,786][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:33:42,786][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:33:42,787][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:33:43,575][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:33:44,029][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:33:44,536][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:33:45,036][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:33:45,534][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:33:46,033][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:33:46,541][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:33:47,039][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:33:47,538][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:33:48,038][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:33:48,537][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:33:49,045][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:33:49,542][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:33:50,041][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:33:50,539][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:33:51,042][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:33:51,544][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:33:52,049][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:33:52,552][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:33:53,056][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:33:53,556][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:33:54,059][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:33:54,561][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:33:55,065][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:33:55,566][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:33:56,066][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:33:56,568][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:33:57,067][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:33:57,566][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:33:58,070][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:33:58,570][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:33:59,080][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:33:59,580][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:34:00,079][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:34:00,578][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:34:01,077][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:34:01,576][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:34:02,081][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:34:02,579][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:34:03,078][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:34:03,578][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:34:04,076][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:34:04,575][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:34:05,074][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:34:05,575][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:34:06,074][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:34:06,574][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:34:07,073][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:34:07,571][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:34:08,073][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:34:08,574][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:34:09,072][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:34:09,572][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:34:10,075][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:34:10,577][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:34:11,078][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:34:11,577][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:34:12,078][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:34:12,575][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:34:13,074][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:34:13,571][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:34:14,069][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:34:14,570][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:34:15,070][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:34:15,567][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10184 tokens. +[2026-03-26 06:34:16,417][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.05%, Current % of VRAM taken: 60.48%, Block Peak % of device VRAM: 62.14%, ΔTime: 00:00:32 +[2026-03-26 06:34:17,175][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:34:17,177][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:34:17,179][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:34:18,042][__main__][INFO] - Iteration 602 took 52s (31.29% Gen, 67.06% Train). Generation: 16s, Training: 35s. Estimated remaining time: 34h 19m 8s. Estimated total time: 43h 47m 45s. Time estimates for 10 more iterations: 8m 45s, 100 more iterations: 1h 27m 35s, 500 more iterations: 7h 17m 57s. +[2026-03-26 06:34:18,044][__main__][INFO] - Starting iteration 602. +[2026-03-26 06:34:18,458][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 60 and human policies 1. +[2026-03-26 06:34:18,458][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:34:33,017][__main__][INFO] - Number of regex retries in iteration 602: 0 +[2026-03-26 06:34:33,018][__main__][INFO] - agents played in iteration 602 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:34:33,794][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:34:33,818][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:34:33,841][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:34:33,865][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:34:33,865][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:34:33,866][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:34:34,667][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:34:35,130][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:34:35,639][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:34:36,142][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:34:36,646][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:34:37,147][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:34:37,650][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:34:38,152][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:34:38,653][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:34:39,150][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:34:39,650][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:34:40,148][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:34:40,646][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:34:41,145][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:34:41,644][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:34:42,143][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:34:42,642][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:34:43,139][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:34:43,638][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:34:44,138][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:34:44,636][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:34:45,136][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:34:45,634][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:34:46,130][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:34:46,625][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:34:47,122][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:34:47,619][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:34:48,120][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:34:48,617][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:34:49,115][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:34:49,612][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:34:50,107][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:34:50,604][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:34:51,105][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:34:51,612][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:34:52,129][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:34:52,634][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:34:53,138][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:34:53,644][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:34:54,145][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:34:54,653][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:34:55,163][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:34:55,674][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:34:56,178][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:34:56,681][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:34:57,184][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:34:57,687][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:34:58,193][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:34:58,700][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:34:59,213][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:34:59,718][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:35:00,220][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:35:00,723][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:35:01,224][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:35:01,726][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:35:02,230][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:35:02,735][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:35:03,240][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:35:03,752][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:35:04,254][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:35:04,757][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:35:05,258][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:35:05,762][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:35:06,263][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:35:06,763][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10192 tokens. +[2026-03-26 06:35:07,606][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.42%, ΔTime: 00:00:32 +[2026-03-26 06:35:08,367][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:35:08,370][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:35:08,371][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:35:09,151][__main__][INFO] - Iteration 603 took 50s (28.72% Gen, 69.74% Train). Generation: 14s, Training: 35s. Estimated remaining time: 32h 45m 14s. Estimated total time: 42h 14m 42s. Time estimates for 10 more iterations: 8m 26s, 100 more iterations: 1h 24m 29s, 500 more iterations: 7h 2m 27s. +[2026-03-26 06:35:09,153][__main__][INFO] - Starting iteration 603. +[2026-03-26 06:35:09,567][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 60 and human policies 1. +[2026-03-26 06:35:09,567][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:35:13,387][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 06:35:23,104][__main__][INFO] - Number of regex retries in iteration 603: 1 +[2026-03-26 06:35:23,104][__main__][INFO] - agents played in iteration 603 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:35:23,866][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:35:23,891][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:35:23,916][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:35:23,940][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:35:23,941][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:35:23,941][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:35:24,742][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:35:25,202][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:35:25,710][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:35:26,220][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:35:26,722][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:35:27,223][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:35:27,725][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:35:28,227][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:35:28,727][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:35:29,230][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:35:29,737][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:35:30,240][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:35:30,741][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:35:31,247][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:35:31,753][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:35:32,257][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:35:32,766][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:35:33,271][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:35:33,774][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:35:34,275][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:35:34,778][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:35:35,279][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:35:35,779][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:35:36,281][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:35:36,781][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:35:37,285][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:35:37,787][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:35:38,289][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:35:38,791][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:35:39,294][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:35:39,797][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:35:40,299][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:35:40,800][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:35:41,300][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:35:41,800][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:35:42,299][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:35:42,801][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:35:43,300][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:35:43,799][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:35:44,302][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:35:44,800][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:35:45,301][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:35:45,801][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:35:46,303][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:35:46,803][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:35:47,305][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:35:47,803][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:35:48,305][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:35:48,805][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:35:49,301][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:35:49,799][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:35:50,299][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:35:50,798][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:35:51,297][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:35:51,797][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:35:52,296][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:35:52,794][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:35:53,294][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:35:53,794][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:35:54,301][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:35:54,801][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:35:55,304][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:35:55,810][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:35:56,318][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:35:56,821][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10265 tokens. +[2026-03-26 06:35:57,673][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.58%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:32 +[2026-03-26 06:35:58,441][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:35:58,443][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:35:58,445][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:35:59,181][__main__][INFO] - Iteration 604 took 49s (27.28% Gen, 71.23% Train). Generation: 13s, Training: 35s. Estimated remaining time: 31h 50m 26s. Estimated total time: 41h 20m 45s. Time estimates for 10 more iterations: 8m 16s, 100 more iterations: 1h 22m 41s, 500 more iterations: 6h 53m 27s. +[2026-03-26 06:35:59,184][__main__][INFO] - Starting iteration 604. +[2026-03-26 06:35:59,609][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 60 and human policies 1. +[2026-03-26 06:35:59,609][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:36:04,696][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 06:36:15,283][__main__][INFO] - Number of regex retries in iteration 604: 1 +[2026-03-26 06:36:15,284][__main__][INFO] - agents played in iteration 604 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:36:16,059][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:36:16,083][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:36:16,107][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:36:16,130][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:36:16,131][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:36:16,131][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:36:16,919][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:36:17,379][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:36:17,892][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:36:18,393][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:36:18,906][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:36:19,406][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:36:19,914][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:36:20,416][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:36:20,922][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:36:21,424][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:36:21,924][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:36:22,426][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:36:22,926][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:36:23,428][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:36:23,929][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:36:24,428][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:36:24,931][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:36:25,430][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:36:25,930][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:36:26,432][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:36:26,932][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:36:27,431][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:36:27,931][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:36:28,435][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:36:28,933][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:36:29,437][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:36:29,936][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:36:30,435][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:36:30,937][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:36:31,440][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:36:31,940][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:36:32,440][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:36:32,939][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:36:33,440][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:36:33,946][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:36:34,447][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:36:34,954][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:36:35,456][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:36:35,965][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:36:36,469][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:36:36,979][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:36:37,481][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:36:37,982][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:36:38,482][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:36:38,982][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:36:39,481][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:36:39,981][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:36:40,481][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:36:40,980][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:36:41,483][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:36:41,990][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:36:42,494][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:36:42,998][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:36:43,501][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:36:44,004][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:36:44,507][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:36:45,009][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:36:45,506][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:36:46,005][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:36:46,506][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:36:47,007][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:36:47,505][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:36:48,003][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:36:48,505][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:36:49,005][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10182 tokens. +[2026-03-26 06:36:49,856][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 60.57%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 +[2026-03-26 06:36:50,607][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:36:50,610][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:36:50,611][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:36:51,371][__main__][INFO] - Iteration 605 took 51s (30.28% Gen, 68.25% Train). Generation: 15s, Training: 35s. Estimated remaining time: 33h 36m 56s. Estimated total time: 43h 8m 7s. Time estimates for 10 more iterations: 8m 37s, 100 more iterations: 1h 26m 16s, 500 more iterations: 7h 11m 21s. +[2026-03-26 06:36:51,373][__main__][INFO] - Starting iteration 605. +[2026-03-26 06:36:51,785][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 60 and human policies 1. +[2026-03-26 06:36:51,786][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:37:05,706][__main__][INFO] - Number of regex retries in iteration 605: 0 +[2026-03-26 06:37:05,707][__main__][INFO] - agents played in iteration 605 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:37:06,479][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:37:06,502][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:37:06,526][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:37:06,549][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:37:06,550][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:37:06,551][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:37:07,351][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:37:07,808][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:37:08,311][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:37:08,808][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:37:09,305][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:37:09,804][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:37:10,301][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:37:10,798][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:37:11,294][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:37:11,789][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:37:12,284][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:37:12,782][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:37:13,278][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:37:13,772][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:37:14,268][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:37:14,763][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:37:15,259][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:37:15,759][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:37:16,256][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:37:16,756][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:37:17,252][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:37:17,750][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:37:18,247][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:37:18,746][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:37:19,245][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:37:19,744][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:37:20,242][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:37:20,740][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:37:21,239][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:37:21,736][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:37:22,234][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:37:22,743][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:37:23,243][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:37:23,751][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:37:24,255][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:37:24,758][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:37:25,266][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:37:25,771][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:37:26,270][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:37:26,771][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:37:27,272][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:37:27,771][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:37:28,269][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:37:28,766][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:37:29,264][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:37:29,761][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:37:30,260][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:37:30,759][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:37:31,256][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:37:31,756][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:37:32,255][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:37:32,754][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:37:33,252][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:37:33,750][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:37:34,248][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:37:34,747][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:37:35,246][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:37:35,746][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:37:36,247][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:37:36,747][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:37:37,245][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:37:37,743][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:37:38,243][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:37:38,744][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:37:39,244][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9987 tokens. +[2026-03-26 06:37:40,079][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.57%, Block Peak % of device VRAM: 62.13%, ΔTime: 00:00:32 +[2026-03-26 06:37:40,824][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:37:40,826][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:37:40,828][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:37:41,686][__main__][INFO] - Iteration 606 took 49s (27.90% Gen, 70.38% Train). Generation: 13s, Training: 35s. Estimated remaining time: 32h 3m 3s. Estimated total time: 41h 35m 4s. Time estimates for 10 more iterations: 8m 19s, 100 more iterations: 1h 23m 10s, 500 more iterations: 6h 55m 50s. +[2026-03-26 06:37:41,688][__main__][INFO] - Starting iteration 606. +[2026-03-26 06:37:42,105][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 60 and human policies 1. +[2026-03-26 06:37:42,106][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:37:46,129][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 06:37:47,208][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 06:37:57,563][__main__][INFO] - Number of regex retries in iteration 606: 2 +[2026-03-26 06:37:57,564][__main__][INFO] - agents played in iteration 606 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:37:58,331][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:37:58,355][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:37:58,379][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:37:58,403][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:37:58,404][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:37:58,405][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:37:59,201][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:37:59,666][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:38:00,187][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:38:00,693][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:38:01,201][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:38:01,707][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:38:02,210][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:38:02,714][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:38:03,234][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:38:03,736][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:38:04,240][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:38:04,747][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:38:05,247][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:38:05,747][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:38:06,253][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:38:06,767][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:38:07,268][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:38:07,776][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:38:08,282][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:38:08,783][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:38:09,286][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:38:09,788][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:38:10,296][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:38:10,811][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:38:11,317][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:38:11,818][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:38:12,321][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:38:12,826][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:38:13,327][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:38:13,828][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:38:14,328][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:38:14,835][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:38:15,343][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:38:15,852][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:38:16,354][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:38:16,858][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:38:17,361][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:38:17,861][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:38:18,361][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:38:18,864][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:38:19,367][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:38:19,867][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:38:20,371][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:38:20,875][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:38:21,388][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:38:21,891][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:38:22,392][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:38:22,899][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:38:23,401][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:38:23,901][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:38:24,401][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:38:24,904][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:38:25,406][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:38:25,905][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:38:26,406][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:38:26,909][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:38:27,414][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:38:27,919][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:38:28,431][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:38:28,931][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:38:29,433][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:38:29,937][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:38:30,438][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:38:30,940][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:38:31,445][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10210 tokens. +[2026-03-26 06:38:32,284][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 60.62%, Block Peak % of device VRAM: 62.23%, ΔTime: 00:00:33 +[2026-03-26 06:38:33,041][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:38:33,044][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:38:33,046][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:38:33,772][__main__][INFO] - Iteration 607 took 51s (29.92% Gen, 68.67% Train). Generation: 15s, Training: 35s. Estimated remaining time: 33h 30m 28s. Estimated total time: 43h 3m 21s. Time estimates for 10 more iterations: 8m 36s, 100 more iterations: 1h 26m 6s, 500 more iterations: 7h 10m 33s. +[2026-03-26 06:38:33,774][__main__][INFO] - Starting iteration 607. +[2026-03-26 06:38:34,181][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 60 and human policies 1. +[2026-03-26 06:38:34,182][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:38:39,273][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 06:38:50,009][__main__][INFO] - Number of regex retries in iteration 607: 1 +[2026-03-26 06:38:50,010][__main__][INFO] - agents played in iteration 607 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:38:50,808][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:38:50,832][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:38:50,855][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:38:50,879][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:38:50,879][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:38:50,880][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:38:51,673][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:38:52,134][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:38:52,640][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:38:53,145][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:38:53,650][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:38:54,149][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:38:54,650][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:38:55,149][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:38:55,648][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:38:56,147][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:38:56,645][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:38:57,149][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:38:57,647][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:38:58,151][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:38:58,660][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:38:59,160][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:38:59,660][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:39:00,160][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:39:00,661][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:39:01,161][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:39:01,661][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:39:02,166][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:39:02,666][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:39:03,164][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:39:03,664][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:39:04,164][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:39:04,664][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:39:05,164][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:39:05,664][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:39:06,165][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:39:06,667][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:39:07,167][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:39:07,669][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:39:08,174][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:39:08,680][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:39:09,185][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:39:09,686][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:39:10,193][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:39:10,697][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:39:11,198][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:39:11,700][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:39:12,201][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:39:12,700][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:39:13,201][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:39:13,699][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:39:14,201][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:39:14,701][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:39:15,201][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:39:15,700][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:39:16,201][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:39:16,701][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:39:17,200][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:39:17,699][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:39:18,199][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:39:18,698][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:39:19,197][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:39:19,699][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:39:20,202][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:39:20,698][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:39:21,194][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:39:21,691][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:39:22,187][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:39:22,685][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:39:23,183][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:39:23,680][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10097 tokens. +[2026-03-26 06:39:24,514][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.02%, Current % of VRAM taken: 60.45%, Block Peak % of device VRAM: 62.10%, ΔTime: 00:00:32 +[2026-03-26 06:39:25,262][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:39:25,264][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:39:25,266][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:39:26,166][__main__][INFO] - Iteration 608 took 51s (30.45% Gen, 67.82% Train). Generation: 15s, Training: 35s. Estimated remaining time: 33h 45m 29s. Estimated total time: 43h 19m 15s. Time estimates for 10 more iterations: 8m 39s, 100 more iterations: 1h 26m 38s, 500 more iterations: 7h 13m 12s. +[2026-03-26 06:39:26,168][__main__][INFO] - Starting iteration 608. +[2026-03-26 06:39:26,575][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 60 and human policies 1. +[2026-03-26 06:39:26,576][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:39:42,416][__main__][INFO] - Number of regex retries in iteration 608: 0 +[2026-03-26 06:39:42,417][__main__][INFO] - agents played in iteration 608 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:39:43,186][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:39:43,210][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:39:43,233][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:39:43,257][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:39:43,257][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:39:43,258][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:39:44,051][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:39:44,514][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:39:45,019][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:39:45,522][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:39:46,023][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:39:46,522][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:39:47,023][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:39:47,527][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:39:48,028][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:39:48,524][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:39:49,022][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:39:49,518][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:39:50,015][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:39:50,512][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:39:51,011][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:39:51,506][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:39:52,004][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:39:52,502][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:39:53,009][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:39:53,510][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:39:54,015][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:39:54,518][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:39:55,018][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:39:55,525][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:39:56,026][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:39:56,536][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:39:57,036][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:39:57,538][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:39:58,040][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:39:58,542][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:39:59,043][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:39:59,546][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:40:00,046][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:40:00,548][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:40:01,047][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:40:01,548][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:40:02,047][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:40:02,546][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:40:03,044][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:40:03,544][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:40:04,044][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:40:04,541][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:40:05,039][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:40:05,536][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:40:06,032][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:40:06,528][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:40:07,024][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:40:07,521][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:40:08,020][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:40:08,518][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:40:09,016][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:40:09,513][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:40:10,009][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:40:10,509][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:40:11,008][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:40:11,506][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:40:12,005][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:40:12,504][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:40:13,004][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:40:13,505][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:40:14,006][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:40:14,506][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:40:15,013][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:40:15,514][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:40:16,016][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10196 tokens. +[2026-03-26 06:40:16,853][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 60.63%, Block Peak % of device VRAM: 62.27%, ΔTime: 00:00:32 +[2026-03-26 06:40:17,611][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:40:17,613][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:40:17,615][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:40:18,480][__main__][INFO] - Iteration 609 took 51s (30.52% Gen, 67.81% Train). Generation: 15s, Training: 35s. Estimated remaining time: 33h 40m 38s. Estimated total time: 43h 15m 16s. Time estimates for 10 more iterations: 8m 39s, 100 more iterations: 1h 26m 30s, 500 more iterations: 7h 12m 32s. +[2026-03-26 06:40:18,482][__main__][INFO] - Starting iteration 609. +[2026-03-26 06:40:18,889][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 60 and human policies 1. +[2026-03-26 06:40:18,890][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:40:30,014][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 06:40:34,550][__main__][INFO] - Number of regex retries in iteration 609: 1 +[2026-03-26 06:40:34,551][__main__][INFO] - agents played in iteration 609 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:40:35,330][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:40:35,355][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:40:35,381][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:40:35,406][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:40:35,406][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:40:35,407][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:40:36,213][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:40:36,679][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:40:37,194][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:40:37,703][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:40:38,210][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:40:38,713][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:40:39,221][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:40:39,723][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:40:40,226][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:40:40,726][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:40:41,229][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:40:41,734][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:40:42,240][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:40:42,755][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:40:43,257][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:40:43,761][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:40:44,263][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:40:44,768][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:40:45,270][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:40:45,774][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:40:46,279][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:40:46,783][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:40:47,290][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:40:47,794][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:40:48,299][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:40:48,807][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:40:49,308][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:40:49,810][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:40:50,311][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:40:50,841][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:40:51,342][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:40:51,848][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:40:52,349][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:40:52,850][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:40:53,349][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:40:53,854][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:40:54,355][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:40:54,857][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:40:55,362][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:40:55,862][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:40:56,362][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:40:56,871][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:40:57,373][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:40:57,879][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:40:58,386][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:40:58,889][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:40:59,392][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:40:59,891][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:41:00,396][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:41:00,898][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:41:01,400][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:41:01,902][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:41:02,404][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:41:02,906][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:41:03,406][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:41:03,905][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:41:04,406][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:41:04,908][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:41:05,410][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:41:05,910][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:41:06,410][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:41:06,909][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:41:07,412][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:41:07,911][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:41:08,409][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10138 tokens. +[2026-03-26 06:41:09,248][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.28%, ΔTime: 00:00:33 +[2026-03-26 06:41:10,040][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:41:10,042][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:41:10,047][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:41:10,876][__main__][INFO] - Iteration 610 took 51s (30.12% Gen, 68.28% Train). Generation: 15s, Training: 35s. Estimated remaining time: 33h 43m 51s. Estimated total time: 43h 19m 21s. Time estimates for 10 more iterations: 8m 39s, 100 more iterations: 1h 26m 38s, 500 more iterations: 7h 13m 13s. +[2026-03-26 06:41:10,878][__main__][INFO] - Starting iteration 610. +[2026-03-26 06:41:11,294][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 60 and human policies 1. +[2026-03-26 06:41:11,294][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:41:26,016][__main__][INFO] - Number of regex retries in iteration 610: 0 +[2026-03-26 06:41:26,018][__main__][INFO] - agents played in iteration 610 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:41:26,815][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:41:26,839][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:41:26,863][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:41:26,886][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:41:26,887][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:41:26,887][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:41:27,690][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:41:28,146][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:41:28,655][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:41:29,160][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:41:29,661][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:41:30,169][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:41:30,681][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:41:31,183][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:41:31,685][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:41:32,190][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:41:32,690][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:41:33,191][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:41:33,699][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:41:34,223][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:41:34,728][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:41:35,228][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:41:35,728][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:41:36,227][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:41:36,725][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:41:37,224][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:41:37,723][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:41:38,221][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:41:38,720][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:41:39,217][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:41:39,722][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:41:40,225][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:41:40,730][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:41:41,237][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:41:41,739][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:41:42,250][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:41:42,751][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:41:43,253][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:41:43,756][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:41:44,255][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:41:44,754][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:41:45,253][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:41:45,754][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:41:46,254][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:41:46,752][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:41:47,254][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:41:47,752][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:41:48,251][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:41:48,748][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:41:49,246][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:41:49,746][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:41:50,246][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:41:50,746][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:41:51,245][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:41:51,743][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:41:52,242][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:41:52,743][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:41:53,243][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:41:53,744][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:41:54,243][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:41:54,742][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:41:55,242][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:41:55,742][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:41:56,241][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:41:56,739][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:41:57,242][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:41:57,739][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:41:58,239][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:41:58,736][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:41:59,237][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:41:59,734][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10036 tokens. +[2026-03-26 06:42:00,573][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.09%, Current % of VRAM taken: 60.52%, Block Peak % of device VRAM: 62.08%, ΔTime: 00:00:32 +[2026-03-26 06:42:01,320][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:42:01,322][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:42:01,324][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:42:02,624][__main__][INFO] - Iteration 611 took 51s (28.68% Gen, 68.78% Train). Generation: 14s, Training: 35s. Estimated remaining time: 33h 10m 17s. Estimated total time: 42h 46m 39s. Time estimates for 10 more iterations: 8m 33s, 100 more iterations: 1h 25m 33s, 500 more iterations: 7h 7m 46s. +[2026-03-26 06:42:02,627][__main__][INFO] - Starting iteration 611. +[2026-03-26 06:42:03,048][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 61 and human policies 1. +[2026-03-26 06:42:03,049][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:42:08,547][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 06:42:18,460][__main__][INFO] - Number of regex retries in iteration 611: 1 +[2026-03-26 06:42:18,461][__main__][INFO] - agents played in iteration 611 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:42:19,250][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:42:19,279][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:42:19,305][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:42:19,332][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:42:19,333][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:42:19,333][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:42:20,080][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:42:20,547][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:42:21,058][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:42:21,565][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:42:22,076][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:42:22,582][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:42:23,116][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:42:23,627][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:42:24,136][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:42:24,641][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:42:25,146][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:42:25,650][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:42:26,153][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:42:26,659][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:42:27,164][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:42:27,668][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:42:28,174][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:42:28,675][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:42:29,180][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:42:29,685][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:42:30,185][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:42:30,689][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:42:31,188][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:42:31,688][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:42:32,190][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:42:32,690][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:42:33,201][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:42:33,702][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:42:34,203][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:42:34,703][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:42:35,206][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:42:35,705][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:42:36,204][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:42:36,706][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:42:37,207][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:42:37,708][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:42:38,206][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:42:38,707][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:42:39,205][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:42:39,705][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:42:40,204][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:42:40,702][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:42:41,202][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:42:41,701][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:42:42,200][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:42:42,701][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:42:43,203][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:42:43,700][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:42:44,199][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:42:44,699][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:42:45,198][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:42:45,696][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:42:46,217][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:42:46,714][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:42:47,212][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:42:47,710][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:42:48,208][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:42:48,706][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:42:49,203][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:42:49,701][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:42:50,201][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:42:50,698][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:42:51,194][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:42:51,690][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:42:52,187][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10221 tokens. +[2026-03-26 06:42:52,968][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 60.54%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:32 +[2026-03-26 06:42:53,780][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:42:53,783][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:42:53,785][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:42:54,589][__main__][INFO] - Iteration 612 took 51s (29.90% Gen, 68.53% Train). Generation: 15s, Training: 35s. Estimated remaining time: 33h 19m 50s. Estimated total time: 42h 57m 5s. Time estimates for 10 more iterations: 8m 35s, 100 more iterations: 1h 25m 54s, 500 more iterations: 7h 9m 30s. +[2026-03-26 06:42:54,592][__main__][INFO] - Starting iteration 612. +[2026-03-26 06:42:55,001][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 61 and human policies 1. +[2026-03-26 06:42:55,002][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:43:09,017][__main__][INFO] - Number of regex retries in iteration 612: 0 +[2026-03-26 06:43:09,018][__main__][INFO] - agents played in iteration 612 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:43:09,798][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:43:09,822][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:43:09,846][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:43:09,871][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:43:09,871][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:43:09,872][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:43:10,632][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:43:11,090][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:43:11,595][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:43:12,105][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:43:12,610][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:43:13,116][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:43:13,617][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:43:14,119][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:43:14,621][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:43:15,120][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:43:15,619][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:43:16,126][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:43:16,627][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:43:17,130][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:43:17,635][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:43:18,148][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:43:18,648][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:43:19,151][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:43:19,657][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:43:20,158][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:43:20,662][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:43:21,162][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:43:21,662][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:43:22,164][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:43:22,664][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:43:23,172][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:43:23,673][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:43:24,176][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:43:24,679][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:43:25,182][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:43:25,685][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:43:26,187][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:43:26,685][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:43:27,187][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:43:27,687][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:43:28,185][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:43:28,682][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:43:29,179][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:43:29,677][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:43:30,173][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:43:30,672][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:43:31,173][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:43:31,673][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:43:32,174][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:43:32,674][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:43:33,173][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:43:33,673][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:43:34,171][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:43:34,669][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:43:35,170][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:43:35,671][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:43:36,172][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:43:36,672][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:43:37,171][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:43:37,669][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:43:38,167][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:43:38,665][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:43:39,166][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:43:39,663][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:43:40,165][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:43:40,662][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:43:41,161][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:43:41,662][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:43:42,163][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:43:42,663][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10197 tokens. +[2026-03-26 06:43:43,446][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 60.60%, Block Peak % of device VRAM: 62.18%, ΔTime: 00:00:32 +[2026-03-26 06:43:44,202][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:43:44,204][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:43:44,205][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:43:44,937][__main__][INFO] - Iteration 613 took 49s (28.07% Gen, 70.46% Train). Generation: 14s, Training: 35s. Estimated remaining time: 31h 58m 43s. Estimated total time: 41h 36m 48s. Time estimates for 10 more iterations: 8m 19s, 100 more iterations: 1h 23m 13s, 500 more iterations: 6h 56m 8s. +[2026-03-26 06:43:44,939][__main__][INFO] - Starting iteration 613. +[2026-03-26 06:43:45,349][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 61 and human policies 1. +[2026-03-26 06:43:45,350][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:43:50,988][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 06:44:00,545][__main__][INFO] - Number of regex retries in iteration 613: 1 +[2026-03-26 06:44:00,546][__main__][INFO] - agents played in iteration 613 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:44:01,323][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:44:01,349][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:44:01,374][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:44:01,399][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:44:01,400][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:44:01,401][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:44:02,167][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:44:02,624][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:44:03,127][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:44:03,625][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:44:04,123][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:44:04,622][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:44:05,120][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:44:05,620][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:44:06,118][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:44:06,614][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:44:07,112][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:44:07,615][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:44:08,116][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:44:08,615][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:44:09,121][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:44:09,627][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:44:10,136][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:44:10,636][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:44:11,143][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:44:11,644][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:44:12,146][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:44:12,646][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:44:13,147][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:44:13,647][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:44:14,154][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:44:14,655][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:44:15,156][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:44:15,657][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:44:16,159][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:44:16,659][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:44:17,163][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:44:17,669][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:44:18,193][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:44:18,695][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:44:19,202][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:44:19,703][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:44:20,206][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:44:20,705][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:44:21,208][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:44:21,706][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:44:22,207][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:44:22,708][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:44:23,207][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:44:23,708][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:44:24,206][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:44:24,705][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:44:25,205][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:44:25,703][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:44:26,202][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:44:26,701][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:44:27,200][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:44:27,698][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:44:28,197][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:44:28,696][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:44:29,194][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:44:29,692][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:44:30,190][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:44:30,686][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:44:31,183][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:44:31,681][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:44:32,180][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:44:32,678][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:44:33,174][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:44:33,670][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:44:34,170][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10045 tokens. +[2026-03-26 06:44:34,952][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 60.60%, Block Peak % of device VRAM: 62.09%, ΔTime: 00:00:32 +[2026-03-26 06:44:35,696][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:44:35,698][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:44:35,700][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:44:36,421][__main__][INFO] - Iteration 614 took 51s (29.75% Gen, 68.83% Train). Generation: 15s, Training: 35s. Estimated remaining time: 32h 54m 40s. Estimated total time: 42h 33m 36s. Time estimates for 10 more iterations: 8m 30s, 100 more iterations: 1h 25m 7s, 500 more iterations: 7h 5m 36s. +[2026-03-26 06:44:36,423][__main__][INFO] - Starting iteration 614. +[2026-03-26 06:44:36,820][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 61 and human policies 1. +[2026-03-26 06:44:36,821][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:44:47,437][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 06:44:50,642][__main__][INFO] - Number of regex retries in iteration 614: 1 +[2026-03-26 06:44:50,642][__main__][INFO] - agents played in iteration 614 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:44:51,413][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:44:51,437][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:44:51,461][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:44:51,485][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:44:51,486][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:44:51,487][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:44:52,229][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:44:52,685][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:44:53,191][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:44:53,692][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:44:54,197][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:44:54,695][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:44:55,192][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:44:55,689][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:44:56,187][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:44:56,688][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:44:57,185][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:44:57,683][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:44:58,182][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:44:58,680][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:44:59,179][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:44:59,678][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:45:00,179][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:45:00,683][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:45:01,184][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:45:01,687][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:45:02,190][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:45:02,701][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:45:03,201][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:45:03,702][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:45:04,202][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:45:04,703][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:45:05,204][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:45:05,704][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:45:06,205][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:45:06,709][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:45:07,214][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:45:07,711][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:45:08,210][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:45:08,710][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:45:09,209][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:45:09,708][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:45:10,206][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:45:10,704][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:45:11,203][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:45:11,724][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:45:12,225][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:45:12,724][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:45:13,227][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:45:13,724][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:45:14,223][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:45:14,725][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:45:15,222][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:45:15,720][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:45:16,218][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:45:16,715][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:45:17,210][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:45:17,706][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:45:18,204][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:45:18,701][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:45:19,200][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:45:19,696][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:45:20,195][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:45:20,690][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:45:21,190][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:45:21,686][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:45:22,183][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:45:22,679][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:45:23,177][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:45:23,673][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:45:24,169][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10055 tokens. +[2026-03-26 06:45:24,953][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 60.51%, Block Peak % of device VRAM: 62.09%, ΔTime: 00:00:32 +[2026-03-26 06:45:25,701][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:45:25,703][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:45:25,705][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:45:26,438][__main__][INFO] - Iteration 615 took 49s (27.86% Gen, 70.66% Train). Generation: 13s, Training: 35s. Estimated remaining time: 31h 41m 9s. Estimated total time: 41h 20m 55s. Time estimates for 10 more iterations: 8m 16s, 100 more iterations: 1h 22m 41s, 500 more iterations: 6h 53m 29s. +[2026-03-26 06:45:26,440][__main__][INFO] - Starting iteration 615. +[2026-03-26 06:45:26,851][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 61 and human policies 1. +[2026-03-26 06:45:26,851][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:45:42,218][__main__][INFO] - Number of regex retries in iteration 615: 0 +[2026-03-26 06:45:42,219][__main__][INFO] - agents played in iteration 615 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:45:43,005][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:45:43,029][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:45:43,053][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:45:43,076][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:45:43,077][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:45:43,077][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:45:43,824][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:45:44,288][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:45:44,790][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:45:45,297][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:45:45,796][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:45:46,295][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:45:46,799][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:45:47,302][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:45:47,815][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:45:48,315][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:45:48,815][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:45:49,319][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:45:49,824][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:45:50,331][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:45:50,832][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:45:51,335][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:45:51,839][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:45:52,339][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:45:52,840][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:45:53,341][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:45:53,843][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:45:54,343][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:45:54,844][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:45:55,347][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:45:55,846][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:45:56,348][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:45:56,850][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:45:57,351][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:45:57,853][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:45:58,354][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:45:58,854][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:45:59,355][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:45:59,864][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:46:00,362][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:46:00,861][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:46:01,360][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:46:01,860][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:46:02,360][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:46:02,859][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:46:03,361][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:46:03,859][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:46:04,360][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:46:04,859][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:46:05,361][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:46:05,868][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:46:06,366][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:46:06,866][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:46:07,371][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:46:07,870][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:46:08,376][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:46:08,879][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:46:09,383][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:46:09,894][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:46:10,396][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:46:10,900][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:46:11,403][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:46:11,907][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:46:12,406][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:46:12,906][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:46:13,410][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:46:13,912][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:46:14,411][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:46:14,911][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:46:15,412][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:46:15,911][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10161 tokens. +[2026-03-26 06:46:16,726][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 60.53%, Block Peak % of device VRAM: 62.24%, ΔTime: 00:00:32 +[2026-03-26 06:46:17,477][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:46:17,479][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:46:17,481][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:46:18,209][__main__][INFO] - Iteration 616 took 51s (29.92% Gen, 68.66% Train). Generation: 15s, Training: 35s. Estimated remaining time: 33h 7m 18s. Estimated total time: 42h 47m 56s. Time estimates for 10 more iterations: 8m 33s, 100 more iterations: 1h 25m 35s, 500 more iterations: 7h 7m 59s. +[2026-03-26 06:46:18,211][__main__][INFO] - Starting iteration 616. +[2026-03-26 06:46:18,629][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 61 and human policies 1. +[2026-03-26 06:46:18,630][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:46:35,860][__main__][INFO] - Number of regex retries in iteration 616: 0 +[2026-03-26 06:46:35,860][__main__][INFO] - agents played in iteration 616 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:46:36,628][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:46:36,652][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:46:36,675][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:46:36,699][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:46:36,699][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:46:36,700][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:46:37,448][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:46:37,906][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:46:38,409][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:46:38,912][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:46:39,414][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:46:39,915][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:46:40,417][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:46:40,916][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:46:41,422][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:46:41,922][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:46:42,423][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:46:42,929][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:46:43,433][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:46:43,934][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:46:44,436][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:46:44,941][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:46:45,445][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:46:45,948][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:46:46,451][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:46:46,953][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:46:47,456][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:46:47,960][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:46:48,473][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:46:48,974][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:46:49,477][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:46:49,978][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:46:50,479][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:46:50,979][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:46:51,480][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:46:51,980][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:46:52,480][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:46:52,978][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:46:53,481][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:46:53,985][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:46:54,486][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:46:54,985][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:46:55,487][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:46:55,985][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:46:56,488][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:46:56,986][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:46:57,486][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:46:57,984][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:46:58,485][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:46:58,983][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:46:59,484][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:46:59,985][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:47:00,483][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:47:00,982][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:47:01,480][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:47:01,984][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:47:02,482][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:47:02,984][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:47:03,482][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:47:03,983][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:47:04,481][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:47:04,981][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:47:05,483][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:47:05,980][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:47:06,480][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:47:06,978][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:47:07,475][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:47:07,976][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:47:08,474][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:47:08,970][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:47:09,468][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10180 tokens. +[2026-03-26 06:47:10,258][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 60.54%, Block Peak % of device VRAM: 62.33%, ΔTime: 00:00:32 +[2026-03-26 06:47:11,014][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:47:11,016][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:47:11,018][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:47:11,753][__main__][INFO] - Iteration 617 took 53s (32.43% Gen, 66.18% Train). Generation: 17s, Training: 35s. Estimated remaining time: 34h 34m 43s. Estimated total time: 44h 16m 14s. Time estimates for 10 more iterations: 8m 51s, 100 more iterations: 1h 28m 32s, 500 more iterations: 7h 22m 42s. +[2026-03-26 06:47:11,756][__main__][INFO] - Starting iteration 617. +[2026-03-26 06:47:12,153][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 61 and human policies 1. +[2026-03-26 06:47:12,154][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:47:23,572][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 06:47:28,227][__main__][INFO] - Number of regex retries in iteration 617: 1 +[2026-03-26 06:47:28,228][__main__][INFO] - agents played in iteration 617 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:47:28,993][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:47:29,017][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:47:29,040][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:47:29,064][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.57%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:47:29,064][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:47:29,065][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:47:29,808][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:47:30,274][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:47:30,777][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:47:31,276][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:47:31,780][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:47:32,280][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:47:32,779][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:47:33,277][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:47:33,775][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:47:34,273][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:47:34,770][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:47:35,268][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:47:35,770][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:47:36,268][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:47:36,766][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:47:37,265][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:47:37,764][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:47:38,261][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:47:38,758][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:47:39,254][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:47:39,750][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:47:40,247][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:47:40,747][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:47:41,244][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:47:41,741][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:47:42,241][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:47:42,741][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:47:43,238][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:47:43,739][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:47:44,237][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:47:44,734][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:47:45,232][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:47:45,729][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:47:46,227][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:47:46,724][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:47:47,225][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:47:47,723][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:47:48,223][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:47:48,725][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:47:49,221][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:47:49,717][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:47:50,215][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:47:50,713][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:47:51,213][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:47:51,709][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:47:52,206][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:47:52,704][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:47:53,201][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:47:53,699][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:47:54,197][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:47:54,694][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:47:55,191][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:47:55,688][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:47:56,190][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:47:56,689][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:47:57,188][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:47:57,687][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:47:58,192][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:47:58,695][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:47:59,194][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:47:59,693][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:48:00,199][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:48:00,700][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:48:01,201][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:48:01,705][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10082 tokens. +[2026-03-26 06:48:02,508][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 60.67%, Block Peak % of device VRAM: 62.18%, ΔTime: 00:00:32 +[2026-03-26 06:48:03,376][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:48:03,379][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:48:03,380][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:48:04,127][__main__][INFO] - Iteration 618 took 51s (30.93% Gen, 67.63% Train). Generation: 16s, Training: 35s. Estimated remaining time: 33h 36m 19s. Estimated total time: 43h 18m 43s. Time estimates for 10 more iterations: 8m 39s, 100 more iterations: 1h 26m 37s, 500 more iterations: 7h 13m 7s. +[2026-03-26 06:48:04,129][__main__][INFO] - Starting iteration 618. +[2026-03-26 06:48:04,537][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 61 and human policies 1. +[2026-03-26 06:48:04,538][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:48:22,003][__main__][INFO] - Number of regex retries in iteration 618: 0 +[2026-03-26 06:48:22,004][__main__][INFO] - agents played in iteration 618 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:48:22,788][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:48:22,814][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:48:22,839][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:48:22,863][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:48:22,864][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:48:22,864][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:48:23,625][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:48:24,077][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:48:24,587][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:48:25,085][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:48:25,584][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:48:26,082][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:48:26,581][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:48:27,080][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:48:27,578][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:48:28,079][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:48:28,578][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:48:29,074][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:48:29,569][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:48:30,069][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:48:30,564][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:48:31,060][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:48:31,557][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:48:32,061][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:48:32,562][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:48:33,062][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:48:33,563][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:48:34,070][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:48:34,574][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:48:35,074][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:48:35,574][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:48:36,072][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:48:36,571][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:48:37,069][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:48:37,569][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:48:38,069][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:48:38,566][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:48:39,065][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:48:39,564][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:48:40,062][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:48:40,559][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:48:41,057][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:48:41,554][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:48:42,050][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:48:42,548][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:48:43,055][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:48:43,556][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:48:44,057][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:48:44,559][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:48:45,058][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:48:45,561][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:48:46,062][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:48:46,562][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:48:47,062][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:48:47,562][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:48:48,067][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:48:48,570][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:48:49,072][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:48:49,574][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:48:50,088][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:48:50,591][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:48:51,095][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:48:51,597][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:48:52,095][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:48:52,595][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:48:53,096][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:48:53,600][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:48:54,098][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:48:54,595][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:48:55,094][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:48:55,592][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10200 tokens. +[2026-03-26 06:48:56,374][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.05%, Current % of VRAM taken: 60.48%, Block Peak % of device VRAM: 62.53%, ΔTime: 00:00:32 +[2026-03-26 06:48:57,122][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:48:57,124][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:48:57,126][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:48:57,856][__main__][INFO] - Iteration 619 took 53s (32.76% Gen, 65.87% Train). Generation: 17s, Training: 35s. Estimated remaining time: 34h 42m 41s. Estimated total time: 44h 25m 59s. Time estimates for 10 more iterations: 8m 53s, 100 more iterations: 1h 28m 51s, 500 more iterations: 7h 24m 19s. +[2026-03-26 06:48:57,859][__main__][INFO] - Starting iteration 619. +[2026-03-26 06:48:58,267][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 61 and human policies 1. +[2026-03-26 06:48:58,268][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:49:14,123][__main__][INFO] - Number of regex retries in iteration 619: 0 +[2026-03-26 06:49:14,124][__main__][INFO] - agents played in iteration 619 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:49:14,908][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:49:14,931][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:49:14,955][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:49:14,978][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:49:14,979][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:49:14,980][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:49:15,719][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:49:16,178][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:49:16,688][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:49:17,196][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:49:17,698][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:49:18,199][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:49:18,699][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:49:19,198][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:49:19,704][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:49:20,209][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:49:20,710][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:49:21,210][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:49:21,710][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:49:22,214][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:49:22,713][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:49:23,214][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:49:23,724][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:49:24,226][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:49:24,727][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:49:25,260][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:49:25,761][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:49:26,261][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:49:26,764][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:49:27,265][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:49:27,768][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:49:28,269][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:49:28,767][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:49:29,265][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:49:29,763][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:49:30,259][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:49:30,757][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:49:31,255][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:49:31,757][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:49:32,256][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:49:32,754][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:49:33,252][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:49:33,751][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:49:34,249][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:49:34,746][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:49:35,246][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:49:35,745][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:49:36,249][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:49:36,752][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:49:37,254][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:49:37,752][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:49:38,254][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:49:38,755][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:49:39,255][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:49:39,756][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:49:40,254][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:49:40,752][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:49:41,250][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:49:41,748][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:49:42,247][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:49:42,748][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:49:43,246][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:49:43,744][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:49:44,242][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:49:44,740][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:49:45,237][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:49:45,734][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:49:46,230][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:49:46,727][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:49:47,226][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:49:47,724][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10123 tokens. +[2026-03-26 06:49:48,504][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:32 +[2026-03-26 06:49:49,257][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:49:49,259][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:49:49,261][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:49:50,107][__main__][INFO] - Iteration 620 took 51s (30.59% Gen, 67.78% Train). Generation: 15s, Training: 35s. Estimated remaining time: 33h 27m 53s. Estimated total time: 43h 12m 3s. Time estimates for 10 more iterations: 8m 38s, 100 more iterations: 1h 26m 24s, 500 more iterations: 7h 12m 0s. +[2026-03-26 06:49:50,110][__main__][INFO] - Starting iteration 620. +[2026-03-26 06:49:50,524][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 61 and human policies 1. +[2026-03-26 06:49:50,524][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:50:04,332][__main__][INFO] - Number of regex retries in iteration 620: 0 +[2026-03-26 06:50:04,333][__main__][INFO] - agents played in iteration 620 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:50:05,118][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:50:05,144][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:50:05,169][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:50:05,194][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:50:05,195][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:50:05,195][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:50:05,967][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:50:06,425][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:50:06,929][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:50:07,428][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:50:07,929][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:50:08,433][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:50:08,938][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:50:09,445][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:50:09,957][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:50:10,456][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:50:10,956][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:50:11,455][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:50:11,954][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:50:12,452][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:50:12,951][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:50:13,452][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:50:13,950][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:50:14,456][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:50:14,954][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:50:15,459][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:50:15,960][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:50:16,463][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:50:16,967][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:50:17,473][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:50:17,974][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:50:18,474][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:50:18,974][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:50:19,475][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:50:19,974][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:50:20,477][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:50:20,981][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:50:21,480][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:50:21,978][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:50:22,479][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:50:22,982][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:50:23,482][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:50:23,984][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:50:24,484][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:50:25,009][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:50:25,507][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:50:26,009][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:50:26,508][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:50:27,012][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:50:27,512][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:50:28,011][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:50:28,510][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:50:29,009][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:50:29,508][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:50:30,009][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:50:30,506][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:50:31,005][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:50:31,504][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:50:32,006][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:50:32,505][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:50:33,004][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:50:33,502][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:50:34,001][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:50:34,499][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:50:34,997][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:50:35,495][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:50:35,993][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:50:36,491][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:50:36,989][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:50:37,488][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:50:37,987][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10031 tokens. +[2026-03-26 06:50:38,778][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 60.53%, Block Peak % of device VRAM: 62.18%, ΔTime: 00:00:32 +[2026-03-26 06:50:39,526][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:50:39,529][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:50:39,530][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:50:41,060][__main__][INFO] - Iteration 621 took 50s (27.32% Gen, 69.64% Train). Generation: 13s, Training: 35s. Estimated remaining time: 32h 22m 14s. Estimated total time: 42h 7m 15s. Time estimates for 10 more iterations: 8m 25s, 100 more iterations: 1h 24m 14s, 500 more iterations: 7h 1m 12s. +[2026-03-26 06:50:41,063][__main__][INFO] - Starting iteration 621. +[2026-03-26 06:50:41,471][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 62 and human policies 1. +[2026-03-26 06:50:41,471][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:50:57,484][__main__][INFO] - Number of regex retries in iteration 621: 0 +[2026-03-26 06:50:57,485][__main__][INFO] - agents played in iteration 621 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:50:58,268][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:50:58,295][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:50:58,321][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:50:58,347][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:50:58,348][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:50:58,348][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:50:59,115][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:50:59,577][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:51:00,084][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:51:00,591][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:51:01,104][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:51:01,608][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:51:02,113][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:51:02,622][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:51:03,124][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:51:03,632][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:51:04,140][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:51:04,651][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:51:05,153][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:51:05,658][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:51:06,158][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:51:06,660][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:51:07,163][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:51:07,668][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:51:08,171][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:51:08,683][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:51:09,184][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:51:09,687][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:51:10,188][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:51:10,690][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:51:11,192][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:51:11,691][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:51:12,191][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:51:12,693][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:51:13,192][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:51:13,693][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:51:14,191][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:51:14,692][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:51:15,196][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:51:15,702][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:51:16,208][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:51:16,714][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:51:17,215][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:51:17,716][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:51:18,217][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:51:18,718][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:51:19,217][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:51:19,717][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:51:20,215][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:51:20,713][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:51:21,211][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:51:21,708][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:51:22,207][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:51:22,708][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:51:23,206][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:51:23,708][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:51:24,206][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:51:24,708][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:51:25,209][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:51:25,706][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:51:26,204][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:51:26,702][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:51:27,201][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:51:27,701][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:51:28,201][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:51:28,700][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:51:29,200][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:51:29,702][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:51:30,201][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:51:30,708][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:51:31,206][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10034 tokens. +[2026-03-26 06:51:32,034][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 60.52%, Block Peak % of device VRAM: 62.12%, ΔTime: 00:00:32 +[2026-03-26 06:51:32,779][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:51:32,781][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:51:32,783][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:51:33,508][__main__][INFO] - Iteration 622 took 52s (30.77% Gen, 67.83% Train). Generation: 16s, Training: 35s. Estimated remaining time: 33h 36m 0s. Estimated total time: 43h 21m 53s. Time estimates for 10 more iterations: 8m 40s, 100 more iterations: 1h 26m 43s, 500 more iterations: 7h 13m 38s. +[2026-03-26 06:51:33,510][__main__][INFO] - Starting iteration 622. +[2026-03-26 06:51:33,935][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 62 and human policies 1. +[2026-03-26 06:51:33,936][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:51:47,398][__main__][INFO] - Number of regex retries in iteration 622: 0 +[2026-03-26 06:51:47,398][__main__][INFO] - agents played in iteration 622 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:51:48,165][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:51:48,189][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:51:48,212][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:51:48,236][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:51:48,237][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:51:48,238][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:51:49,053][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:51:49,514][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:51:50,019][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:51:50,518][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:51:51,015][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:51:51,516][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:51:52,022][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:51:52,529][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:51:53,042][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:51:53,543][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:51:54,047][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:51:54,552][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:51:55,058][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:51:55,557][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:51:56,058][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:51:56,558][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:51:57,057][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:51:57,561][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:51:58,060][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:51:58,561][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:51:59,068][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:51:59,601][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:52:00,105][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:52:00,605][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:52:01,108][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:52:01,617][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:52:02,123][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:52:02,626][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:52:03,128][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:52:03,630][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:52:04,136][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:52:04,637][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:52:05,140][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:52:05,640][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:52:06,140][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:52:06,641][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:52:07,142][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:52:07,642][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:52:08,142][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:52:08,642][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:52:09,142][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:52:09,641][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:52:10,145][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:52:10,654][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:52:11,154][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:52:11,658][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:52:12,167][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:52:12,668][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:52:13,168][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:52:13,669][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:52:14,171][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:52:14,672][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:52:15,174][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:52:15,675][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:52:16,175][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:52:16,677][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:52:17,177][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:52:17,675][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:52:18,177][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:52:18,675][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:52:19,175][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:52:19,672][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:52:20,169][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:52:20,667][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:52:21,165][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10070 tokens. +[2026-03-26 06:52:22,019][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 60.51%, Block Peak % of device VRAM: 62.14%, ΔTime: 00:00:32 +[2026-03-26 06:52:22,772][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:52:22,775][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:52:22,776][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:52:23,505][__main__][INFO] - Iteration 623 took 49s (27.16% Gen, 71.37% Train). Generation: 13s, Training: 35s. Estimated remaining time: 31h 31m 47s. Estimated total time: 41h 18m 30s. Time estimates for 10 more iterations: 8m 15s, 100 more iterations: 1h 22m 37s, 500 more iterations: 6h 53m 5s. +[2026-03-26 06:52:23,507][__main__][INFO] - Starting iteration 623. +[2026-03-26 06:52:23,914][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 62 and human policies 1. +[2026-03-26 06:52:23,914][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:52:42,561][__main__][INFO] - Number of regex retries in iteration 623: 0 +[2026-03-26 06:52:42,562][__main__][INFO] - agents played in iteration 623 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:52:43,344][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:52:43,368][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:52:43,392][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:52:43,415][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:52:43,416][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:52:43,416][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:52:44,218][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:52:44,680][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:52:45,184][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:52:45,688][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:52:46,188][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:52:46,693][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:52:47,192][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:52:47,703][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:52:48,203][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:52:48,708][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:52:49,211][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:52:49,710][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:52:50,212][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:52:50,713][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:52:51,215][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:52:51,713][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:52:52,212][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:52:52,711][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:52:53,212][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:52:53,710][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:52:54,208][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:52:54,711][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:52:55,209][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:52:55,707][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:52:56,206][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:52:56,708][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:52:57,209][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:52:57,707][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:52:58,205][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:52:58,706][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:52:59,206][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:52:59,706][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:53:00,207][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:53:00,705][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:53:01,208][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:53:01,708][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:53:02,210][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:53:02,710][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:53:03,209][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:53:03,707][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:53:04,207][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:53:04,706][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:53:05,206][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:53:05,711][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:53:06,210][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:53:06,719][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:53:07,226][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:53:07,727][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:53:08,227][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:53:08,729][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:53:09,229][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:53:09,730][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:53:10,231][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:53:10,732][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:53:11,232][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:53:11,730][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:53:12,230][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:53:12,729][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:53:13,251][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:53:13,748][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:53:14,248][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:53:14,747][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:53:15,253][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:53:15,750][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:53:16,249][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10218 tokens. +[2026-03-26 06:53:17,079][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.57%, Block Peak % of device VRAM: 62.15%, ΔTime: 00:00:32 +[2026-03-26 06:53:17,828][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:53:17,830][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:53:17,832][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:53:18,657][__main__][INFO] - Iteration 624 took 54s (34.06% Gen, 64.43% Train). Generation: 18s, Training: 35s. Estimated remaining time: 35h 49m 32s. Estimated total time: 45h 37m 11s. Time estimates for 10 more iterations: 9m 7s, 100 more iterations: 1h 31m 14s, 500 more iterations: 7h 36m 11s. +[2026-03-26 06:53:18,659][__main__][INFO] - Starting iteration 624. +[2026-03-26 06:53:19,068][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 62 and human policies 1. +[2026-03-26 06:53:19,069][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:53:34,259][__main__][INFO] - Number of regex retries in iteration 624: 0 +[2026-03-26 06:53:34,260][__main__][INFO] - agents played in iteration 624 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:53:35,032][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:53:35,056][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:53:35,080][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:53:35,103][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:53:35,104][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:53:35,104][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:53:35,914][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:53:36,368][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:53:36,870][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:53:37,368][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:53:37,867][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:53:38,363][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:53:38,862][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:53:39,360][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:53:39,858][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:53:40,355][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:53:40,852][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:53:41,349][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:53:41,850][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:53:42,347][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:53:42,844][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:53:43,342][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:53:43,840][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:53:44,339][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:53:44,850][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:53:45,350][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:53:45,850][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:53:46,356][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:53:46,859][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:53:47,358][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:53:47,866][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:53:48,371][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:53:48,898][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:53:49,414][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:53:49,914][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:53:50,414][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:53:50,916][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:53:51,416][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:53:51,917][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:53:52,416][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:53:52,915][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:53:53,420][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:53:53,920][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:53:54,419][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:53:54,919][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:53:55,419][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:53:55,918][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:53:56,416][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:53:56,914][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:53:57,415][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:53:57,913][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:53:58,411][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:53:58,909][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:53:59,409][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:53:59,906][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:54:00,404][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:54:00,902][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:54:01,401][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:54:01,901][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:54:02,398][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:54:02,896][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:54:03,393][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:54:03,893][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:54:04,391][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:54:04,888][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:54:05,390][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:54:05,887][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:54:06,390][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:54:06,894][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:54:07,391][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:54:07,889][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10144 tokens. +[2026-03-26 06:54:08,734][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.58%, Block Peak % of device VRAM: 62.24%, ΔTime: 00:00:32 +[2026-03-26 06:54:09,478][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:54:09,480][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:54:09,482][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:54:10,210][__main__][INFO] - Iteration 625 took 51s (29.70% Gen, 68.87% Train). Generation: 15s, Training: 35s. Estimated remaining time: 32h 48m 38s. Estimated total time: 42h 37m 8s. Time estimates for 10 more iterations: 8m 31s, 100 more iterations: 1h 25m 14s, 500 more iterations: 7h 6m 11s. +[2026-03-26 06:54:10,212][__main__][INFO] - Starting iteration 625. +[2026-03-26 06:54:10,629][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 62 and human policies 1. +[2026-03-26 06:54:10,630][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:54:25,392][__main__][INFO] - Number of regex retries in iteration 625: 0 +[2026-03-26 06:54:25,392][__main__][INFO] - agents played in iteration 625 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:54:26,163][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:54:26,188][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:54:26,213][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:54:26,238][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:54:26,238][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:54:26,239][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:54:27,053][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:54:27,509][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:54:28,018][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:54:28,516][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:54:29,027][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:54:29,540][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:54:30,051][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:54:30,555][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:54:31,058][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:54:31,564][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:54:32,067][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:54:32,574][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:54:33,082][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:54:33,587][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:54:34,088][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:54:34,590][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:54:35,090][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:54:35,596][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:54:36,099][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:54:36,602][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:54:37,107][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:54:37,617][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:54:38,129][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:54:38,632][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:54:39,137][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:54:39,643][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:54:40,148][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:54:40,651][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:54:41,154][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:54:41,662][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:54:42,171][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:54:42,679][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:54:43,187][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:54:43,693][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:54:44,195][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:54:44,697][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:54:45,197][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:54:45,697][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:54:46,203][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:54:46,703][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:54:47,203][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:54:47,704][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:54:48,202][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:54:48,701][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:54:49,200][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:54:49,698][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:54:50,195][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:54:50,694][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:54:51,193][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:54:51,693][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:54:52,193][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:54:52,692][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:54:53,195][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:54:53,693][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:54:54,193][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:54:54,691][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:54:55,194][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:54:55,694][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:54:56,192][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:54:56,691][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:54:57,189][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:54:57,687][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:54:58,184][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:54:58,684][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:54:59,182][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10069 tokens. +[2026-03-26 06:55:00,036][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.03%, Current % of VRAM taken: 60.46%, Block Peak % of device VRAM: 62.00%, ΔTime: 00:00:32 +[2026-03-26 06:55:00,797][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:55:00,800][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:55:00,801][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:55:01,534][__main__][INFO] - Iteration 626 took 50s (29.00% Gen, 69.56% Train). Generation: 14s, Training: 35s. Estimated remaining time: 32h 35m 54s. Estimated total time: 42h 25m 16s. Time estimates for 10 more iterations: 8m 29s, 100 more iterations: 1h 24m 50s, 500 more iterations: 7h 4m 12s. +[2026-03-26 06:55:01,536][__main__][INFO] - Starting iteration 626. +[2026-03-26 06:55:01,954][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 62 and human policies 1. +[2026-03-26 06:55:01,955][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:55:17,210][__main__][INFO] - Number of regex retries in iteration 626: 0 +[2026-03-26 06:55:17,211][__main__][INFO] - agents played in iteration 626 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:55:17,984][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:55:18,009][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:55:18,033][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:55:18,057][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:55:18,058][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:55:18,058][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:55:18,889][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:55:19,344][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:55:19,851][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:55:20,352][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:55:20,860][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:55:21,362][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:55:21,869][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:55:22,377][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:55:22,880][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:55:23,386][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:55:23,893][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:55:24,400][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:55:24,902][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:55:25,407][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:55:25,912][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:55:26,415][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:55:26,917][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:55:27,423][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:55:27,924][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:55:28,428][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:55:28,932][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:55:29,438][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:55:29,946][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:55:30,455][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:55:30,963][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:55:31,469][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:55:31,978][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:55:32,482][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:55:32,984][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:55:33,485][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:55:33,987][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:55:34,489][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:55:34,991][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:55:35,493][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:55:35,996][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:55:36,502][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:55:37,007][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:55:37,508][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:55:38,011][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:55:38,513][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:55:39,013][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:55:39,516][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:55:40,016][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:55:40,516][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:55:41,016][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:55:41,518][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:55:42,019][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:55:42,517][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:55:43,021][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:55:43,520][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:55:44,018][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:55:44,514][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:55:45,012][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:55:45,512][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:55:46,008][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:55:46,506][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:55:47,005][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:55:47,504][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:55:48,007][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:55:48,504][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:55:49,002][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:55:49,503][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:55:50,000][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:55:50,497][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:55:50,997][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10184 tokens. +[2026-03-26 06:55:51,862][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 60.60%, Block Peak % of device VRAM: 62.21%, ΔTime: 00:00:32 +[2026-03-26 06:55:52,606][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:55:52,609][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:55:52,610][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:55:53,338][__main__][INFO] - Iteration 627 took 51s (29.69% Gen, 68.89% Train). Generation: 15s, Training: 35s. Estimated remaining time: 32h 59m 1s. Estimated total time: 42h 49m 14s. Time estimates for 10 more iterations: 8m 33s, 100 more iterations: 1h 25m 38s, 500 more iterations: 7h 8m 12s. +[2026-03-26 06:55:53,340][__main__][INFO] - Starting iteration 627. +[2026-03-26 06:55:53,759][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 62 and human policies 1. +[2026-03-26 06:55:53,759][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:56:07,526][__main__][INFO] - Number of regex retries in iteration 627: 0 +[2026-03-26 06:56:07,527][__main__][INFO] - agents played in iteration 627 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:56:08,307][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:56:08,331][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:56:08,354][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:56:08,378][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:56:08,378][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:56:08,379][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:56:09,180][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:56:09,638][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:56:10,145][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:56:10,649][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:56:11,161][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:56:11,663][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:56:12,166][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:56:12,672][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:56:13,172][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:56:13,674][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:56:14,177][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:56:14,677][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:56:15,177][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:56:15,677][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:56:16,179][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:56:16,677][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:56:17,178][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:56:17,677][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:56:18,179][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:56:18,678][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:56:19,181][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:56:19,682][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:56:20,182][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:56:20,681][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:56:21,181][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:56:21,682][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:56:22,181][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:56:22,678][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:56:23,179][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:56:23,679][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:56:24,180][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:56:24,684][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:56:25,185][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:56:25,695][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:56:26,199][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:56:26,699][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:56:27,200][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:56:27,701][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:56:28,204][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:56:28,705][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:56:29,209][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:56:29,708][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:56:30,209][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:56:30,710][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:56:31,216][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:56:31,716][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:56:32,216][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:56:32,722][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:56:33,225][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:56:33,749][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:56:34,250][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:56:34,752][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:56:35,254][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:56:35,756][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:56:36,259][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:56:36,765][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:56:37,267][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:56:37,765][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:56:38,265][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:56:38,763][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:56:39,262][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:56:39,761][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:56:40,259][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:56:40,757][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:56:41,255][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10079 tokens. +[2026-03-26 06:56:42,112][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 60.53%, Block Peak % of device VRAM: 62.18%, ΔTime: 00:00:32 +[2026-03-26 06:56:42,888][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:56:42,895][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:56:42,897][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:56:43,590][__main__][INFO] - Iteration 628 took 49s (27.63% Gen, 70.98% Train). Generation: 13s, Training: 35s. Estimated remaining time: 31h 40m 34s. Estimated total time: 41h 31m 37s. Time estimates for 10 more iterations: 8m 18s, 100 more iterations: 1h 23m 3s, 500 more iterations: 6h 55m 16s. +[2026-03-26 06:56:43,593][__main__][INFO] - Starting iteration 628. +[2026-03-26 06:56:43,997][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 62 and human policies 1. +[2026-03-26 06:56:43,998][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:56:57,544][__main__][INFO] - Number of regex retries in iteration 628: 0 +[2026-03-26 06:56:57,545][__main__][INFO] - agents played in iteration 628 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:56:58,320][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:56:58,343][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:56:58,367][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:56:58,390][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:56:58,391][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:56:58,392][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:56:59,210][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:56:59,666][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:57:00,169][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:57:00,667][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:57:01,173][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:57:01,676][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:57:02,175][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:57:02,674][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:57:03,172][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:57:03,670][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:57:04,172][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:57:04,671][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:57:05,169][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:57:05,667][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:57:06,165][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:57:06,666][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:57:07,168][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:57:07,667][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:57:08,173][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:57:08,673][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:57:09,173][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:57:09,680][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:57:10,181][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:57:10,693][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:57:11,194][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:57:11,695][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:57:12,197][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:57:12,699][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:57:13,202][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:57:13,704][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:57:14,208][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:57:14,709][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:57:15,211][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:57:15,715][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:57:16,218][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:57:16,721][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:57:17,223][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:57:17,726][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:57:18,228][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:57:18,735][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:57:19,238][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:57:19,736][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:57:20,236][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:57:20,734][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:57:21,234][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:57:21,732][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:57:22,230][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:57:22,729][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:57:23,227][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:57:23,725][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:57:24,222][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:57:24,723][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:57:25,222][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:57:25,721][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:57:26,218][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:57:26,716][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:57:27,214][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:57:27,711][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:57:28,208][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:57:28,710][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:57:29,211][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:57:29,710][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:57:30,213][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:57:30,713][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:57:31,212][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10022 tokens. +[2026-03-26 06:57:32,094][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 60.49%, Block Peak % of device VRAM: 62.28%, ΔTime: 00:00:32 +[2026-03-26 06:57:32,860][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:57:32,863][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:57:32,865][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:57:33,727][__main__][INFO] - Iteration 629 took 49s (27.24% Gen, 71.02% Train). Generation: 13s, Training: 35s. Estimated remaining time: 31h 34m 36s. Estimated total time: 41h 26m 29s. Time estimates for 10 more iterations: 8m 17s, 100 more iterations: 1h 22m 52s, 500 more iterations: 6h 54m 24s. +[2026-03-26 06:57:33,730][__main__][INFO] - Starting iteration 629. +[2026-03-26 06:57:34,149][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 62 and human policies 1. +[2026-03-26 06:57:34,150][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:57:47,084][__main__][INFO] - Number of regex retries in iteration 629: 0 +[2026-03-26 06:57:47,085][__main__][INFO] - agents played in iteration 629 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:57:47,939][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:57:47,965][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:57:47,990][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:57:48,015][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:57:48,016][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:57:48,016][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:57:48,847][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:57:49,311][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:57:49,824][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:57:50,333][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:57:50,839][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:57:51,341][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:57:51,844][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:57:52,346][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:57:52,851][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:57:53,358][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:57:53,871][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:57:54,376][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:57:54,878][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:57:55,379][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:57:55,878][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:57:56,377][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:57:56,877][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:57:57,387][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:57:57,892][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:57:58,405][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:57:58,906][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:57:59,410][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:57:59,911][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:58:00,413][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:58:00,914][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:58:01,415][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:58:01,921][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:58:02,430][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:58:02,938][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:58:03,451][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:58:03,953][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:58:04,456][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:58:04,959][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:58:05,463][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:58:05,965][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:58:06,468][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:58:06,972][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:58:07,474][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:58:07,979][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:58:08,487][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:58:09,017][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:58:09,525][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:58:10,028][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:58:10,530][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:58:11,034][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:58:11,535][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:58:12,036][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:58:12,539][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:58:13,041][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:58:13,542][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:58:14,043][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:58:14,542][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:58:15,044][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:58:15,547][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:58:16,050][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:58:16,556][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:58:17,061][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:58:17,561][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:58:18,062][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:58:18,563][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:58:19,064][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:58:19,565][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:58:20,067][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:58:20,568][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:58:21,069][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10049 tokens. +[2026-03-26 06:58:21,939][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 60.54%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:33 +[2026-03-26 06:58:22,693][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:58:22,695][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:58:22,697][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:58:23,448][__main__][INFO] - Iteration 630 took 49s (26.24% Gen, 72.24% Train). Generation: 12s, Training: 35s. Estimated remaining time: 31h 12m 15s. Estimated total time: 41h 4m 58s. Time estimates for 10 more iterations: 8m 12s, 100 more iterations: 1h 22m 9s, 500 more iterations: 6h 50m 49s. +[2026-03-26 06:58:23,451][__main__][INFO] - Starting iteration 630. +[2026-03-26 06:58:23,849][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 62 and human policies 1. +[2026-03-26 06:58:23,849][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:58:38,150][__main__][INFO] - Number of regex retries in iteration 630: 0 +[2026-03-26 06:58:38,151][__main__][INFO] - agents played in iteration 630 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:58:38,939][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:58:38,963][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:58:38,987][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:58:39,010][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:58:39,011][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:58:39,011][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:58:39,818][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:58:40,276][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:58:40,781][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:58:41,283][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:58:41,788][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:58:42,290][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:58:42,796][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:58:43,311][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:58:43,811][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:58:44,318][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:58:44,819][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:58:45,322][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:58:45,824][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:58:46,324][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:58:46,829][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:58:47,329][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:58:47,833][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:58:48,335][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:58:48,835][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:58:49,334][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:58:49,833][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:58:50,331][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:58:50,829][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:58:51,329][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:58:51,829][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:58:52,332][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:58:52,834][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:58:53,332][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:58:53,834][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:58:54,336][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:58:54,833][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:58:55,335][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:58:55,834][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:58:56,333][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:58:56,832][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:58:57,330][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:58:57,829][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:58:58,331][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:58:58,833][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:58:59,334][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:58:59,835][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:59:00,332][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:59:00,829][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:59:01,327][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:59:01,828][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:59:02,327][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:59:02,825][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:59:03,324][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:59:03,822][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:59:04,319][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:59:04,816][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:59:05,312][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:59:05,810][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:59:06,306][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:59:06,804][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:59:07,302][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 06:59:07,799][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 06:59:08,300][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 06:59:08,800][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 06:59:09,297][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 06:59:09,798][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 06:59:10,301][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 06:59:10,798][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 06:59:11,297][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 06:59:11,799][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10155 tokens. +[2026-03-26 06:59:12,638][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.08%, ΔTime: 00:00:32 +[2026-03-26 06:59:13,387][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 06:59:13,389][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 06:59:13,391][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 06:59:14,762][__main__][INFO] - Iteration 631 took 50s (28.09% Gen, 69.21% Train). Generation: 14s, Training: 35s. Estimated remaining time: 32h 32m 7s. Estimated total time: 42h 25m 41s. Time estimates for 10 more iterations: 8m 29s, 100 more iterations: 1h 24m 51s, 500 more iterations: 7h 4m 16s. +[2026-03-26 06:59:14,764][__main__][INFO] - Starting iteration 631. +[2026-03-26 06:59:15,172][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 63 and human policies 1. +[2026-03-26 06:59:15,172][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 06:59:30,619][__main__][INFO] - Number of regex retries in iteration 631: 0 +[2026-03-26 06:59:30,620][__main__][INFO] - agents played in iteration 631 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 06:59:31,387][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:59:31,411][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:59:31,434][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:59:31,458][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 06:59:31,458][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 06:59:31,459][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 06:59:32,274][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 06:59:32,733][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 06:59:33,235][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 06:59:33,732][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 06:59:34,230][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 06:59:34,727][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 06:59:35,224][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 06:59:35,727][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 06:59:36,231][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 06:59:36,734][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 06:59:37,233][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 06:59:37,735][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 06:59:38,233][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 06:59:38,732][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 06:59:39,232][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 06:59:39,734][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 06:59:40,234][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 06:59:40,746][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 06:59:41,251][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 06:59:41,753][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 06:59:42,253][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 06:59:42,752][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 06:59:43,251][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 06:59:43,751][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 06:59:44,257][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 06:59:44,757][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 06:59:45,258][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 06:59:45,757][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 06:59:46,257][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 06:59:46,757][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 06:59:47,255][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 06:59:47,755][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 06:59:48,254][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 06:59:48,753][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 06:59:49,279][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 06:59:49,775][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 06:59:50,274][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 06:59:50,773][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 06:59:51,275][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 06:59:51,777][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 06:59:52,277][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 06:59:52,775][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 06:59:53,274][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 06:59:53,775][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 06:59:54,279][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 06:59:54,780][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 06:59:55,280][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 06:59:55,784][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 06:59:56,287][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 06:59:56,796][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 06:59:57,300][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 06:59:57,805][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 06:59:58,312][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 06:59:58,816][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 06:59:59,316][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 06:59:59,816][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:00:00,317][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:00:00,819][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:00:01,319][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:00:01,819][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:00:02,322][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:00:02,824][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:00:03,327][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:00:03,828][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:00:04,328][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10140 tokens. +[2026-03-26 07:00:05,192][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 60.56%, Block Peak % of device VRAM: 62.13%, ΔTime: 00:00:32 +[2026-03-26 07:00:05,948][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:00:05,950][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:00:05,952][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:00:06,684][__main__][INFO] - Iteration 632 took 51s (29.99% Gen, 68.59% Train). Generation: 15s, Training: 35s. Estimated remaining time: 33h 1m 14s. Estimated total time: 42h 55m 40s. Time estimates for 10 more iterations: 8m 35s, 100 more iterations: 1h 25m 51s, 500 more iterations: 7h 9m 16s. +[2026-03-26 07:00:06,687][__main__][INFO] - Starting iteration 632. +[2026-03-26 07:00:07,095][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 63 and human policies 1. +[2026-03-26 07:00:07,095][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:00:22,040][__main__][INFO] - Number of regex retries in iteration 632: 0 +[2026-03-26 07:00:22,041][__main__][INFO] - agents played in iteration 632 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:00:22,813][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:00:22,836][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:00:22,860][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:00:22,883][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:00:22,884][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:00:22,884][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:00:23,697][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:00:24,152][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:00:24,662][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:00:25,160][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:00:25,658][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:00:26,156][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:00:26,654][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:00:27,156][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:00:27,656][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:00:28,157][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:00:28,661][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:00:29,160][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:00:29,660][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:00:30,157][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:00:30,661][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:00:31,168][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:00:31,670][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:00:32,169][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:00:32,674][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:00:33,173][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:00:33,672][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:00:34,177][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:00:34,677][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:00:35,177][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:00:35,689][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:00:36,193][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:00:36,695][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:00:37,196][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:00:37,701][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:00:38,200][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:00:38,703][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:00:39,205][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:00:39,706][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:00:40,207][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:00:40,707][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:00:41,210][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:00:41,710][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:00:42,210][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:00:42,711][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:00:43,211][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:00:43,711][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:00:44,210][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:00:44,709][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:00:45,208][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:00:45,708][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:00:46,208][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:00:46,706][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:00:47,206][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:00:47,707][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:00:48,206][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:00:48,707][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:00:49,207][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:00:49,706][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:00:50,206][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:00:50,705][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:00:51,203][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:00:51,702][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:00:52,200][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:00:52,698][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:00:53,197][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:00:53,695][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:00:54,192][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:00:54,691][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:00:55,194][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:00:55,697][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10098 tokens. +[2026-03-26 07:00:56,565][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:32 +[2026-03-26 07:00:57,327][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:00:57,330][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:00:57,332][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:00:58,077][__main__][INFO] - Iteration 633 took 50s (29.31% Gen, 69.22% Train). Generation: 14s, Training: 35s. Estimated remaining time: 32h 33m 50s. Estimated total time: 42h 29m 8s. Time estimates for 10 more iterations: 8m 29s, 100 more iterations: 1h 24m 58s, 500 more iterations: 7h 4m 51s. +[2026-03-26 07:00:58,080][__main__][INFO] - Starting iteration 633. +[2026-03-26 07:00:58,500][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 63 and human policies 1. +[2026-03-26 07:00:58,500][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:01:14,133][__main__][INFO] - Number of regex retries in iteration 633: 0 +[2026-03-26 07:01:14,134][__main__][INFO] - agents played in iteration 633 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:01:14,903][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:01:14,929][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:01:14,954][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:01:14,979][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:01:14,979][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:01:14,980][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:01:15,794][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:01:16,259][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:01:16,767][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:01:17,281][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:01:17,801][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:01:18,307][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:01:18,814][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:01:19,319][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:01:19,825][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:01:20,327][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:01:20,832][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:01:21,337][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:01:21,851][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:01:22,354][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:01:22,858][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:01:23,361][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:01:23,864][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:01:24,367][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:01:24,871][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:01:25,371][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:01:25,875][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:01:26,380][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:01:26,882][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:01:27,383][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:01:27,884][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:01:28,385][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:01:28,887][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:01:29,387][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:01:29,893][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:01:30,394][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:01:30,899][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:01:31,402][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:01:31,902][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:01:32,406][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:01:32,909][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:01:33,411][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:01:33,913][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:01:34,417][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:01:34,923][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:01:35,439][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:01:35,940][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:01:36,445][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:01:36,948][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:01:37,450][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:01:37,952][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:01:38,453][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:01:38,954][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:01:39,455][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:01:39,956][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:01:40,458][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:01:40,958][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:01:41,461][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:01:41,960][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:01:42,461][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:01:42,962][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:01:43,462][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:01:43,963][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:01:44,466][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:01:44,966][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:01:45,465][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:01:45,969][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:01:46,468][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:01:46,967][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:01:47,471][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:01:47,971][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10224 tokens. +[2026-03-26 07:01:48,816][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.24%, ΔTime: 00:00:33 +[2026-03-26 07:01:49,568][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:01:49,570][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:01:49,572][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:01:50,303][__main__][INFO] - Iteration 634 took 51s (30.18% Gen, 68.41% Train). Generation: 15s, Training: 35s. Estimated remaining time: 33h 14m 2s. Estimated total time: 43h 10m 12s. Time estimates for 10 more iterations: 8m 38s, 100 more iterations: 1h 26m 20s, 500 more iterations: 7h 11m 42s. +[2026-03-26 07:01:50,306][__main__][INFO] - Starting iteration 634. +[2026-03-26 07:01:50,711][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 63 and human policies 1. +[2026-03-26 07:01:50,712][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:02:00,872][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 07:02:05,427][__main__][INFO] - Number of regex retries in iteration 634: 1 +[2026-03-26 07:02:05,428][__main__][INFO] - agents played in iteration 634 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:02:06,187][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:02:06,211][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:02:06,235][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:02:06,259][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:02:06,259][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:02:06,260][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:02:07,081][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:02:07,554][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:02:08,064][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:02:08,567][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:02:09,076][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:02:09,576][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:02:10,090][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:02:10,597][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:02:11,102][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:02:11,603][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:02:12,103][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:02:12,603][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:02:13,102][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:02:13,601][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:02:14,100][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:02:14,601][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:02:15,104][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:02:15,605][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:02:16,105][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:02:16,605][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:02:17,106][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:02:17,611][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:02:18,116][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:02:18,622][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:02:19,121][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:02:19,620][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:02:20,125][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:02:20,624][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:02:21,125][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:02:21,627][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:02:22,126][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:02:22,624][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:02:23,122][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:02:23,625][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:02:24,123][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:02:24,625][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:02:25,123][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:02:25,622][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:02:26,121][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:02:26,619][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:02:27,117][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:02:27,613][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:02:28,109][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:02:28,608][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:02:29,105][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:02:29,602][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:02:30,102][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:02:30,598][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:02:31,094][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:02:31,591][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:02:32,091][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:02:32,589][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:02:33,087][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:02:33,589][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:02:34,086][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:02:34,583][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:02:35,082][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:02:35,585][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:02:36,085][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:02:36,586][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:02:37,087][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:02:37,588][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:02:38,089][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:02:38,589][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:02:39,089][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10180 tokens. +[2026-03-26 07:02:39,949][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:32 +[2026-03-26 07:02:40,701][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:02:40,703][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:02:40,705][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:02:41,438][__main__][INFO] - Iteration 635 took 50s (29.01% Gen, 69.54% Train). Generation: 14s, Training: 35s. Estimated remaining time: 32h 19m 20s. Estimated total time: 42h 16m 21s. Time estimates for 10 more iterations: 8m 27s, 100 more iterations: 1h 24m 32s, 500 more iterations: 7h 2m 43s. +[2026-03-26 07:02:41,440][__main__][INFO] - Starting iteration 635. +[2026-03-26 07:02:41,860][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 63 and human policies 1. +[2026-03-26 07:02:41,860][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:02:47,041][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 07:02:47,042][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 07:02:56,926][__main__][INFO] - Number of regex retries in iteration 635: 2 +[2026-03-26 07:02:56,927][__main__][INFO] - agents played in iteration 635 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:02:57,696][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:02:57,720][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:02:57,744][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:02:57,768][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:02:57,769][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:02:57,770][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:02:58,590][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:02:59,049][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:02:59,569][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:03:00,081][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:03:00,584][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:03:01,089][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:03:01,593][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:03:02,098][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:03:02,601][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:03:03,106][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:03:03,613][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:03:04,122][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:03:04,624][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:03:05,127][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:03:05,630][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:03:06,132][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:03:06,636][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:03:07,137][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:03:07,637][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:03:08,137][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:03:08,636][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:03:09,135][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:03:09,633][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:03:10,135][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:03:10,634][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:03:11,138][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:03:11,644][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:03:12,153][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:03:12,654][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:03:13,158][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:03:13,659][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:03:14,159][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:03:14,658][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:03:15,159][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:03:15,659][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:03:16,157][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:03:16,656][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:03:17,156][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:03:17,654][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:03:18,152][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:03:18,650][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:03:19,150][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:03:19,657][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:03:20,162][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:03:20,669][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:03:21,182][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:03:21,684][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:03:22,187][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:03:22,692][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:03:23,196][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:03:23,696][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:03:24,196][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:03:24,696][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:03:25,197][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:03:25,696][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:03:26,196][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:03:26,696][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:03:27,196][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:03:27,700][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:03:28,205][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:03:28,710][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:03:29,217][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:03:29,719][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:03:30,221][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:03:30,723][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10059 tokens. +[2026-03-26 07:03:31,599][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 60.54%, Block Peak % of device VRAM: 62.11%, ΔTime: 00:00:33 +[2026-03-26 07:03:32,339][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:03:32,341][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:03:32,343][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:03:33,075][__main__][INFO] - Iteration 636 took 51s (29.42% Gen, 69.15% Train). Generation: 15s, Training: 35s. Estimated remaining time: 32h 42m 56s. Estimated total time: 42h 40m 49s. Time estimates for 10 more iterations: 8m 32s, 100 more iterations: 1h 25m 21s, 500 more iterations: 7h 6m 48s. +[2026-03-26 07:03:33,078][__main__][INFO] - Starting iteration 636. +[2026-03-26 07:03:33,500][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 63 and human policies 1. +[2026-03-26 07:03:33,501][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:03:48,500][__main__][INFO] - Number of regex retries in iteration 636: 0 +[2026-03-26 07:03:48,501][__main__][INFO] - agents played in iteration 636 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:03:49,269][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:03:49,295][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:03:49,322][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:03:49,349][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:03:49,349][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:03:49,350][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:03:50,183][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:03:50,645][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:03:51,153][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:03:51,660][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:03:52,174][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:03:52,679][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:03:53,183][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:03:53,688][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:03:54,196][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:03:54,700][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:03:55,203][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:03:55,710][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:03:56,224][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:03:56,730][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:03:57,233][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:03:57,740][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:03:58,241][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:03:58,749][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:03:59,252][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:03:59,756][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:04:00,266][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:04:00,777][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:04:01,286][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:04:01,790][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:04:02,293][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:04:02,795][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:04:03,296][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:04:03,799][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:04:04,305][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:04:04,805][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:04:05,306][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:04:05,812][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:04:06,314][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:04:06,825][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:04:07,327][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:04:07,831][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:04:08,332][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:04:08,836][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:04:09,335][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:04:09,837][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:04:10,337][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:04:10,837][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:04:11,337][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:04:11,839][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:04:12,337][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:04:12,835][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:04:13,334][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:04:13,834][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:04:14,335][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:04:14,836][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:04:15,340][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:04:15,844][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:04:16,346][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:04:16,847][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:04:17,349][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:04:17,851][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:04:18,352][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:04:18,849][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:04:19,351][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:04:19,850][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:04:20,348][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:04:20,846][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:04:21,346][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:04:21,844][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:04:22,343][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10151 tokens. +[2026-03-26 07:04:23,207][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.05%, Current % of VRAM taken: 60.48%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:33 +[2026-03-26 07:04:23,954][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:04:23,956][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:04:23,957][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:04:24,690][__main__][INFO] - Iteration 637 took 51s (29.30% Gen, 69.26% Train). Generation: 15s, Training: 35s. Estimated remaining time: 32h 40m 46s. Estimated total time: 42h 39m 31s. Time estimates for 10 more iterations: 8m 31s, 100 more iterations: 1h 25m 19s, 500 more iterations: 7h 6m 35s. +[2026-03-26 07:04:24,692][__main__][INFO] - Starting iteration 637. +[2026-03-26 07:04:25,095][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 63 and human policies 1. +[2026-03-26 07:04:25,095][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:04:40,638][__main__][INFO] - Number of regex retries in iteration 637: 0 +[2026-03-26 07:04:40,639][__main__][INFO] - agents played in iteration 637 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:04:41,406][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:04:41,431][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:04:41,455][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:04:41,480][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:04:41,481][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:04:41,481][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:04:42,304][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:04:42,769][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:04:43,277][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:04:43,776][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:04:44,282][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:04:44,799][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:04:45,306][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:04:45,809][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:04:46,313][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:04:46,817][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:04:47,319][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:04:47,825][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:04:48,339][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:04:48,852][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:04:49,356][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:04:49,860][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:04:50,365][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:04:50,866][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:04:51,367][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:04:51,873][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:04:52,378][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:04:52,892][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:04:53,395][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:04:53,897][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:04:54,402][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:04:54,906][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:04:55,415][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:04:55,918][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:04:56,422][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:04:56,930][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:04:57,435][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:04:57,949][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:04:58,453][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:04:58,955][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:04:59,457][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:04:59,962][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:05:00,462][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:05:00,965][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:05:01,468][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:05:01,973][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:05:02,473][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:05:02,972][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:05:03,476][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:05:03,983][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:05:04,488][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:05:04,996][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:05:05,501][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:05:06,001][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:05:06,502][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:05:07,004][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:05:07,503][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:05:08,003][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:05:08,504][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:05:09,004][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:05:09,504][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:05:10,005][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:05:10,509][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:05:11,014][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:05:11,519][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:05:12,029][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:05:12,531][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:05:13,032][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:05:13,536][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:05:14,040][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:05:14,540][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10196 tokens. +[2026-03-26 07:05:15,408][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 60.50%, Block Peak % of device VRAM: 62.33%, ΔTime: 00:00:33 +[2026-03-26 07:05:16,171][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:05:16,173][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:05:16,175][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:05:16,914][__main__][INFO] - Iteration 638 took 51s (30.00% Gen, 68.57% Train). Generation: 15s, Training: 35s. Estimated remaining time: 33h 11m 24s. Estimated total time: 43h 11m 1s. Time estimates for 10 more iterations: 8m 38s, 100 more iterations: 1h 26m 22s, 500 more iterations: 7h 11m 50s. +[2026-03-26 07:05:16,917][__main__][INFO] - Starting iteration 638. +[2026-03-26 07:05:17,332][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 63 and human policies 1. +[2026-03-26 07:05:17,332][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:05:28,673][__main__][INFO] - Number of regex retries in iteration 638: 0 +[2026-03-26 07:05:28,674][__main__][INFO] - agents played in iteration 638 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:05:29,517][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:05:29,541][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:05:29,565][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:05:29,588][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:05:29,589][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:05:29,589][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:05:30,398][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:05:30,856][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:05:31,364][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:05:31,868][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:05:32,367][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:05:32,866][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:05:33,367][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:05:33,877][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:05:34,380][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:05:34,886][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:05:35,400][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:05:35,901][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:05:36,403][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:05:36,905][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:05:37,407][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:05:37,910][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:05:38,411][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:05:38,913][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:05:39,413][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:05:39,913][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:05:40,413][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:05:40,913][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:05:41,414][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:05:41,913][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:05:42,413][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:05:42,913][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:05:43,412][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:05:43,912][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:05:44,412][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:05:44,911][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:05:45,410][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:05:45,909][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:05:46,411][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:05:46,912][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:05:47,414][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:05:47,912][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:05:48,417][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:05:48,925][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:05:49,429][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:05:49,936][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:05:50,441][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:05:50,947][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:05:51,450][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:05:51,952][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:05:52,454][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:05:52,955][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:05:53,458][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:05:53,959][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:05:54,459][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:05:54,967][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:05:55,467][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:05:55,968][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:05:56,468][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:05:56,969][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:05:57,469][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:05:57,970][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:05:58,469][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:05:58,971][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:05:59,473][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:05:59,973][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:06:00,473][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:06:00,973][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:06:01,473][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:06:01,974][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:06:02,475][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10093 tokens. +[2026-03-26 07:06:03,344][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 60.56%, Block Peak % of device VRAM: 62.11%, ΔTime: 00:00:32 +[2026-03-26 07:06:04,091][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:06:04,093][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:06:04,095][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:06:04,951][__main__][INFO] - Iteration 639 took 47s (23.82% Gen, 74.38% Train). Generation: 11s, Training: 35s. Estimated remaining time: 29h 40m 34s. Estimated total time: 39h 40m 59s. Time estimates for 10 more iterations: 7m 56s, 100 more iterations: 1h 19m 21s, 500 more iterations: 6h 36m 49s. +[2026-03-26 07:06:04,953][__main__][INFO] - Starting iteration 639. +[2026-03-26 07:06:05,367][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 63 and human policies 1. +[2026-03-26 07:06:05,368][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:06:19,918][__main__][INFO] - Number of regex retries in iteration 639: 0 +[2026-03-26 07:06:19,919][__main__][INFO] - agents played in iteration 639 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:06:20,700][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:06:20,723][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:06:20,747][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:06:20,771][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:06:20,771][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:06:20,772][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:06:21,593][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:06:22,063][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:06:22,565][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:06:23,067][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:06:23,566][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:06:24,066][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:06:24,566][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:06:25,067][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:06:25,572][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:06:26,070][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:06:26,568][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:06:27,066][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:06:27,563][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:06:28,060][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:06:28,557][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:06:29,054][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:06:29,552][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:06:30,058][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:06:30,561][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:06:31,062][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:06:31,562][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:06:32,066][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:06:32,586][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:06:33,087][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:06:33,589][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:06:34,088][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:06:34,588][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:06:35,087][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:06:35,586][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:06:36,091][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:06:36,590][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:06:37,092][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:06:37,593][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:06:38,096][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:06:38,597][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:06:39,099][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:06:39,607][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:06:40,108][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:06:40,613][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:06:41,113][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:06:41,614][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:06:42,113][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:06:42,613][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:06:43,113][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:06:43,611][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:06:44,111][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:06:44,609][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:06:45,109][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:06:45,608][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:06:46,107][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:06:46,607][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:06:47,107][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:06:47,606][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:06:48,106][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:06:48,605][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:06:49,105][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:06:49,603][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:06:50,106][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:06:50,606][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:06:51,105][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:06:51,607][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:06:52,105][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:06:52,607][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:06:53,106][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:06:53,605][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10121 tokens. +[2026-03-26 07:06:54,448][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.46%, ΔTime: 00:00:32 +[2026-03-26 07:06:55,197][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:06:55,200][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:06:55,201][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:06:56,056][__main__][INFO] - Iteration 640 took 50s (28.71% Gen, 69.60% Train). Generation: 14s, Training: 35s. Estimated remaining time: 32h 13m 14s. Estimated total time: 42h 14m 30s. Time estimates for 10 more iterations: 8m 26s, 100 more iterations: 1h 24m 29s, 500 more iterations: 7h 2m 25s. +[2026-03-26 07:06:56,059][__main__][INFO] - Starting iteration 640. +[2026-03-26 07:06:56,461][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 63 and human policies 1. +[2026-03-26 07:06:56,462][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:07:10,739][__main__][INFO] - Number of regex retries in iteration 640: 0 +[2026-03-26 07:07:10,741][__main__][INFO] - agents played in iteration 640 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:07:11,507][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:07:11,533][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:07:11,558][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:07:11,583][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:07:11,584][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:07:11,585][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:07:12,406][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:07:12,866][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:07:13,371][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:07:13,872][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:07:14,373][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:07:14,873][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:07:15,375][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:07:15,875][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:07:16,377][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:07:16,889][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:07:17,398][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:07:17,902][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:07:18,407][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:07:18,916][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:07:19,417][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:07:19,922][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:07:20,428][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:07:20,936][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:07:21,437][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:07:21,939][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:07:22,440][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:07:22,945][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:07:23,449][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:07:23,950][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:07:24,457][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:07:24,964][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:07:25,471][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:07:26,001][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:07:26,507][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:07:27,010][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:07:27,511][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:07:28,013][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:07:28,516][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:07:29,024][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:07:29,537][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:07:30,039][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:07:30,541][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:07:31,044][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:07:31,548][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:07:32,053][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:07:32,554][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:07:33,053][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:07:33,551][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:07:34,050][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:07:34,550][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:07:35,049][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:07:35,549][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:07:36,048][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:07:36,550][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:07:37,051][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:07:37,557][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:07:38,063][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:07:38,573][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:07:39,075][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:07:39,580][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:07:40,081][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:07:40,583][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:07:41,084][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:07:41,586][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:07:42,086][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:07:42,587][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:07:43,087][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:07:43,586][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:07:44,089][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:07:44,589][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10062 tokens. +[2026-03-26 07:07:45,448][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.20%, ΔTime: 00:00:33 +[2026-03-26 07:07:46,203][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:07:46,205][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:07:46,207][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:07:47,603][__main__][INFO] - Iteration 641 took 51s (27.92% Gen, 69.35% Train). Generation: 14s, Training: 35s. Estimated remaining time: 32h 35m 0s. Estimated total time: 42h 37m 7s. Time estimates for 10 more iterations: 8m 31s, 100 more iterations: 1h 25m 14s, 500 more iterations: 7h 6m 11s. +[2026-03-26 07:07:47,605][__main__][INFO] - Starting iteration 641. +[2026-03-26 07:07:48,015][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 64 and human policies 1. +[2026-03-26 07:07:48,015][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:08:04,815][mllm.models.large_language_model_local][WARNING] - Response Propagation: 0 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 07:08:13,083][__main__][INFO] - Number of regex retries in iteration 641: 1 +[2026-03-26 07:08:13,084][__main__][INFO] - agents played in iteration 641 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:08:13,844][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:08:13,868][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:08:13,892][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:08:13,915][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:08:13,916][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:08:13,916][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:08:14,728][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:08:15,183][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:08:15,692][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:08:16,195][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:08:16,703][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:08:17,214][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:08:17,716][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:08:18,221][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:08:18,721][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:08:19,230][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:08:19,735][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:08:20,238][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:08:20,745][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:08:21,253][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:08:21,757][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:08:22,264][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:08:22,769][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:08:23,271][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:08:23,774][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:08:24,276][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:08:24,777][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:08:25,304][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:08:25,806][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:08:26,307][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:08:26,811][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:08:27,310][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:08:27,814][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:08:28,314][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:08:28,818][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:08:29,317][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:08:29,817][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:08:30,326][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:08:30,838][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:08:31,346][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:08:31,849][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:08:32,355][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:08:32,855][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:08:33,356][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:08:33,856][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:08:34,357][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:08:34,857][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:08:35,356][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:08:35,855][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:08:36,353][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:08:36,852][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:08:37,351][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:08:37,855][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:08:38,359][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:08:38,867][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:08:39,368][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:08:39,868][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:08:40,371][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:08:40,871][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:08:41,373][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:08:41,875][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:08:42,374][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:08:42,874][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:08:43,376][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:08:43,877][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:08:44,382][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:08:44,881][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:08:45,381][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:08:45,885][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:08:46,385][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:08:46,893][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10180 tokens. +[2026-03-26 07:08:47,767][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 60.65%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:33 +[2026-03-26 07:08:48,534][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:08:48,538][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:08:48,540][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:08:49,311][__main__][INFO] - Iteration 642 took 1m 1s (40.90% Gen, 57.84% Train). Generation: 25s, Training: 35s. Estimated remaining time: 41h 1m 41s. Estimated total time: 51h 4m 50s. Time estimates for 10 more iterations: 10m 12s, 100 more iterations: 1h 42m 9s, 500 more iterations: 8h 30m 48s. +[2026-03-26 07:08:49,313][__main__][INFO] - Starting iteration 642. +[2026-03-26 07:08:49,900][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 64 and human policies 1. +[2026-03-26 07:08:49,901][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:09:11,885][__main__][INFO] - Number of regex retries in iteration 642: 0 +[2026-03-26 07:09:11,886][__main__][INFO] - agents played in iteration 642 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:09:12,662][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:09:12,685][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:09:12,709][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:09:12,732][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:09:12,733][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:09:12,733][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:09:13,546][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:09:14,000][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:09:14,500][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:09:14,996][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:09:15,494][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:09:15,997][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:09:16,494][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:09:17,005][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:09:17,505][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:09:18,009][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:09:18,509][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:09:19,009][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:09:19,510][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:09:20,009][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:09:20,507][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:09:21,006][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:09:21,509][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:09:22,011][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:09:22,511][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:09:23,035][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:09:23,537][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:09:24,040][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:09:24,539][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:09:25,038][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:09:25,535][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:09:26,033][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:09:26,532][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:09:27,031][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:09:27,529][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:09:28,029][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:09:28,531][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:09:29,031][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:09:29,534][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:09:30,045][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:09:30,546][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:09:31,048][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:09:31,549][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:09:32,050][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:09:32,551][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:09:33,051][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:09:33,551][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:09:34,054][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:09:34,552][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:09:35,052][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:09:35,551][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:09:36,050][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:09:36,549][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:09:37,051][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:09:37,550][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:09:38,049][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:09:38,547][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:09:39,049][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:09:39,547][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:09:40,046][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:09:40,547][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:09:41,045][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:09:41,542][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:09:42,040][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:09:42,537][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:09:43,034][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:09:43,531][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:09:44,029][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:09:44,526][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:09:45,025][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:09:45,528][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10087 tokens. +[2026-03-26 07:09:46,397][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 60.49%, Block Peak % of device VRAM: 62.20%, ΔTime: 00:00:32 +[2026-03-26 07:09:47,157][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:09:47,159][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:09:47,161][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:09:48,023][__main__][INFO] - Iteration 643 took 58s (37.82% Gen, 60.69% Train). Generation: 21s, Training: 35s. Estimated remaining time: 38h 22m 1s. Estimated total time: 48h 26m 9s. Time estimates for 10 more iterations: 9m 41s, 100 more iterations: 1h 36m 52s, 500 more iterations: 8h 4m 21s. +[2026-03-26 07:09:48,025][__main__][INFO] - Starting iteration 643. +[2026-03-26 07:09:48,424][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 64 and human policies 1. +[2026-03-26 07:09:48,425][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:10:02,721][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 07:10:05,469][__main__][INFO] - Number of regex retries in iteration 643: 1 +[2026-03-26 07:10:05,470][__main__][INFO] - agents played in iteration 643 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:10:06,231][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:10:06,254][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:10:06,278][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:10:06,301][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:10:06,302][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:10:06,303][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:10:07,127][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:10:07,582][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:10:08,081][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:10:08,578][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:10:09,074][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:10:09,569][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:10:10,065][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:10:10,566][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:10:11,069][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:10:11,573][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:10:12,072][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:10:12,571][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:10:13,076][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:10:13,576][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:10:14,076][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:10:14,574][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:10:15,073][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:10:15,575][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:10:16,075][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:10:16,576][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:10:17,076][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:10:17,575][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:10:18,096][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:10:18,598][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:10:19,096][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:10:19,594][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:10:20,092][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:10:20,589][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:10:21,086][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:10:21,584][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:10:22,081][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:10:22,579][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:10:23,078][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:10:23,574][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:10:24,071][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:10:24,566][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:10:25,063][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:10:25,558][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:10:26,057][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:10:26,556][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:10:27,052][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:10:27,553][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:10:28,059][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:10:28,557][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:10:29,065][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:10:29,563][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:10:30,062][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:10:30,561][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:10:31,060][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:10:31,558][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:10:32,058][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:10:32,557][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:10:33,058][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:10:33,556][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:10:34,054][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:10:34,551][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:10:35,050][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:10:35,556][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:10:36,057][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:10:36,560][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:10:37,061][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:10:37,562][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:10:38,062][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:10:38,563][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:10:39,064][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10024 tokens. +[2026-03-26 07:10:39,916][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.21%, ΔTime: 00:00:32 +[2026-03-26 07:10:40,670][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:10:40,672][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:10:40,673][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:10:41,533][__main__][INFO] - Iteration 644 took 53s (32.09% Gen, 66.28% Train). Generation: 17s, Training: 35s. Estimated remaining time: 34h 10m 29s. Estimated total time: 44h 15m 30s. Time estimates for 10 more iterations: 8m 51s, 100 more iterations: 1h 28m 31s, 500 more iterations: 7h 22m 35s. +[2026-03-26 07:10:41,535][__main__][INFO] - Starting iteration 644. +[2026-03-26 07:10:41,944][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 64 and human policies 1. +[2026-03-26 07:10:41,944][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:10:57,666][__main__][INFO] - Number of regex retries in iteration 644: 0 +[2026-03-26 07:10:57,667][__main__][INFO] - agents played in iteration 644 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:10:58,445][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:10:58,470][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:10:58,494][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:10:58,518][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:10:58,519][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:10:58,519][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:10:59,336][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:10:59,792][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:11:00,301][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:11:00,801][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:11:01,303][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:11:01,811][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:11:02,311][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:11:02,813][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:11:03,312][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:11:03,814][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:11:04,315][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:11:04,816][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:11:05,317][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:11:05,818][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:11:06,329][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:11:06,842][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:11:07,344][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:11:07,844][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:11:08,346][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:11:08,848][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:11:09,349][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:11:09,853][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:11:10,352][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:11:10,852][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:11:11,353][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:11:11,853][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:11:12,362][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:11:12,873][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:11:13,377][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:11:13,876][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:11:14,377][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:11:14,875][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:11:15,374][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:11:15,876][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:11:16,373][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:11:16,870][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:11:17,371][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:11:17,868][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:11:18,369][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:11:18,869][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:11:19,371][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:11:19,880][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:11:20,381][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:11:20,880][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:11:21,383][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:11:21,889][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:11:22,389][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:11:22,888][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:11:23,388][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:11:23,887][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:11:24,387][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:11:24,887][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:11:25,386][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:11:25,889][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:11:26,386][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:11:26,884][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:11:27,383][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:11:27,879][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:11:28,375][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:11:28,873][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:11:29,372][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:11:29,870][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:11:30,369][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:11:30,869][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:11:31,369][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10216 tokens. +[2026-03-26 07:11:32,235][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 60.53%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:32 +[2026-03-26 07:11:32,983][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:11:32,985][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:11:32,987][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:11:33,723][__main__][INFO] - Iteration 645 took 51s (30.36% Gen, 68.21% Train). Generation: 15s, Training: 35s. Estimated remaining time: 33h 3m 6s. Estimated total time: 43h 8m 59s. Time estimates for 10 more iterations: 8m 37s, 100 more iterations: 1h 26m 17s, 500 more iterations: 7h 11m 29s. +[2026-03-26 07:11:33,725][__main__][INFO] - Starting iteration 645. +[2026-03-26 07:11:34,125][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 64 and human policies 1. +[2026-03-26 07:11:34,126][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:11:50,238][__main__][INFO] - Number of regex retries in iteration 645: 0 +[2026-03-26 07:11:50,239][__main__][INFO] - agents played in iteration 645 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:11:50,997][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:11:51,022][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:11:51,046][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:11:51,070][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:11:51,071][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:11:51,072][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:11:51,885][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:11:52,345][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:11:52,852][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:11:53,353][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:11:53,854][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:11:54,363][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:11:54,865][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:11:55,365][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:11:55,864][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:11:56,363][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:11:56,862][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:11:57,362][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:11:57,862][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:11:58,361][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:11:58,860][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:11:59,359][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:11:59,858][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:12:00,358][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:12:00,864][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:12:01,372][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:12:01,878][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:12:02,393][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:12:02,895][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:12:03,398][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:12:03,900][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:12:04,403][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:12:04,903][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:12:05,404][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:12:05,907][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:12:06,407][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:12:06,908][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:12:07,407][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:12:07,907][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:12:08,409][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:12:08,910][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:12:09,409][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:12:09,909][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:12:10,411][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:12:10,908][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:12:11,409][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:12:11,907][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:12:12,407][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:12:12,906][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:12:13,405][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:12:13,904][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:12:14,402][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:12:14,900][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:12:15,400][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:12:15,899][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:12:16,395][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:12:16,890][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:12:17,386][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:12:17,883][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:12:18,379][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:12:18,876][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:12:19,371][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:12:19,867][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:12:20,369][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:12:20,867][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:12:21,363][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:12:21,862][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:12:22,360][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:12:22,857][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:12:23,354][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:12:23,855][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10099 tokens. +[2026-03-26 07:12:24,706][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.25%, Current % of VRAM taken: 60.67%, Block Peak % of device VRAM: 62.24%, ΔTime: 00:00:32 +[2026-03-26 07:12:25,457][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:12:25,459][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:12:25,461][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:12:26,194][__main__][INFO] - Iteration 646 took 52s (30.94% Gen, 67.64% Train). Generation: 16s, Training: 35s. Estimated remaining time: 33h 16m 43s. Estimated total time: 43h 23m 29s. Time estimates for 10 more iterations: 8m 40s, 100 more iterations: 1h 26m 46s, 500 more iterations: 7h 13m 54s. +[2026-03-26 07:12:26,197][__main__][INFO] - Starting iteration 646. +[2026-03-26 07:12:26,610][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 64 and human policies 1. +[2026-03-26 07:12:26,611][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:12:30,629][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 07:12:40,038][__main__][INFO] - Number of regex retries in iteration 646: 1 +[2026-03-26 07:12:40,039][__main__][INFO] - agents played in iteration 646 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:12:40,798][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:12:40,822][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:12:40,845][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:12:40,868][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:12:40,869][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:12:40,870][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:12:41,688][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:12:42,141][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:12:42,646][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:12:43,146][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:12:43,650][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:12:44,149][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:12:44,649][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:12:45,148][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:12:45,647][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:12:46,146][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:12:46,643][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:12:47,146][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:12:47,646][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:12:48,152][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:12:48,668][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:12:49,168][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:12:49,669][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:12:50,170][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:12:50,672][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:12:51,173][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:12:51,674][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:12:52,178][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:12:52,680][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:12:53,181][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:12:53,680][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:12:54,179][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:12:54,679][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:12:55,178][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:12:55,677][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:12:56,176][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:12:56,676][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:12:57,177][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:12:57,675][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:12:58,173][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:12:58,673][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:12:59,172][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:12:59,671][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:13:00,170][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:13:00,668][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:13:01,168][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:13:01,666][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:13:02,172][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:13:02,673][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:13:03,177][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:13:03,680][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:13:04,184][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:13:04,688][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:13:05,190][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:13:05,692][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:13:06,196][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:13:06,698][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:13:07,200][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:13:07,702][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:13:08,206][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:13:08,708][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:13:09,214][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:13:09,716][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:13:10,214][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:13:10,712][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:13:11,211][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:13:11,710][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:13:12,209][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:13:12,707][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:13:13,205][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:13:13,707][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10165 tokens. +[2026-03-26 07:13:14,581][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 60.62%, Block Peak % of device VRAM: 62.36%, ΔTime: 00:00:32 +[2026-03-26 07:13:15,355][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:13:15,357][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:13:15,359][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:13:16,095][__main__][INFO] - Iteration 647 took 49s (27.13% Gen, 71.37% Train). Generation: 13s, Training: 35s. Estimated remaining time: 31h 6m 41s. Estimated total time: 41h 14m 17s. Time estimates for 10 more iterations: 8m 14s, 100 more iterations: 1h 22m 28s, 500 more iterations: 6h 52m 22s. +[2026-03-26 07:13:16,098][__main__][INFO] - Starting iteration 647. +[2026-03-26 07:13:16,508][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 64 and human policies 1. +[2026-03-26 07:13:16,509][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:13:44,157][__main__][INFO] - Number of regex retries in iteration 647: 0 +[2026-03-26 07:13:44,158][__main__][INFO] - agents played in iteration 647 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:13:45,147][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:13:45,171][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:13:45,194][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:13:45,218][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:13:45,218][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:13:45,219][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:13:46,109][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:13:46,752][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:13:47,257][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:13:47,766][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:13:48,273][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:13:48,778][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:13:49,278][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:13:49,781][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:13:50,285][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:13:50,784][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:13:51,284][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:13:51,785][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:13:52,284][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:13:52,782][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:13:53,280][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:13:53,779][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:13:54,278][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:13:54,781][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:13:55,278][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:13:55,776][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:13:56,274][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:13:56,776][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:13:57,273][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:13:57,772][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:13:58,274][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:13:58,770][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:13:59,268][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:13:59,764][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:14:00,261][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:14:00,758][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:14:01,255][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:14:01,754][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:14:02,252][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:14:02,754][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:14:03,253][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:14:03,751][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:14:04,250][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:14:04,749][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:14:05,254][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:14:05,752][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:14:06,251][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:14:06,750][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:14:07,251][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:14:07,749][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:14:08,248][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:14:08,746][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:14:09,247][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:14:09,746][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:14:10,246][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:14:10,744][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:14:11,241][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:14:11,743][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:14:12,240][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:14:12,739][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:14:13,236][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:14:13,734][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:14:14,233][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:14:14,734][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:14:15,233][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:14:15,733][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:14:16,232][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:14:16,729][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:14:17,230][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:14:17,731][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:14:18,227][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10218 tokens. +[2026-03-26 07:14:19,343][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.09%, Current % of VRAM taken: 60.52%, Block Peak % of device VRAM: 62.49%, ΔTime: 00:00:33 +[2026-03-26 07:14:20,232][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:14:20,234][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:14:20,235][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:14:20,960][__main__][INFO] - Iteration 648 took 1m 4s (42.90% Gen, 55.97% Train). Generation: 27s, Training: 36s. Estimated remaining time: 43h 33m 58s. Estimated total time: 53h 42m 39s. Time estimates for 10 more iterations: 10m 44s, 100 more iterations: 1h 47m 25s, 500 more iterations: 8h 57m 6s. +[2026-03-26 07:14:20,980][__main__][INFO] - Starting iteration 648. +[2026-03-26 07:14:21,387][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 64 and human policies 1. +[2026-03-26 07:14:21,388][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:15:12,170][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 07:15:18,809][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 07:15:20,841][__main__][INFO] - Number of regex retries in iteration 648: 2 +[2026-03-26 07:15:20,842][__main__][INFO] - agents played in iteration 648 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:15:21,619][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:15:21,642][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:15:21,665][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:15:21,689][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:15:21,689][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:15:21,690][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:15:22,511][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:15:23,728][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:15:24,229][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:15:24,728][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:15:25,228][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:15:25,725][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:15:26,224][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:15:26,723][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:15:27,221][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:15:27,718][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:15:28,215][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:15:28,712][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:15:29,208][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:15:29,705][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:15:30,202][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:15:30,700][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:15:31,198][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:15:31,697][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:15:32,200][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:15:32,697][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:15:33,196][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:15:33,694][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:15:34,196][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:15:34,695][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:15:35,192][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:15:35,690][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:15:36,192][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:15:36,690][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:15:37,188][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:15:37,686][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:15:38,183][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:15:38,681][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:15:39,178][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:15:39,673][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:15:40,169][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:15:40,669][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:15:41,169][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:15:41,668][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:15:42,169][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:15:42,665][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:15:43,162][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:15:43,660][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:15:44,157][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:15:44,658][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:15:45,155][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:15:45,658][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:15:46,157][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:15:46,656][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:15:47,166][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:15:47,667][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:15:48,165][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:15:48,662][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:15:49,161][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:15:49,658][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:15:50,154][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:15:50,651][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:15:51,149][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:15:51,649][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:15:52,149][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:15:52,648][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:15:53,146][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:15:53,645][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:15:54,143][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:15:54,640][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:15:55,137][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10072 tokens. +[2026-03-26 07:15:56,004][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 60.50%, Block Peak % of device VRAM: 62.18%, ΔTime: 00:00:33 +[2026-03-26 07:15:56,738][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:15:56,740][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:15:56,741][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:15:57,435][__main__][INFO] - Iteration 649 took 1m 36s (61.90% Gen, 37.38% Train). Generation: 59s, Training: 35s. Estimated remaining time: 69h 52m 7s. Estimated total time: 80h 2m 24s. Time estimates for 10 more iterations: 16m 0s, 100 more iterations: 2h 40m 4s, 500 more iterations: 13h 20m 24s. +[2026-03-26 07:15:57,437][__main__][INFO] - Starting iteration 649. +[2026-03-26 07:15:57,849][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 64 and human policies 1. +[2026-03-26 07:15:57,850][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:16:52,775][__main__][INFO] - Number of regex retries in iteration 649: 0 +[2026-03-26 07:16:52,776][__main__][INFO] - agents played in iteration 649 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:16:53,536][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:16:53,560][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:16:53,583][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:16:53,607][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:16:53,607][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:16:53,608][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:16:54,418][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:16:54,874][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:16:55,382][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:16:55,884][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:16:56,395][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:16:56,895][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:16:57,400][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:16:57,902][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:16:58,402][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:16:58,902][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:16:59,406][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:16:59,915][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:17:00,417][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:17:00,922][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:17:01,421][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:17:01,921][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:17:02,428][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:17:02,927][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:17:03,425][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:17:03,927][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:17:04,425][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:17:04,931][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:17:05,430][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:17:05,930][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:17:06,432][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:17:06,937][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:17:07,449][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:17:07,950][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:17:08,451][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:17:08,954][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:17:09,454][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:17:09,954][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:17:10,453][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:17:10,953][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:17:11,452][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:17:11,954][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:17:12,467][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:17:12,967][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:17:13,467][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:17:13,969][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:17:14,469][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:17:14,970][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:17:15,472][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:17:15,972][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:17:16,472][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:17:16,973][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:17:17,471][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:17:17,972][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:17:18,479][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:17:18,987][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:17:19,491][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:17:19,990][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:17:20,490][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:17:20,989][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:17:21,490][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:17:21,993][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:17:22,490][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:17:22,989][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:17:23,488][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:17:23,986][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:17:24,486][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:17:24,988][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:17:25,491][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:17:25,990][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:17:26,497][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10149 tokens. +[2026-03-26 07:17:27,361][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.21%, ΔTime: 00:00:32 +[2026-03-26 07:17:28,116][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:17:28,118][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:17:28,120][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:17:28,844][__main__][INFO] - Iteration 650 took 1m 30s (60.36% Gen, 38.84% Train). Generation: 54s, Training: 35s. Estimated remaining time: 65h 37m 58s. Estimated total time: 75h 49m 47s. Time estimates for 10 more iterations: 15m 9s, 100 more iterations: 2h 31m 39s, 500 more iterations: 12h 38m 17s. +[2026-03-26 07:17:28,847][__main__][INFO] - Starting iteration 650. +[2026-03-26 07:17:29,252][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 64 and human policies 1. +[2026-03-26 07:17:29,253][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:18:01,035][__main__][INFO] - Number of regex retries in iteration 650: 0 +[2026-03-26 07:18:01,036][__main__][INFO] - agents played in iteration 650 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:18:01,808][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:18:01,831][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:18:01,855][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:18:01,878][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:18:01,879][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:18:01,879][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:18:02,694][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:18:03,157][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:18:03,663][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:18:04,179][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:18:04,683][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:18:05,186][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:18:05,689][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:18:06,195][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:18:06,702][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:18:07,206][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:18:07,710][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:18:08,209][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:18:08,711][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:18:09,211][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:18:09,709][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:18:10,208][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:18:10,705][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:18:11,208][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:18:11,711][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:18:12,213][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:18:12,714][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:18:13,215][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:18:13,715][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:18:14,215][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:18:14,715][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:18:15,214][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:18:15,715][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:18:16,215][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:18:16,716][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:18:17,214][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:18:17,713][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:18:18,213][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:18:18,713][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:18:19,210][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:18:19,707][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:18:20,205][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:18:20,705][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:18:21,204][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:18:21,702][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:18:22,199][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:18:22,698][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:18:23,194][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:18:23,692][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:18:24,189][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:18:24,687][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:18:25,185][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:18:25,684][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:18:26,182][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:18:26,683][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:18:27,182][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:18:27,682][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:18:28,178][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:18:28,673][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:18:29,170][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:18:29,667][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:18:30,164][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:18:30,661][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:18:31,158][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:18:31,656][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:18:32,155][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:18:32,654][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:18:33,152][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:18:33,650][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:18:34,148][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:18:34,646][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10146 tokens. +[2026-03-26 07:18:35,499][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 60.52%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 +[2026-03-26 07:18:36,251][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:18:36,253][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:18:36,255][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:18:37,897][__main__][INFO] - Iteration 651 took 1m 8s (46.30% Gen, 51.30% Train). Generation: 31s, Training: 35s. Estimated remaining time: 46h 59m 18s. Estimated total time: 57h 12m 16s. Time estimates for 10 more iterations: 11m 26s, 100 more iterations: 1h 54m 24s, 500 more iterations: 9h 32m 2s. +[2026-03-26 07:18:37,899][__main__][INFO] - Starting iteration 651. +[2026-03-26 07:18:38,312][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 65 and human policies 1. +[2026-03-26 07:18:38,313][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:18:46,692][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 07:18:56,539][__main__][INFO] - Number of regex retries in iteration 651: 1 +[2026-03-26 07:18:56,540][__main__][INFO] - agents played in iteration 651 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:18:57,315][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:18:57,341][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:18:57,367][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:18:57,393][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:18:57,393][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:18:57,394][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:18:58,221][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:18:58,680][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:18:59,190][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:18:59,691][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:19:00,192][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:19:00,692][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:19:01,193][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:19:01,693][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:19:02,196][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:19:02,695][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:19:03,198][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:19:03,701][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:19:04,220][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:19:04,721][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:19:05,222][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:19:05,723][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:19:06,227][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:19:06,726][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:19:07,227][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:19:07,728][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:19:08,232][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:19:08,734][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:19:09,232][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:19:09,732][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:19:10,234][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:19:10,735][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:19:11,241][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:19:11,743][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:19:12,243][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:19:12,746][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:19:13,248][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:19:13,768][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:19:14,272][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:19:14,770][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:19:15,272][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:19:15,769][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:19:16,268][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:19:16,769][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:19:17,267][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:19:17,766][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:19:18,267][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:19:18,765][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:19:19,263][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:19:19,763][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:19:20,265][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:19:20,762][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:19:21,261][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:19:21,762][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:19:22,258][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:19:22,753][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:19:23,250][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:19:23,747][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:19:24,244][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:19:24,740][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:19:25,235][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:19:25,732][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:19:26,228][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:19:26,726][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:19:27,227][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:19:27,724][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:19:28,220][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:19:28,717][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:19:29,213][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:19:29,711][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:19:30,208][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10157 tokens. +[2026-03-26 07:19:31,062][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 60.54%, Block Peak % of device VRAM: 62.37%, ΔTime: 00:00:32 +[2026-03-26 07:19:31,826][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:19:31,828][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:19:31,829][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:19:32,638][__main__][INFO] - Iteration 652 took 54s (33.55% Gen, 64.96% Train). Generation: 18s, Training: 35s. Estimated remaining time: 35h 2m 29s. Estimated total time: 45h 16m 21s. Time estimates for 10 more iterations: 9m 3s, 100 more iterations: 1h 30m 32s, 500 more iterations: 7h 32m 43s. +[2026-03-26 07:19:32,640][__main__][INFO] - Starting iteration 652. +[2026-03-26 07:19:33,046][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 65 and human policies 1. +[2026-03-26 07:19:33,046][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:19:54,726][__main__][INFO] - Number of regex retries in iteration 652: 0 +[2026-03-26 07:19:54,727][__main__][INFO] - agents played in iteration 652 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:19:55,497][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:19:55,521][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:19:55,544][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:19:55,567][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:19:55,568][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:19:55,568][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:19:56,387][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:19:56,840][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:19:57,345][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:19:57,845][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:19:58,343][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:19:58,847][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:19:59,351][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:19:59,851][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:20:00,362][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:20:00,865][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:20:01,369][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:20:01,873][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:20:02,375][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:20:02,875][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:20:03,383][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:20:03,882][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:20:04,381][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:20:04,880][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:20:05,380][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:20:05,878][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:20:06,378][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:20:06,876][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:20:07,376][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:20:07,875][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:20:08,374][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:20:08,877][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:20:09,399][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:20:09,905][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:20:10,406][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:20:10,909][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:20:11,413][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:20:11,919][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:20:12,423][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:20:12,932][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:20:13,434][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:20:13,934][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:20:14,434][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:20:14,934][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:20:15,435][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:20:15,934][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:20:16,438][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:20:16,937][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:20:17,440][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:20:17,939][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:20:18,438][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:20:18,938][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:20:19,440][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:20:19,938][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:20:20,436][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:20:20,933][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:20:21,435][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:20:21,933][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:20:22,433][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:20:22,931][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:20:23,429][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:20:23,927][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:20:24,425][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:20:24,923][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:20:25,421][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:20:25,920][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:20:26,418][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:20:26,916][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:20:27,414][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:20:27,913][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:20:28,413][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10144 tokens. +[2026-03-26 07:20:29,271][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 60.66%, Block Peak % of device VRAM: 62.27%, ΔTime: 00:00:32 +[2026-03-26 07:20:30,025][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:20:30,027][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:20:30,029][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:20:30,782][__main__][INFO] - Iteration 653 took 57s (37.55% Gen, 61.14% Train). Generation: 21s, Training: 35s. Estimated remaining time: 37h 52m 2s. Estimated total time: 48h 6m 52s. Time estimates for 10 more iterations: 9m 37s, 100 more iterations: 1h 36m 13s, 500 more iterations: 8h 1m 8s. +[2026-03-26 07:20:30,784][__main__][INFO] - Starting iteration 653. +[2026-03-26 07:20:31,183][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 65 and human policies 1. +[2026-03-26 07:20:31,184][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:20:39,508][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 07:20:45,523][__main__][INFO] - Number of regex retries in iteration 653: 1 +[2026-03-26 07:20:45,524][__main__][INFO] - agents played in iteration 653 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:20:46,296][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:20:46,320][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:20:46,343][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:20:46,367][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:20:46,367][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:20:46,368][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:20:47,194][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:20:47,647][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:20:48,150][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:20:48,649][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:20:49,148][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:20:49,646][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:20:50,144][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:20:50,640][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:20:51,145][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:20:51,646][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:20:52,147][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:20:52,647][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:20:53,148][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:20:53,647][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:20:54,148][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:20:54,651][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:20:55,152][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:20:55,657][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:20:56,156][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:20:56,658][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:20:57,168][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:20:57,667][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:20:58,167][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:20:58,668][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:20:59,168][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:20:59,668][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:21:00,167][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:21:00,665][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:21:01,163][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:21:01,662][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:21:02,166][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:21:02,664][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:21:03,163][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:21:03,660][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:21:04,160][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:21:04,658][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:21:05,155][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:21:05,652][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:21:06,150][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:21:06,650][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:21:07,148][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:21:07,649][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:21:08,149][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:21:08,653][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:21:09,153][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:21:09,653][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:21:10,153][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:21:10,653][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:21:11,155][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:21:11,651][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:21:12,148][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:21:12,645][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:21:13,141][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:21:13,638][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:21:14,134][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:21:14,631][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:21:15,128][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:21:15,625][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:21:16,122][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:21:16,619][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:21:17,118][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:21:17,618][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:21:18,116][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:21:18,615][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:21:19,115][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10075 tokens. +[2026-03-26 07:21:19,981][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 +[2026-03-26 07:21:20,730][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:21:20,732][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:21:20,733][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:21:21,462][__main__][INFO] - Iteration 654 took 50s (28.52% Gen, 70.03% Train). Generation: 14s, Training: 35s. Estimated remaining time: 31h 38m 16s. Estimated total time: 41h 53m 57s. Time estimates for 10 more iterations: 8m 22s, 100 more iterations: 1h 23m 47s, 500 more iterations: 6h 58m 59s. +[2026-03-26 07:21:21,464][__main__][INFO] - Starting iteration 654. +[2026-03-26 07:21:21,873][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 65 and human policies 1. +[2026-03-26 07:21:21,874][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:21:38,058][__main__][INFO] - Number of regex retries in iteration 654: 0 +[2026-03-26 07:21:38,059][__main__][INFO] - agents played in iteration 654 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:21:38,838][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:21:38,861][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:21:38,885][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:21:38,909][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:21:38,909][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:21:38,910][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:21:39,726][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:21:40,184][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:21:40,686][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:21:41,186][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:21:41,690][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:21:42,188][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:21:42,686][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:21:43,186][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:21:43,685][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:21:44,184][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:21:44,681][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:21:45,178][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:21:45,682][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:21:46,199][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:21:46,697][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:21:47,199][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:21:47,696][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:21:48,193][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:21:48,694][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:21:49,191][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:21:49,690][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:21:50,190][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:21:50,689][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:21:51,188][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:21:51,689][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:21:52,190][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:21:52,687][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:21:53,184][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:21:53,679][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:21:54,176][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:21:54,674][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:21:55,172][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:21:55,672][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:21:56,170][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:21:56,668][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:21:57,164][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:21:57,661][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:21:58,160][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:21:58,657][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:21:59,156][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:21:59,654][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:22:00,156][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:22:00,656][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:22:01,156][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:22:01,657][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:22:02,156][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:22:02,656][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:22:03,157][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:22:03,657][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:22:04,161][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:22:04,662][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:22:05,162][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:22:05,663][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:22:06,170][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:22:06,672][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:22:07,174][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:22:07,687][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:22:08,190][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:22:08,692][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:22:09,191][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:22:09,693][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:22:10,191][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:22:10,691][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:22:11,193][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:22:11,692][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10289 tokens. +[2026-03-26 07:22:12,571][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.57%, Block Peak % of device VRAM: 62.23%, ΔTime: 00:00:32 +[2026-03-26 07:22:13,330][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:22:13,333][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:22:13,334][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:22:14,055][__main__][INFO] - Iteration 655 took 52s (31.02% Gen, 67.60% Train). Generation: 16s, Training: 35s. Estimated remaining time: 33h 12m 34s. Estimated total time: 43h 29m 7s. Time estimates for 10 more iterations: 8m 41s, 100 more iterations: 1h 26m 58s, 500 more iterations: 7h 14m 51s. +[2026-03-26 07:22:14,057][__main__][INFO] - Starting iteration 655. +[2026-03-26 07:22:14,467][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 65 and human policies 1. +[2026-03-26 07:22:14,468][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:22:27,673][__main__][INFO] - Number of regex retries in iteration 655: 0 +[2026-03-26 07:22:27,674][__main__][INFO] - agents played in iteration 655 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:22:28,434][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:22:28,458][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:22:28,482][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:22:28,505][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:22:28,506][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:22:28,506][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:22:29,319][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:22:29,778][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:22:30,281][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:22:30,782][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:22:31,299][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:22:31,805][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:22:32,310][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:22:32,811][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:22:33,313][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:22:33,812][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:22:34,314][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:22:34,826][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:22:35,332][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:22:35,832][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:22:36,334][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:22:36,834][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:22:37,334][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:22:37,833][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:22:38,334][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:22:38,833][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:22:39,333][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:22:39,834][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:22:40,333][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:22:40,832][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:22:41,331][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:22:41,831][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:22:42,330][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:22:42,835][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:22:43,334][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:22:43,837][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:22:44,335][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:22:44,832][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:22:45,331][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:22:45,831][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:22:46,331][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:22:46,829][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:22:47,328][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:22:47,827][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:22:48,327][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:22:48,824][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:22:49,322][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:22:49,819][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:22:50,316][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:22:50,816][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:22:51,314][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:22:51,837][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:22:52,335][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:22:52,833][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:22:53,332][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:22:53,829][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:22:54,335][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:22:54,832][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:22:55,331][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:22:55,830][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:22:56,328][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:22:56,826][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:22:57,326][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:22:57,823][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:22:58,318][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:22:58,814][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:22:59,311][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:22:59,807][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:23:00,303][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:23:00,799][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:23:01,297][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9990 tokens. +[2026-03-26 07:23:02,159][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.05%, Current % of VRAM taken: 60.48%, Block Peak % of device VRAM: 62.10%, ΔTime: 00:00:32 +[2026-03-26 07:23:02,918][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:23:02,920][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:23:02,922][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:23:03,647][__main__][INFO] - Iteration 656 took 49s (26.85% Gen, 71.67% Train). Generation: 13s, Training: 35s. Estimated remaining time: 30h 41m 39s. Estimated total time: 40h 59m 2s. Time estimates for 10 more iterations: 8m 11s, 100 more iterations: 1h 21m 58s, 500 more iterations: 6h 49m 50s. +[2026-03-26 07:23:03,650][__main__][INFO] - Starting iteration 656. +[2026-03-26 07:23:04,051][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 65 and human policies 1. +[2026-03-26 07:23:04,052][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:23:20,563][__main__][INFO] - Number of regex retries in iteration 656: 0 +[2026-03-26 07:23:20,564][__main__][INFO] - agents played in iteration 656 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:23:21,328][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:23:21,352][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:23:21,376][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:23:21,400][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:23:21,400][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:23:21,401][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:23:22,229][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:23:22,682][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:23:23,191][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:23:23,704][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:23:24,204][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:23:24,704][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:23:25,207][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:23:25,706][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:23:26,207][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:23:26,708][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:23:27,209][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:23:27,710][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:23:28,210][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:23:28,710][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:23:29,214][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:23:29,713][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:23:30,213][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:23:30,712][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:23:31,214][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:23:31,720][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:23:32,224][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:23:32,736][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:23:33,236][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:23:33,736][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:23:34,237][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:23:34,739][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:23:35,241][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:23:35,742][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:23:36,247][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:23:36,748][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:23:37,250][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:23:37,751][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:23:38,270][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:23:38,776][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:23:39,276][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:23:39,776][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:23:40,276][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:23:40,779][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:23:41,283][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:23:41,782][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:23:42,281][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:23:42,780][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:23:43,282][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:23:43,781][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:23:44,279][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:23:44,781][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:23:45,278][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:23:45,780][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:23:46,277][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:23:46,776][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:23:47,274][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:23:47,773][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:23:48,274][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:23:48,771][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:23:49,270][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:23:49,768][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:23:50,270][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:23:50,768][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:23:51,266][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:23:51,764][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:23:52,262][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:23:52,760][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:23:53,261][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:23:53,758][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:23:54,258][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10143 tokens. +[2026-03-26 07:23:55,122][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 60.57%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:32 +[2026-03-26 07:23:55,872][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:23:55,875][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:23:55,876][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:23:56,598][__main__][INFO] - Iteration 657 took 52s (31.42% Gen, 67.20% Train). Generation: 16s, Training: 35s. Estimated remaining time: 33h 29m 6s. Estimated total time: 43h 47m 23s. Time estimates for 10 more iterations: 8m 45s, 100 more iterations: 1h 27m 34s, 500 more iterations: 7h 17m 53s. +[2026-03-26 07:23:56,600][__main__][INFO] - Starting iteration 657. +[2026-03-26 07:23:57,003][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 65 and human policies 1. +[2026-03-26 07:23:57,004][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:24:10,095][__main__][INFO] - Number of regex retries in iteration 657: 0 +[2026-03-26 07:24:10,095][__main__][INFO] - agents played in iteration 657 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:24:10,854][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:24:10,878][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:24:10,901][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:24:10,924][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:24:10,925][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:24:10,926][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:24:11,736][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:24:12,193][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:24:12,696][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:24:13,193][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:24:13,691][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:24:14,196][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:24:14,706][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:24:15,207][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:24:15,707][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:24:16,209][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:24:16,710][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:24:17,211][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:24:17,710][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:24:18,214][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:24:18,717][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:24:19,216][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:24:19,716][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:24:20,215][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:24:20,721][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:24:21,230][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:24:21,730][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:24:22,230][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:24:22,730][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:24:23,231][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:24:23,731][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:24:24,240][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:24:24,754][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:24:25,256][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:24:25,757][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:24:26,258][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:24:26,758][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:24:27,258][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:24:27,758][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:24:28,258][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:24:28,760][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:24:29,275][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:24:29,777][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:24:30,278][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:24:30,779][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:24:31,279][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:24:31,782][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:24:32,283][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:24:32,782][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:24:33,281][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:24:33,780][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:24:34,284][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:24:34,795][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:24:35,298][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:24:35,798][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:24:36,300][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:24:36,800][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:24:37,299][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:24:37,800][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:24:38,299][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:24:38,798][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:24:39,298][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:24:39,797][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:24:40,296][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:24:40,796][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:24:41,296][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:24:41,795][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:24:42,294][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:24:42,793][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:24:43,292][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:24:43,791][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10004 tokens. +[2026-03-26 07:24:44,656][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 60.52%, Block Peak % of device VRAM: 61.95%, ΔTime: 00:00:32 +[2026-03-26 07:24:45,405][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:24:45,407][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:24:45,409][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:24:46,132][__main__][INFO] - Iteration 658 took 49s (26.65% Gen, 71.88% Train). Generation: 13s, Training: 35s. Estimated remaining time: 30h 37m 24s. Estimated total time: 40h 56m 29s. Time estimates for 10 more iterations: 8m 11s, 100 more iterations: 1h 21m 52s, 500 more iterations: 6h 49m 24s. +[2026-03-26 07:24:46,136][__main__][INFO] - Starting iteration 658. +[2026-03-26 07:24:46,547][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 65 and human policies 1. +[2026-03-26 07:24:46,547][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:24:50,875][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 07:25:01,408][__main__][INFO] - Number of regex retries in iteration 658: 1 +[2026-03-26 07:25:01,409][__main__][INFO] - agents played in iteration 658 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:25:02,190][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:25:02,214][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:25:02,237][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:25:02,261][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:25:02,261][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:25:02,262][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:25:03,075][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:25:03,539][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:25:04,058][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:25:04,561][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:25:05,065][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:25:05,565][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:25:06,069][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:25:06,574][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:25:07,077][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:25:07,584][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:25:08,086][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:25:08,589][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:25:09,088][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:25:09,588][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:25:10,087][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:25:10,587][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:25:11,085][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:25:11,584][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:25:12,085][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:25:12,587][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:25:13,095][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:25:13,599][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:25:14,099][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:25:14,600][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:25:15,100][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:25:15,600][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:25:16,100][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:25:16,598][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:25:17,096][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:25:17,592][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:25:18,091][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:25:18,591][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:25:19,090][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:25:19,589][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:25:20,088][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:25:20,590][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:25:21,087][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:25:21,585][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:25:22,089][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:25:22,588][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:25:23,087][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:25:23,587][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:25:24,084][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:25:24,581][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:25:25,078][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:25:25,575][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:25:26,072][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:25:26,569][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:25:27,069][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:25:27,570][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:25:28,070][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:25:28,567][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:25:29,071][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:25:29,570][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:25:30,072][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:25:30,569][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:25:31,069][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:25:31,567][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:25:32,066][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:25:32,564][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:25:33,064][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:25:33,561][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:25:34,059][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:25:34,560][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:25:35,057][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10193 tokens. +[2026-03-26 07:25:35,918][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 60.54%, Block Peak % of device VRAM: 62.11%, ΔTime: 00:00:32 +[2026-03-26 07:25:36,665][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:25:36,667][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:25:36,669][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:25:37,396][__main__][INFO] - Iteration 659 took 50s (29.23% Gen, 69.34% Train). Generation: 14s, Training: 35s. Estimated remaining time: 32h 2m 32s. Estimated total time: 42h 22m 29s. Time estimates for 10 more iterations: 8m 28s, 100 more iterations: 1h 24m 44s, 500 more iterations: 7h 3m 44s. +[2026-03-26 07:25:37,398][__main__][INFO] - Starting iteration 659. +[2026-03-26 07:25:37,822][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 65 and human policies 1. +[2026-03-26 07:25:37,822][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:25:42,529][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 07:25:52,380][__main__][INFO] - Number of regex retries in iteration 659: 1 +[2026-03-26 07:25:52,381][__main__][INFO] - agents played in iteration 659 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:25:53,154][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:25:53,178][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:25:53,201][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:25:53,225][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:25:53,225][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:25:53,226][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:25:54,046][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:25:54,500][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:25:55,001][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:25:55,498][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:25:55,995][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:25:56,491][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:25:56,987][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:25:57,484][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:25:57,980][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:25:58,481][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:25:58,978][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:25:59,474][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:25:59,975][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:26:00,475][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:26:00,980][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:26:01,482][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:26:02,003][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:26:02,508][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:26:03,008][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:26:03,508][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:26:04,007][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:26:04,509][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:26:05,009][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:26:05,508][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:26:06,008][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:26:06,506][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:26:07,004][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:26:07,504][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:26:08,002][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:26:08,499][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:26:08,997][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:26:09,495][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:26:09,994][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:26:10,491][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:26:10,988][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:26:11,486][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:26:11,981][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:26:12,479][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:26:12,976][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:26:13,476][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:26:13,973][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:26:14,470][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:26:14,967][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:26:15,469][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:26:15,967][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:26:16,469][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:26:16,969][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:26:17,475][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:26:17,978][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:26:18,478][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:26:18,979][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:26:19,482][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:26:19,983][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:26:20,485][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:26:20,985][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:26:21,485][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:26:21,986][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:26:22,488][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:26:22,990][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:26:23,491][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:26:23,993][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:26:24,494][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:26:24,995][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:26:25,496][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:26:25,997][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10080 tokens. +[2026-03-26 07:26:26,875][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 60.64%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 +[2026-03-26 07:26:27,621][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:26:27,624][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:26:27,625][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:26:28,350][__main__][INFO] - Iteration 660 took 50s (28.81% Gen, 69.75% Train). Generation: 14s, Training: 35s. Estimated remaining time: 31h 45m 37s. Estimated total time: 42h 6m 25s. Time estimates for 10 more iterations: 8m 25s, 100 more iterations: 1h 24m 12s, 500 more iterations: 7h 1m 4s. +[2026-03-26 07:26:28,352][__main__][INFO] - Starting iteration 660. +[2026-03-26 07:26:28,764][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 65 and human policies 1. +[2026-03-26 07:26:28,764][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:26:44,487][__main__][INFO] - Number of regex retries in iteration 660: 0 +[2026-03-26 07:26:44,488][__main__][INFO] - agents played in iteration 660 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:26:45,270][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:26:45,294][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:26:45,317][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:26:45,340][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:26:45,341][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:26:45,342][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:26:46,163][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:26:46,626][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:26:47,147][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:26:47,651][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:26:48,157][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:26:48,663][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:26:49,166][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:26:49,668][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:26:50,185][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:26:50,690][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:26:51,191][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:26:51,691][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:26:52,192][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:26:52,694][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:26:53,193][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:26:53,691][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:26:54,191][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:26:54,690][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:26:55,189][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:26:55,689][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:26:56,188][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:26:56,688][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:26:57,186][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:26:57,687][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:26:58,184][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:26:58,681][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:26:59,178][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:26:59,675][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:27:00,177][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:27:00,677][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:27:01,194][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:27:01,695][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:27:02,198][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:27:02,701][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:27:03,203][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:27:03,705][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:27:04,208][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:27:04,709][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:27:05,211][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:27:05,710][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:27:06,214][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:27:06,716][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:27:07,216][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:27:07,725][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:27:08,237][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:27:08,741][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:27:09,243][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:27:09,748][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:27:10,252][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:27:10,754][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:27:11,256][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:27:11,754][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:27:12,257][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:27:12,758][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:27:13,267][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:27:13,766][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:27:14,270][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:27:14,769][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:27:15,267][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:27:15,767][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:27:16,266][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:27:16,764][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:27:17,263][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:27:17,763][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:27:18,277][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10216 tokens. +[2026-03-26 07:27:19,169][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 60.53%, Block Peak % of device VRAM: 62.27%, ΔTime: 00:00:33 +[2026-03-26 07:27:19,916][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:27:19,920][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:27:19,921][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:27:21,225][__main__][INFO] - Iteration 661 took 52s (29.97% Gen, 67.54% Train). Generation: 15s, Training: 35s. Estimated remaining time: 33h 21m 24s. Estimated total time: 43h 43m 5s. Time estimates for 10 more iterations: 8m 44s, 100 more iterations: 1h 27m 26s, 500 more iterations: 7h 17m 10s. +[2026-03-26 07:27:21,227][__main__][INFO] - Starting iteration 661. +[2026-03-26 07:27:21,640][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 66 and human policies 1. +[2026-03-26 07:27:21,641][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:27:35,304][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 07:27:36,198][__main__][INFO] - Number of regex retries in iteration 661: 1 +[2026-03-26 07:27:36,199][__main__][INFO] - agents played in iteration 661 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:27:36,970][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:27:36,994][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:27:37,017][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:27:37,041][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:27:37,041][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:27:37,042][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:27:37,859][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:27:38,314][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:27:38,826][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:27:39,328][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:27:39,835][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:27:40,344][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:27:40,846][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:27:41,350][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:27:41,854][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:27:42,359][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:27:42,863][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:27:43,367][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:27:43,870][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:27:44,372][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:27:44,875][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:27:45,379][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:27:45,883][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:27:46,382][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:27:46,884][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:27:47,381][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:27:47,880][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:27:48,384][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:27:48,887][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:27:49,388][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:27:49,895][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:27:50,397][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:27:50,900][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:27:51,399][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:27:51,906][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:27:52,406][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:27:52,905][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:27:53,410][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:27:53,910][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:27:54,410][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:27:54,913][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:27:55,412][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:27:55,911][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:27:56,410][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:27:56,909][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:27:57,408][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:27:57,908][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:27:58,411][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:27:58,910][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:27:59,410][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:27:59,910][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:28:00,409][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:28:00,908][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:28:01,410][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:28:01,918][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:28:02,419][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:28:02,933][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:28:03,434][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:28:03,936][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:28:04,441][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:28:04,940][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:28:05,442][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:28:05,945][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:28:06,444][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:28:06,944][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:28:07,443][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:28:07,941][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:28:08,439][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:28:08,937][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:28:09,436][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:28:09,936][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10143 tokens. +[2026-03-26 07:28:10,812][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.40%, ΔTime: 00:00:32 +[2026-03-26 07:28:11,561][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:28:11,563][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:28:11,565][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:28:12,419][__main__][INFO] - Iteration 662 took 50s (28.67% Gen, 69.65% Train). Generation: 14s, Training: 35s. Estimated remaining time: 31h 56m 26s. Estimated total time: 42h 18m 58s. Time estimates for 10 more iterations: 8m 27s, 100 more iterations: 1h 24m 37s, 500 more iterations: 7h 3m 9s. +[2026-03-26 07:28:12,421][__main__][INFO] - Starting iteration 662. +[2026-03-26 07:28:12,824][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 66 and human policies 1. +[2026-03-26 07:28:12,825][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:28:27,918][__main__][INFO] - Number of regex retries in iteration 662: 0 +[2026-03-26 07:28:27,919][__main__][INFO] - agents played in iteration 662 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:28:28,688][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:28:28,713][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:28:28,738][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:28:28,762][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:28:28,763][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:28:28,763][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:28:29,590][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:28:30,050][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:28:30,555][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:28:31,056][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:28:31,556][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:28:32,063][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:28:32,577][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:28:33,079][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:28:33,582][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:28:34,081][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:28:34,581][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:28:35,080][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:28:35,579][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:28:36,079][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:28:36,579][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:28:37,078][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:28:37,577][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:28:38,076][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:28:38,574][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:28:39,074][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:28:39,573][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:28:40,078][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:28:40,590][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:28:41,090][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:28:41,592][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:28:42,092][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:28:42,592][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:28:43,092][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:28:43,591][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:28:44,090][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:28:44,589][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:28:45,090][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:28:45,590][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:28:46,095][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:28:46,597][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:28:47,106][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:28:47,605][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:28:48,106][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:28:48,606][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:28:49,104][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:28:49,608][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:28:50,107][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:28:50,609][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:28:51,107][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:28:51,604][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:28:52,102][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:28:52,603][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:28:53,103][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:28:53,614][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:28:54,116][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:28:54,618][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:28:55,123][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:28:55,626][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:28:56,130][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:28:56,631][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:28:57,134][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:28:57,637][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:28:58,138][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:28:58,637][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:28:59,137][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:28:59,637][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:29:00,139][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:29:00,645][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:29:01,148][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:29:01,657][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10141 tokens. +[2026-03-26 07:29:02,525][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:32 +[2026-03-26 07:29:03,294][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:29:03,296][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:29:03,298][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:29:04,028][__main__][INFO] - Iteration 663 took 51s (29.48% Gen, 69.09% Train). Generation: 15s, Training: 35s. Estimated remaining time: 32h 16m 49s. Estimated total time: 42h 40m 13s. Time estimates for 10 more iterations: 8m 32s, 100 more iterations: 1h 25m 20s, 500 more iterations: 7h 6m 42s. +[2026-03-26 07:29:04,030][__main__][INFO] - Starting iteration 663. +[2026-03-26 07:29:04,428][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 66 and human policies 1. +[2026-03-26 07:29:04,429][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:29:15,789][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 07:29:18,458][__main__][INFO] - Number of regex retries in iteration 663: 1 +[2026-03-26 07:29:18,459][__main__][INFO] - agents played in iteration 663 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:29:19,223][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:29:19,246][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:29:19,270][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:29:19,294][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:29:19,294][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:29:19,295][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:29:20,126][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:29:20,583][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:29:21,085][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:29:21,582][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:29:22,078][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:29:22,575][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:29:23,072][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:29:23,568][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:29:24,068][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:29:24,569][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:29:25,068][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:29:25,571][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:29:26,069][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:29:26,573][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:29:27,078][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:29:27,577][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:29:28,088][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:29:28,595][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:29:29,119][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:29:29,619][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:29:30,123][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:29:30,628][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:29:31,128][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:29:31,631][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:29:32,132][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:29:32,635][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:29:33,139][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:29:33,641][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:29:34,141][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:29:34,644][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:29:35,146][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:29:35,648][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:29:36,149][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:29:36,650][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:29:37,152][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:29:37,653][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:29:38,155][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:29:38,657][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:29:39,158][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:29:39,659][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:29:40,160][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:29:40,661][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:29:41,160][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:29:41,664][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:29:42,165][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:29:42,673][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:29:43,175][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:29:43,679][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:29:44,179][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:29:44,682][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:29:45,183][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:29:45,683][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:29:46,186][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:29:46,688][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:29:47,189][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:29:47,691][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:29:48,189][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:29:48,687][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:29:49,185][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:29:49,685][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:29:50,184][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:29:50,683][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:29:51,181][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:29:51,679][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:29:52,177][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10120 tokens. +[2026-03-26 07:29:53,051][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 60.53%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:32 +[2026-03-26 07:29:53,803][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:29:53,806][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:29:53,807][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:29:54,636][__main__][INFO] - Iteration 664 took 50s (27.94% Gen, 70.40% Train). Generation: 14s, Training: 35s. Estimated remaining time: 31h 26m 12s. Estimated total time: 41h 50m 26s. Time estimates for 10 more iterations: 8m 22s, 100 more iterations: 1h 23m 40s, 500 more iterations: 6h 58m 24s. +[2026-03-26 07:29:54,638][__main__][INFO] - Starting iteration 664. +[2026-03-26 07:29:55,047][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 66 and human policies 1. +[2026-03-26 07:29:55,048][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:30:09,513][__main__][INFO] - Number of regex retries in iteration 664: 0 +[2026-03-26 07:30:09,514][__main__][INFO] - agents played in iteration 664 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:30:10,292][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:30:10,316][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:30:10,340][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:30:10,363][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:30:10,364][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:30:10,365][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:30:11,191][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:30:11,652][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:30:12,155][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:30:12,653][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:30:13,152][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:30:13,653][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:30:14,155][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:30:14,653][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:30:15,152][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:30:15,653][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:30:16,154][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:30:16,655][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:30:17,156][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:30:17,655][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:30:18,161][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:30:18,662][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:30:19,162][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:30:19,659][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:30:20,159][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:30:20,658][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:30:21,156][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:30:21,654][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:30:22,152][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:30:22,653][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:30:23,150][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:30:23,650][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:30:24,148][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:30:24,648][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:30:25,146][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:30:25,645][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:30:26,140][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:30:26,636][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:30:27,132][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:30:27,628][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:30:28,126][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:30:28,625][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:30:29,123][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:30:29,620][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:30:30,122][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:30:30,620][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:30:31,122][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:30:31,637][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:30:32,140][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:30:32,638][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:30:33,137][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:30:33,636][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:30:34,135][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:30:34,634][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:30:35,136][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:30:35,634][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:30:36,132][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:30:36,630][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:30:37,126][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:30:37,630][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:30:38,127][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:30:38,624][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:30:39,120][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:30:39,619][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:30:40,120][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:30:40,619][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:30:41,117][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:30:41,616][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:30:42,118][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:30:42,638][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:30:43,136][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10101 tokens. +[2026-03-26 07:30:43,992][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 60.62%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:32 +[2026-03-26 07:30:44,753][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:30:44,756][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:30:44,757][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:30:45,440][__main__][INFO] - Iteration 665 took 50s (28.71% Gen, 69.94% Train). Generation: 14s, Training: 35s. Estimated remaining time: 31h 34m 35s. Estimated total time: 41h 59m 41s. Time estimates for 10 more iterations: 8m 23s, 100 more iterations: 1h 23m 59s, 500 more iterations: 6h 59m 56s. +[2026-03-26 07:30:45,443][__main__][INFO] - Starting iteration 665. +[2026-03-26 07:30:45,856][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 66 and human policies 1. +[2026-03-26 07:30:45,857][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:31:02,663][__main__][INFO] - Number of regex retries in iteration 665: 0 +[2026-03-26 07:31:02,664][__main__][INFO] - agents played in iteration 665 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:31:03,445][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:31:03,469][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:31:03,493][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:31:03,517][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:31:03,517][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:31:03,518][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:31:04,339][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:31:04,793][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:31:05,295][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:31:05,792][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:31:06,289][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:31:06,790][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:31:07,286][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:31:07,781][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:31:08,278][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:31:08,781][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:31:09,299][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:31:09,803][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:31:10,305][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:31:10,805][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:31:11,309][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:31:11,812][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:31:12,318][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:31:12,830][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:31:13,335][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:31:13,838][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:31:14,340][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:31:14,840][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:31:15,338][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:31:15,838][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:31:16,337][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:31:16,839][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:31:17,341][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:31:17,842][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:31:18,344][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:31:18,844][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:31:19,345][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:31:19,848][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:31:20,351][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:31:20,850][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:31:21,348][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:31:21,846][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:31:22,343][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:31:22,840][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:31:23,340][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:31:23,837][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:31:24,336][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:31:24,837][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:31:25,338][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:31:25,838][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:31:26,338][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:31:26,837][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:31:27,337][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:31:27,836][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:31:28,337][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:31:28,833][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:31:29,334][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:31:29,831][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:31:30,328][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:31:30,830][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:31:31,328][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:31:31,827][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:31:32,325][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:31:32,826][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:31:33,323][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:31:33,821][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:31:34,320][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:31:34,818][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:31:35,315][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:31:35,814][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:31:36,314][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10195 tokens. +[2026-03-26 07:31:37,190][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 60.54%, Block Peak % of device VRAM: 62.04%, ΔTime: 00:00:32 +[2026-03-26 07:31:37,944][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:31:37,947][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:31:37,948][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:31:38,725][__main__][INFO] - Iteration 666 took 52s (31.79% Gen, 66.74% Train). Generation: 16s, Training: 35s. Estimated remaining time: 33h 37m 30s. Estimated total time: 44h 3m 29s. Time estimates for 10 more iterations: 8m 48s, 100 more iterations: 1h 28m 6s, 500 more iterations: 7h 20m 34s. +[2026-03-26 07:31:38,727][__main__][INFO] - Starting iteration 666. +[2026-03-26 07:31:39,137][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 66 and human policies 1. +[2026-03-26 07:31:39,137][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:31:53,900][__main__][INFO] - Number of regex retries in iteration 666: 0 +[2026-03-26 07:31:53,901][__main__][INFO] - agents played in iteration 666 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:31:54,685][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:31:54,711][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:31:54,736][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:31:54,761][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:31:54,762][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:31:54,762][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:31:55,596][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:31:56,056][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:31:56,566][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:31:57,067][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:31:57,567][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:31:58,073][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:31:58,575][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:31:59,078][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:31:59,584][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:32:00,084][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:32:00,582][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:32:01,085][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:32:01,585][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:32:02,085][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:32:02,584][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:32:03,083][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:32:03,581][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:32:04,083][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:32:04,581][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:32:05,079][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:32:05,578][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:32:06,077][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:32:06,577][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:32:07,076][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:32:07,573][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:32:08,079][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:32:08,583][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:32:09,093][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:32:09,594][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:32:10,097][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:32:10,597][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:32:11,099][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:32:11,603][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:32:12,102][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:32:12,599][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:32:13,097][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:32:13,594][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:32:14,092][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:32:14,591][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:32:15,089][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:32:15,586][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:32:16,085][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:32:16,586][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:32:17,087][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:32:17,591][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:32:18,092][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:32:18,604][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:32:19,104][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:32:19,607][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:32:20,111][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:32:20,615][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:32:21,119][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:32:21,627][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:32:22,130][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:32:22,633][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:32:23,136][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:32:23,637][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:32:24,136][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:32:24,636][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:32:25,138][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:32:25,639][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:32:26,138][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:32:26,639][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:32:27,143][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:32:27,646][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10124 tokens. +[2026-03-26 07:32:28,538][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.05%, Current % of VRAM taken: 60.48%, Block Peak % of device VRAM: 62.42%, ΔTime: 00:00:32 +[2026-03-26 07:32:29,299][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:32:29,301][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:32:29,303][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:32:30,035][__main__][INFO] - Iteration 667 took 50s (29.01% Gen, 69.55% Train). Generation: 14s, Training: 35s. Estimated remaining time: 31h 58m 8s. Estimated total time: 42h 24m 58s. Time estimates for 10 more iterations: 8m 28s, 100 more iterations: 1h 24m 49s, 500 more iterations: 7h 4m 9s. +[2026-03-26 07:32:30,038][__main__][INFO] - Starting iteration 667. +[2026-03-26 07:32:30,453][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 66 and human policies 1. +[2026-03-26 07:32:30,453][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:32:48,081][__main__][INFO] - Number of regex retries in iteration 667: 0 +[2026-03-26 07:32:48,082][__main__][INFO] - agents played in iteration 667 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:32:48,847][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:32:48,871][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:32:48,894][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:32:48,917][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:32:48,918][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:32:48,918][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:32:49,735][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:32:50,191][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:32:50,694][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:32:51,193][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:32:51,692][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:32:52,195][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:32:52,695][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:32:53,195][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:32:53,696][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:32:54,195][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:32:54,693][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:32:55,192][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:32:55,690][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:32:56,189][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:32:56,690][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:32:57,188][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:32:57,686][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:32:58,188][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:32:58,684][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:32:59,181][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:32:59,678][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:33:00,176][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:33:00,674][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:33:01,172][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:33:01,670][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:33:02,168][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:33:02,669][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:33:03,165][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:33:03,664][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:33:04,162][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:33:04,660][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:33:05,158][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:33:05,655][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:33:06,153][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:33:06,655][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:33:07,154][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:33:07,657][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:33:08,156][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:33:08,665][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:33:09,169][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:33:09,669][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:33:10,172][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:33:10,672][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:33:11,172][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:33:11,673][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:33:12,174][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:33:12,673][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:33:13,172][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:33:13,672][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:33:14,178][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:33:14,680][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:33:15,183][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:33:15,686][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:33:16,187][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:33:16,690][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:33:17,189][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:33:17,691][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:33:18,190][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:33:18,688][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:33:19,190][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:33:19,687][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:33:20,188][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:33:20,688][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:33:21,186][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:33:21,685][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10137 tokens. +[2026-03-26 07:33:22,542][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 60.53%, Block Peak % of device VRAM: 62.18%, ΔTime: 00:00:32 +[2026-03-26 07:33:23,298][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:33:23,300][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:33:23,302][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:33:24,071][__main__][INFO] - Iteration 668 took 53s (32.88% Gen, 65.69% Train). Generation: 17s, Training: 35s. Estimated remaining time: 34h 13m 14s. Estimated total time: 44h 40m 57s. Time estimates for 10 more iterations: 8m 56s, 100 more iterations: 1h 29m 21s, 500 more iterations: 7h 26m 49s. +[2026-03-26 07:33:24,073][__main__][INFO] - Starting iteration 668. +[2026-03-26 07:33:24,489][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 66 and human policies 1. +[2026-03-26 07:33:24,490][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:33:35,196][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 07:33:38,278][__main__][INFO] - Number of regex retries in iteration 668: 1 +[2026-03-26 07:33:38,279][__main__][INFO] - agents played in iteration 668 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:33:39,051][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:33:39,075][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:33:39,099][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:33:39,122][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:33:39,123][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:33:39,123][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:33:39,942][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:33:40,400][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:33:40,903][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:33:41,402][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:33:41,902][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:33:42,402][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:33:42,901][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:33:43,400][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:33:43,899][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:33:44,396][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:33:44,896][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:33:45,394][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:33:45,892][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:33:46,390][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:33:46,889][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:33:47,386][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:33:47,884][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:33:48,380][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:33:48,876][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:33:49,374][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:33:49,876][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:33:50,381][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:33:50,877][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:33:51,374][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:33:51,871][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:33:52,373][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:33:52,870][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:33:53,369][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:33:53,871][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:33:54,370][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:33:54,868][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:33:55,365][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:33:55,862][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:33:56,383][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:33:56,880][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:33:57,378][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:33:57,876][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:33:58,375][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:33:58,875][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:33:59,390][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:33:59,894][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:34:00,397][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:34:00,899][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:34:01,399][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:34:01,900][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:34:02,402][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:34:02,907][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:34:03,413][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:34:03,915][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:34:04,416][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:34:04,923][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:34:05,421][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:34:05,921][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:34:06,420][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:34:06,919][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:34:07,420][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:34:07,918][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:34:08,416][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:34:08,916][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:34:09,416][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:34:09,914][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:34:10,412][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:34:10,910][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:34:11,409][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:34:11,909][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10068 tokens. +[2026-03-26 07:34:12,793][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.08%, ΔTime: 00:00:32 +[2026-03-26 07:34:13,604][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:34:13,607][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:34:13,609][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:34:14,338][__main__][INFO] - Iteration 669 took 49s (27.66% Gen, 70.87% Train). Generation: 13s, Training: 35s. Estimated remaining time: 31h 3m 53s. Estimated total time: 41h 32m 27s. Time estimates for 10 more iterations: 8m 18s, 100 more iterations: 1h 23m 4s, 500 more iterations: 6h 55m 24s. +[2026-03-26 07:34:14,340][__main__][INFO] - Starting iteration 669. +[2026-03-26 07:34:14,744][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 66 and human policies 1. +[2026-03-26 07:34:14,745][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:34:29,624][__main__][INFO] - Number of regex retries in iteration 669: 0 +[2026-03-26 07:34:29,625][__main__][INFO] - agents played in iteration 669 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:34:30,392][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:34:30,416][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:34:30,440][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:34:30,464][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:34:30,465][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:34:30,465][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:34:31,288][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:34:31,752][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:34:32,260][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:34:32,764][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:34:33,271][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:34:33,776][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:34:34,295][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:34:34,799][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:34:35,301][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:34:35,800][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:34:36,305][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:34:36,805][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:34:37,307][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:34:37,809][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:34:38,309][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:34:38,811][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:34:39,310][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:34:39,810][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:34:40,310][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:34:40,811][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:34:41,316][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:34:41,824][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:34:42,326][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:34:42,826][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:34:43,327][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:34:43,831][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:34:44,332][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:34:44,835][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:34:45,334][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:34:45,837][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:34:46,336][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:34:46,838][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:34:47,340][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:34:47,837][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:34:48,335][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:34:48,833][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:34:49,332][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:34:49,831][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:34:50,330][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:34:50,828][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:34:51,330][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:34:51,832][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:34:52,330][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:34:52,831][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:34:53,329][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:34:53,828][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:34:54,325][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:34:54,824][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:34:55,323][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:34:55,824][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:34:56,321][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:34:56,817][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:34:57,314][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:34:57,811][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:34:58,311][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:34:58,806][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:34:59,302][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:34:59,800][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:35:00,300][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:35:00,801][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:35:01,300][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:35:01,803][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:35:02,314][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:35:02,817][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:35:03,318][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10163 tokens. +[2026-03-26 07:35:04,186][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 60.54%, Block Peak % of device VRAM: 62.27%, ΔTime: 00:00:32 +[2026-03-26 07:35:04,943][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:35:04,946][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:35:04,947][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:35:05,683][__main__][INFO] - Iteration 670 took 50s (29.21% Gen, 69.34% Train). Generation: 14s, Training: 35s. Estimated remaining time: 31h 57m 32s. Estimated total time: 42h 26m 57s. Time estimates for 10 more iterations: 8m 29s, 100 more iterations: 1h 24m 53s, 500 more iterations: 7h 4m 29s. +[2026-03-26 07:35:05,685][__main__][INFO] - Starting iteration 670. +[2026-03-26 07:35:06,101][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 66 and human policies 1. +[2026-03-26 07:35:06,101][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:35:20,543][__main__][INFO] - Number of regex retries in iteration 670: 0 +[2026-03-26 07:35:20,544][__main__][INFO] - agents played in iteration 670 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:35:21,353][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:35:21,386][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:35:21,414][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:35:21,441][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:35:21,441][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:35:21,442][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:35:22,274][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:35:22,733][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:35:23,238][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:35:23,736][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:35:24,233][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:35:24,735][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:35:25,232][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:35:25,728][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:35:26,225][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:35:26,723][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:35:27,224][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:35:27,744][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:35:28,246][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:35:28,747][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:35:29,247][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:35:29,746][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:35:30,246][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:35:30,749][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:35:31,257][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:35:31,766][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:35:32,270][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:35:32,774][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:35:33,276][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:35:33,777][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:35:34,279][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:35:34,799][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:35:35,311][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:35:35,824][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:35:36,325][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:35:36,826][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:35:37,330][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:35:37,833][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:35:38,333][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:35:38,836][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:35:39,334][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:35:39,836][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:35:40,341][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:35:40,854][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:35:41,358][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:35:41,857][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:35:42,362][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:35:42,860][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:35:43,359][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:35:43,857][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:35:44,355][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:35:44,853][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:35:45,352][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:35:45,852][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:35:46,355][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:35:46,868][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:35:47,367][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:35:47,868][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:35:48,367][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:35:48,867][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:35:49,367][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:35:49,870][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:35:50,368][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:35:50,866][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:35:51,365][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:35:51,863][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:35:52,362][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:35:52,860][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:35:53,358][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:35:53,857][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:35:54,356][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10117 tokens. +[2026-03-26 07:35:55,221][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 60.54%, Block Peak % of device VRAM: 62.24%, ΔTime: 00:00:32 +[2026-03-26 07:35:55,969][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:35:55,971][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:35:55,973][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:35:57,287][__main__][INFO] - Iteration 671 took 51s (28.22% Gen, 69.21% Train). Generation: 14s, Training: 35s. Estimated remaining time: 32h 9m 5s. Estimated total time: 42h 39m 22s. Time estimates for 10 more iterations: 8m 31s, 100 more iterations: 1h 25m 18s, 500 more iterations: 7h 6m 33s. +[2026-03-26 07:35:57,290][__main__][INFO] - Starting iteration 671. +[2026-03-26 07:35:57,704][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 67 and human policies 1. +[2026-03-26 07:35:57,705][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:36:07,964][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 20 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 07:36:12,069][__main__][INFO] - Number of regex retries in iteration 671: 1 +[2026-03-26 07:36:12,070][__main__][INFO] - agents played in iteration 671 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:36:12,842][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:36:12,866][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:36:12,891][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:36:12,915][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:36:12,916][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:36:12,916][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:36:13,758][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:36:14,221][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:36:14,722][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:36:15,220][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:36:15,719][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:36:16,217][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:36:16,713][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:36:17,212][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:36:17,710][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:36:18,213][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:36:18,721][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:36:19,237][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:36:19,739][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:36:20,241][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:36:20,744][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:36:21,245][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:36:21,745][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:36:22,247][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:36:22,749][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:36:23,252][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:36:23,752][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:36:24,255][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:36:24,754][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:36:25,256][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:36:25,757][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:36:26,257][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:36:26,759][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:36:27,258][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:36:27,758][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:36:28,256][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:36:28,759][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:36:29,262][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:36:29,762][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:36:30,267][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:36:30,769][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:36:31,268][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:36:31,768][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:36:32,272][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:36:32,772][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:36:33,278][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:36:33,780][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:36:34,278][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:36:34,777][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:36:35,275][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:36:35,774][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:36:36,272][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:36:36,775][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:36:37,274][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:36:37,770][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:36:38,268][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:36:38,766][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:36:39,265][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:36:39,765][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:36:40,261][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:36:40,758][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:36:41,254][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:36:41,752][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:36:42,249][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:36:42,745][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:36:43,241][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:36:43,737][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:36:44,235][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:36:44,733][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:36:45,232][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:36:45,729][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10088 tokens. +[2026-03-26 07:36:46,593][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.05%, Current % of VRAM taken: 60.48%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 +[2026-03-26 07:36:47,345][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:36:47,347][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:36:47,349][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:36:48,076][__main__][INFO] - Iteration 672 took 50s (28.52% Gen, 70.03% Train). Generation: 14s, Training: 35s. Estimated remaining time: 31h 27m 31s. Estimated total time: 41h 58m 39s. Time estimates for 10 more iterations: 8m 23s, 100 more iterations: 1h 23m 57s, 500 more iterations: 6h 59m 46s. +[2026-03-26 07:36:48,079][__main__][INFO] - Starting iteration 672. +[2026-03-26 07:36:48,489][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 67 and human policies 1. +[2026-03-26 07:36:48,490][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:37:05,201][__main__][INFO] - Number of regex retries in iteration 672: 0 +[2026-03-26 07:37:05,202][__main__][INFO] - agents played in iteration 672 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:37:05,982][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:37:06,007][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:37:06,032][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:37:06,057][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:37:06,058][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:37:06,058][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:37:06,903][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:37:07,361][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:37:07,871][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:37:08,376][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:37:08,883][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:37:09,384][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:37:09,886][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:37:10,386][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:37:10,887][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:37:11,385][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:37:11,889][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:37:12,397][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:37:12,896][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:37:13,397][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:37:13,899][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:37:14,399][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:37:14,898][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:37:15,396][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:37:15,894][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:37:16,399][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:37:16,909][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:37:17,407][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:37:17,910][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:37:18,410][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:37:18,908][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:37:19,409][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:37:19,909][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:37:20,410][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:37:20,912][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:37:21,417][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:37:21,926][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:37:22,428][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:37:22,929][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:37:23,432][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:37:23,934][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:37:24,440][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:37:24,940][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:37:25,440][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:37:25,946][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:37:26,445][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:37:26,946][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:37:27,445][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:37:27,944][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:37:28,445][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:37:28,943][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:37:29,441][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:37:29,943][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:37:30,441][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:37:30,942][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:37:31,441][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:37:31,944][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:37:32,442][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:37:32,940][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:37:33,441][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:37:33,937][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:37:34,436][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:37:34,938][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:37:35,441][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:37:35,955][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:37:36,461][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:37:36,965][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:37:37,466][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:37:37,966][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:37:38,465][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:37:38,968][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10151 tokens. +[2026-03-26 07:37:39,845][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 60.61%, Block Peak % of device VRAM: 62.15%, ΔTime: 00:00:32 +[2026-03-26 07:37:40,623][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:37:40,625][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:37:40,627][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:37:41,340][__main__][INFO] - Iteration 673 took 52s (31.62% Gen, 67.03% Train). Generation: 16s, Training: 35s. Estimated remaining time: 33h 30m 33s. Estimated total time: 44h 2m 34s. Time estimates for 10 more iterations: 8m 48s, 100 more iterations: 1h 28m 5s, 500 more iterations: 7h 20m 25s. +[2026-03-26 07:37:41,343][__main__][INFO] - Starting iteration 673. +[2026-03-26 07:37:41,756][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 67 and human policies 1. +[2026-03-26 07:37:41,757][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:38:02,038][__main__][INFO] - Number of regex retries in iteration 673: 0 +[2026-03-26 07:38:02,038][__main__][INFO] - agents played in iteration 673 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:38:02,837][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:38:02,860][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:38:02,884][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:38:02,907][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:38:02,908][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:38:02,908][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:38:03,744][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:38:04,203][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:38:04,707][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:38:05,207][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:38:05,707][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:38:06,212][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:38:06,717][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:38:07,221][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:38:07,731][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:38:08,234][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:38:08,734][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:38:09,235][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:38:09,735][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:38:10,237][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:38:10,737][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:38:11,237][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:38:11,737][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:38:12,237][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:38:12,737][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:38:13,235][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:38:13,734][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:38:14,233][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:38:14,734][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:38:15,236][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:38:15,734][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:38:16,234][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:38:16,735][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:38:17,236][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:38:17,740][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:38:18,249][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:38:18,752][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:38:19,254][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:38:19,755][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:38:20,256][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:38:20,757][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:38:21,257][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:38:21,757][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:38:22,258][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:38:22,757][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:38:23,256][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:38:23,755][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:38:24,255][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:38:24,754][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:38:25,254][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:38:25,753][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:38:26,252][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:38:26,752][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:38:27,251][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:38:27,751][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:38:28,253][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:38:28,755][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:38:29,257][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:38:29,755][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:38:30,257][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:38:30,757][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:38:31,257][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:38:31,759][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:38:32,260][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:38:32,759][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:38:33,259][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:38:33,758][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:38:34,260][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:38:34,758][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:38:35,256][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:38:35,756][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10023 tokens. +[2026-03-26 07:38:36,620][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.58%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:32 +[2026-03-26 07:38:37,378][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:38:37,380][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:38:37,382][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:38:38,236][__main__][INFO] - Iteration 674 took 56s (35.91% Gen, 62.58% Train). Generation: 20s, Training: 35s. Estimated remaining time: 36h 31m 1s. Estimated total time: 47h 3m 59s. Time estimates for 10 more iterations: 9m 24s, 100 more iterations: 1h 34m 7s, 500 more iterations: 7h 50m 39s. +[2026-03-26 07:38:38,238][__main__][INFO] - Starting iteration 674. +[2026-03-26 07:38:38,648][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 67 and human policies 1. +[2026-03-26 07:38:38,648][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:39:05,668][__main__][INFO] - Number of regex retries in iteration 674: 0 +[2026-03-26 07:39:05,669][__main__][INFO] - agents played in iteration 674 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:39:06,459][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:39:06,483][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:39:06,507][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:39:06,530][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:39:06,531][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:39:06,531][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:39:07,356][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:39:07,816][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:39:08,333][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:39:08,834][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:39:09,339][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:39:09,839][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:39:10,339][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:39:10,840][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:39:11,356][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:39:11,859][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:39:12,359][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:39:12,859][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:39:13,361][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:39:13,861][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:39:14,360][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:39:14,859][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:39:15,361][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:39:15,863][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:39:16,361][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:39:16,860][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:39:17,357][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:39:17,862][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:39:18,373][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:39:18,876][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:39:19,378][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:39:19,878][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:39:20,377][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:39:20,876][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:39:21,375][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:39:21,877][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:39:22,377][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:39:22,876][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:39:23,376][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:39:23,877][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:39:24,382][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:39:24,882][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:39:25,392][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:39:25,893][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:39:26,394][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:39:26,896][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:39:27,415][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:39:27,920][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:39:28,419][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:39:28,923][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:39:29,421][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:39:29,923][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:39:30,420][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:39:30,919][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:39:31,417][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:39:31,920][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:39:32,422][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:39:32,920][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:39:33,417][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:39:33,916][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:39:34,414][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:39:34,915][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:39:35,413][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:39:35,910][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:39:36,408][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:39:36,906][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:39:37,403][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:39:37,898][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:39:38,395][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:39:38,893][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:39:39,394][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10149 tokens. +[2026-03-26 07:39:40,290][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.05%, Current % of VRAM taken: 60.47%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:32 +[2026-03-26 07:39:42,748][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:39:42,750][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:39:42,752][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:39:43,480][__main__][INFO] - Iteration 675 took 1m 4s (41.68% Gen, 57.20% Train). Generation: 27s, Training: 37s. Estimated remaining time: 43h 27m 35s. Estimated total time: 54h 1m 38s. Time estimates for 10 more iterations: 10m 48s, 100 more iterations: 1h 48m 3s, 500 more iterations: 9h 0m 16s. +[2026-03-26 07:39:43,482][__main__][INFO] - Starting iteration 675. +[2026-03-26 07:39:43,894][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 67 and human policies 1. +[2026-03-26 07:39:43,895][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:40:05,331][__main__][INFO] - Number of regex retries in iteration 675: 0 +[2026-03-26 07:40:05,332][__main__][INFO] - agents played in iteration 675 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:40:06,086][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:40:06,110][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:40:06,134][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:40:06,159][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:40:06,159][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:40:06,160][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:40:06,998][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:40:07,454][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:40:07,983][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:40:08,486][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:40:08,987][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:40:09,491][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:40:09,994][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:40:10,495][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:40:10,996][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:40:11,496][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:40:11,995][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:40:12,493][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:40:12,992][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:40:13,495][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:40:13,995][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:40:14,494][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:40:14,999][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:40:15,507][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:40:16,012][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:40:16,511][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:40:17,011][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:40:17,511][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:40:18,010][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:40:18,513][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:40:19,012][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:40:19,513][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:40:20,015][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:40:20,517][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:40:21,017][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:40:21,517][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:40:22,018][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:40:22,518][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:40:23,021][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:40:23,530][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:40:24,035][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:40:24,538][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:40:25,042][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:40:25,546][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:40:26,050][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:40:26,556][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:40:27,060][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:40:27,560][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:40:28,059][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:40:28,558][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:40:29,060][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:40:29,564][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:40:30,070][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:40:30,572][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:40:31,073][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:40:31,574][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:40:32,080][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:40:32,580][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:40:33,081][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:40:33,581][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:40:34,085][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:40:34,584][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:40:35,085][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:40:35,583][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:40:36,081][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:40:36,584][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:40:37,094][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:40:37,596][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:40:38,095][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:40:38,593][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:40:39,093][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10065 tokens. +[2026-03-26 07:40:39,996][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 60.49%, Block Peak % of device VRAM: 62.13%, ΔTime: 00:00:33 +[2026-03-26 07:40:40,753][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:40:40,755][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:40:40,756][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:40:41,523][__main__][INFO] - Iteration 676 took 57s (37.20% Gen, 61.47% Train). Generation: 21s, Training: 35s. Estimated remaining time: 37h 26m 26s. Estimated total time: 48h 1m 27s. Time estimates for 10 more iterations: 9m 36s, 100 more iterations: 1h 36m 2s, 500 more iterations: 8h 0m 14s. +[2026-03-26 07:40:41,525][__main__][INFO] - Starting iteration 676. +[2026-03-26 07:40:41,931][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 67 and human policies 1. +[2026-03-26 07:40:41,932][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:41:02,471][__main__][INFO] - Number of regex retries in iteration 676: 0 +[2026-03-26 07:41:02,472][__main__][INFO] - agents played in iteration 676 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:41:03,245][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:41:03,271][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:41:03,302][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:41:03,330][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:41:03,331][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:41:03,331][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:41:04,191][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:41:04,653][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:41:05,161][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:41:05,678][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:41:06,180][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:41:06,686][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:41:07,191][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:41:07,692][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:41:08,194][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:41:08,697][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:41:09,208][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:41:09,712][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:41:10,214][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:41:10,717][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:41:11,218][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:41:11,719][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:41:12,218][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:41:12,720][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:41:13,233][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:41:13,736][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:41:14,240][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:41:14,742][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:41:15,246][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:41:15,748][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:41:16,248][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:41:16,752][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:41:17,254][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:41:17,759][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:41:18,267][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:41:18,767][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:41:19,271][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:41:19,772][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:41:20,274][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:41:20,777][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:41:21,278][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:41:21,780][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:41:22,282][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:41:22,788][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:41:23,293][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:41:23,801][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:41:24,305][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:41:24,803][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:41:25,310][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:41:25,811][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:41:26,310][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:41:26,809][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:41:27,307][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:41:27,805][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:41:28,304][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:41:28,805][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:41:29,307][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:41:29,809][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:41:30,316][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:41:30,821][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:41:31,323][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:41:31,828][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:41:32,329][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:41:32,835][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:41:33,334][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:41:33,834][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:41:34,333][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:41:34,832][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:41:35,331][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:41:35,830][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:41:36,329][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10171 tokens. +[2026-03-26 07:41:37,210][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 60.50%, Block Peak % of device VRAM: 62.16%, ΔTime: 00:00:33 +[2026-03-26 07:41:37,965][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:41:37,968][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:41:37,969][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:41:38,812][__main__][INFO] - Iteration 677 took 56s (36.11% Gen, 62.41% Train). Generation: 20s, Training: 35s. Estimated remaining time: 36h 48m 4s. Estimated total time: 47h 24m 3s. Time estimates for 10 more iterations: 9m 28s, 100 more iterations: 1h 34m 48s, 500 more iterations: 7h 54m 0s. +[2026-03-26 07:41:38,814][__main__][INFO] - Starting iteration 677. +[2026-03-26 07:41:39,223][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 67 and human policies 1. +[2026-03-26 07:41:39,223][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:41:52,812][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 07:42:01,129][__main__][INFO] - Number of regex retries in iteration 677: 1 +[2026-03-26 07:42:01,130][__main__][INFO] - agents played in iteration 677 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:42:01,902][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:42:01,926][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:42:01,951][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:42:01,976][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:42:01,976][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:42:01,977][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:42:02,873][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:42:03,335][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:42:03,839][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:42:04,343][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:42:04,842][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:42:05,363][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:42:05,869][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:42:06,381][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:42:06,887][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:42:07,391][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:42:07,893][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:42:08,403][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:42:08,912][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:42:09,418][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:42:09,922][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:42:10,423][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:42:10,924][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:42:11,429][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:42:11,937][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:42:12,450][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:42:12,953][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:42:13,457][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:42:13,959][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:42:14,460][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:42:14,961][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:42:15,462][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:42:15,961][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:42:16,462][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:42:16,966][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:42:17,469][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:42:17,967][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:42:18,467][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:42:18,969][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:42:19,482][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:42:19,986][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:42:20,488][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:42:20,990][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:42:21,495][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:42:21,994][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:42:22,494][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:42:22,994][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:42:23,495][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:42:23,995][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:42:24,500][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:42:25,009][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:42:25,512][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:42:26,013][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:42:26,515][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:42:27,017][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:42:27,517][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:42:28,016][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:42:28,516][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:42:29,016][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:42:29,518][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:42:30,017][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:42:30,514][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:42:31,017][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:42:31,521][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:42:32,030][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:42:32,530][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:42:33,034][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:42:33,535][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:42:34,037][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:42:34,537][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:42:35,039][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10086 tokens. +[2026-03-26 07:42:35,928][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 60.64%, Block Peak % of device VRAM: 62.17%, ΔTime: 00:00:33 +[2026-03-26 07:42:36,688][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:42:36,690][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:42:36,691][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:42:37,414][__main__][INFO] - Iteration 678 took 58s (37.65% Gen, 61.11% Train). Generation: 21s, Training: 35s. Estimated remaining time: 37h 52m 38s. Estimated total time: 48h 29m 36s. Time estimates for 10 more iterations: 9m 41s, 100 more iterations: 1h 36m 59s, 500 more iterations: 8h 4m 56s. +[2026-03-26 07:42:37,416][__main__][INFO] - Starting iteration 678. +[2026-03-26 07:42:37,818][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 67 and human policies 1. +[2026-03-26 07:42:37,818][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:42:57,653][__main__][INFO] - Number of regex retries in iteration 678: 0 +[2026-03-26 07:42:57,654][__main__][INFO] - agents played in iteration 678 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:42:58,419][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:42:58,443][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:42:58,467][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:42:58,491][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:42:58,491][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:42:58,492][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:42:59,325][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:42:59,782][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:43:00,285][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:43:00,785][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:43:01,284][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:43:01,784][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:43:02,283][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:43:02,781][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:43:03,279][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:43:03,776][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:43:04,275][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:43:04,773][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:43:05,274][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:43:05,780][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:43:06,281][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:43:06,779][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:43:07,278][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:43:07,780][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:43:08,283][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:43:08,791][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:43:09,295][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:43:09,796][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:43:10,298][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:43:10,799][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:43:11,300][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:43:11,798][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:43:12,296][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:43:12,796][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:43:13,295][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:43:13,793][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:43:14,292][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:43:14,792][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:43:15,293][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:43:15,791][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:43:16,292][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:43:16,803][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:43:17,302][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:43:17,806][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:43:18,305][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:43:18,803][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:43:19,302][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:43:19,802][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:43:20,303][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:43:20,803][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:43:21,302][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:43:21,800][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:43:22,297][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:43:22,798][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:43:23,297][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:43:23,795][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:43:24,293][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:43:24,791][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:43:25,289][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:43:25,791][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:43:26,291][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:43:26,788][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:43:27,285][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:43:27,787][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:43:28,306][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:43:28,806][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:43:29,306][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:43:29,810][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:43:30,311][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:43:30,810][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:43:31,311][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10096 tokens. +[2026-03-26 07:43:32,201][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 60.61%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 +[2026-03-26 07:43:32,959][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:43:32,962][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:43:32,963][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:43:33,693][__main__][INFO] - Iteration 679 took 55s (35.50% Gen, 63.19% Train). Generation: 19s, Training: 35s. Estimated remaining time: 35h 55m 55s. Estimated total time: 46h 33m 48s. Time estimates for 10 more iterations: 9m 18s, 100 more iterations: 1h 33m 7s, 500 more iterations: 7h 45m 38s. +[2026-03-26 07:43:33,695][__main__][INFO] - Starting iteration 679. +[2026-03-26 07:43:34,094][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 67 and human policies 1. +[2026-03-26 07:43:34,094][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:43:49,527][__main__][INFO] - Number of regex retries in iteration 679: 0 +[2026-03-26 07:43:49,528][__main__][INFO] - agents played in iteration 679 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:43:50,301][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:43:50,328][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:43:50,353][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:43:50,378][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:43:50,379][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:43:50,379][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:43:51,235][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:43:51,699][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:43:52,203][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:43:52,707][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:43:53,212][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:43:53,731][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:43:54,237][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:43:54,738][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:43:55,242][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:43:55,744][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:43:56,244][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:43:56,744][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:43:57,246][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:43:57,754][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:43:58,254][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:43:58,763][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:43:59,274][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:43:59,776][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:44:00,276][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:44:00,777][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:44:01,278][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:44:01,779][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:44:02,283][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:44:02,782][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:44:03,281][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:44:03,789][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:44:04,292][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:44:04,794][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:44:05,296][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:44:05,798][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:44:06,299][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:44:06,800][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:44:07,302][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:44:07,800][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:44:08,300][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:44:08,802][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:44:09,323][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:44:09,822][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:44:10,320][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:44:10,819][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:44:11,318][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:44:11,820][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:44:12,321][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:44:12,819][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:44:13,317][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:44:13,815][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:44:14,316][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:44:14,813][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:44:15,312][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:44:15,813][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:44:16,312][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:44:16,811][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:44:17,310][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:44:17,808][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:44:18,305][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:44:18,804][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:44:19,301][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:44:19,799][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:44:20,297][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:44:20,794][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:44:21,292][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:44:21,788][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:44:22,289][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:44:22,785][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:44:23,282][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10118 tokens. +[2026-03-26 07:44:24,163][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 60.54%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:32 +[2026-03-26 07:44:24,917][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:44:24,919][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:44:24,921][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:44:25,689][__main__][INFO] - Iteration 680 took 51s (29.91% Gen, 68.60% Train). Generation: 15s, Training: 35s. Estimated remaining time: 32h 21m 3s. Estimated total time: 42h 59m 49s. Time estimates for 10 more iterations: 8m 35s, 100 more iterations: 1h 25m 59s, 500 more iterations: 7h 9m 58s. +[2026-03-26 07:44:25,692][__main__][INFO] - Starting iteration 680. +[2026-03-26 07:44:26,106][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 67 and human policies 1. +[2026-03-26 07:44:26,106][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:44:39,733][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 1 z ball did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 07:44:40,571][__main__][INFO] - Number of regex retries in iteration 680: 1 +[2026-03-26 07:44:40,572][__main__][INFO] - agents played in iteration 680 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:44:41,342][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:44:41,366][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:44:41,389][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:44:41,412][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:44:41,413][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:44:41,414][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:44:42,248][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:44:42,704][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:44:43,208][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:44:43,710][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:44:44,219][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:44:44,722][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:44:45,221][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:44:45,722][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:44:46,221][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:44:46,723][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:44:47,223][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:44:47,730][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:44:48,232][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:44:48,732][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:44:49,232][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:44:49,735][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:44:50,235][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:44:50,735][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:44:51,235][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:44:51,735][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:44:52,238][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:44:52,767][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:44:53,268][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:44:53,773][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:44:54,276][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:44:54,776][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:44:55,277][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:44:55,783][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:44:56,282][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:44:56,786][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:44:57,285][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:44:57,783][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:44:58,283][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:44:58,785][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:44:59,285][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:44:59,783][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:45:00,284][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:45:00,783][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:45:01,282][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:45:01,781][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:45:02,282][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:45:02,781][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:45:03,280][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:45:03,780][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:45:04,279][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:45:04,780][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:45:05,278][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:45:05,777][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:45:06,276][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:45:06,775][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:45:07,274][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:45:07,773][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:45:08,271][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:45:08,769][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:45:09,268][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:45:09,765][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:45:10,263][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:45:10,763][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:45:11,260][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:45:11,759][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:45:12,258][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:45:12,757][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:45:13,258][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:45:13,756][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:45:14,255][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10052 tokens. +[2026-03-26 07:45:15,112][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 60.54%, Block Peak % of device VRAM: 62.11%, ΔTime: 00:00:32 +[2026-03-26 07:45:15,876][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:45:15,878][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:45:15,880][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:45:17,347][__main__][INFO] - Iteration 681 took 51s (28.23% Gen, 68.90% Train). Generation: 14s, Training: 35s. Estimated remaining time: 32h 2m 31s. Estimated total time: 42h 42m 8s. Time estimates for 10 more iterations: 8m 32s, 100 more iterations: 1h 25m 24s, 500 more iterations: 7h 7m 1s. +[2026-03-26 07:45:17,350][__main__][INFO] - Starting iteration 681. +[2026-03-26 07:45:17,761][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 68 and human policies 1. +[2026-03-26 07:45:17,761][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:45:22,706][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 07:45:32,842][__main__][INFO] - Number of regex retries in iteration 681: 1 +[2026-03-26 07:45:32,843][__main__][INFO] - agents played in iteration 681 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:45:33,615][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:45:33,638][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:45:33,662][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:45:33,685][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:45:33,686][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:45:33,686][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:45:34,513][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:45:34,975][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:45:35,494][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:45:35,997][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:45:36,500][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:45:37,001][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:45:37,503][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:45:38,003][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:45:38,504][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:45:39,004][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:45:39,504][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:45:40,007][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:45:40,515][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:45:41,019][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:45:41,529][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:45:42,033][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:45:42,534][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:45:43,035][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:45:43,534][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:45:44,033][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:45:44,536][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:45:45,039][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:45:45,549][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:45:46,050][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:45:46,551][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:45:47,051][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:45:47,550][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:45:48,050][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:45:48,548][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:45:49,047][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:45:49,547][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:45:50,046][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:45:50,545][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:45:51,043][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:45:51,541][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:45:52,042][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:45:52,547][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:45:53,054][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:45:53,554][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:45:54,055][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:45:54,556][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:45:55,057][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:45:55,558][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:45:56,060][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:45:56,561][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:45:57,063][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:45:57,566][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:45:58,071][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:45:58,575][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:45:59,076][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:45:59,579][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:46:00,080][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:46:00,579][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:46:01,103][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:46:01,605][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:46:02,109][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:46:02,607][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:46:03,107][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:46:03,608][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:46:04,108][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:46:04,607][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:46:05,107][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:46:05,606][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:46:06,109][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:46:06,608][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10051 tokens. +[2026-03-26 07:46:07,479][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.58%, Block Peak % of device VRAM: 62.09%, ΔTime: 00:00:32 +[2026-03-26 07:46:08,257][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:46:08,260][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:46:08,261][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:46:08,992][__main__][INFO] - Iteration 682 took 51s (29.44% Gen, 69.13% Train). Generation: 15s, Training: 35s. Estimated remaining time: 32h 1m 8s. Estimated total time: 42h 41m 37s. Time estimates for 10 more iterations: 8m 32s, 100 more iterations: 1h 25m 23s, 500 more iterations: 7h 6m 56s. +[2026-03-26 07:46:08,997][__main__][INFO] - Starting iteration 682. +[2026-03-26 07:46:09,406][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 68 and human policies 1. +[2026-03-26 07:46:09,407][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:46:24,277][__main__][INFO] - Number of regex retries in iteration 682: 0 +[2026-03-26 07:46:24,278][__main__][INFO] - agents played in iteration 682 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:46:25,043][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:46:25,067][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:46:25,091][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:46:25,115][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:46:25,115][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:46:25,116][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:46:25,967][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:46:26,420][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:46:26,923][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:46:27,421][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:46:27,920][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:46:28,418][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:46:28,916][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:46:29,415][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:46:29,918][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:46:30,417][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:46:30,916][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:46:31,416][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:46:31,922][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:46:32,426][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:46:32,929][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:46:33,437][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:46:33,939][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:46:34,439][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:46:34,939][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:46:35,440][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:46:35,943][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:46:36,444][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:46:36,943][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:46:37,444][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:46:37,942][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:46:38,440][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:46:38,939][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:46:39,440][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:46:39,938][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:46:40,436][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:46:40,938][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:46:41,439][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:46:41,949][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:46:42,455][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:46:42,955][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:46:43,456][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:46:43,956][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:46:44,457][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:46:44,959][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:46:45,460][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:46:45,961][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:46:46,465][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:46:46,967][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:46:47,468][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:46:47,968][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:46:48,470][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:46:48,973][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:46:49,482][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:46:49,980][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:46:50,482][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:46:50,981][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:46:51,480][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:46:51,982][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:46:52,481][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:46:52,980][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:46:53,480][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:46:53,979][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:46:54,478][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:46:54,980][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:46:55,479][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:46:55,977][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:46:56,476][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:46:56,979][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:46:57,477][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:46:57,977][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10154 tokens. +[2026-03-26 07:46:58,860][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.57%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:32 +[2026-03-26 07:46:59,622][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:46:59,624][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:46:59,626][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:47:00,360][__main__][INFO] - Iteration 683 took 50s (29.19% Gen, 69.37% Train). Generation: 14s, Training: 35s. Estimated remaining time: 31h 46m 22s. Estimated total time: 42h 27m 42s. Time estimates for 10 more iterations: 8m 29s, 100 more iterations: 1h 24m 55s, 500 more iterations: 7h 4m 37s. +[2026-03-26 07:47:00,362][__main__][INFO] - Starting iteration 683. +[2026-03-26 07:47:00,775][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 68 and human policies 1. +[2026-03-26 07:47:00,776][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:47:14,752][__main__][INFO] - Number of regex retries in iteration 683: 0 +[2026-03-26 07:47:14,752][__main__][INFO] - agents played in iteration 683 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:47:15,529][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:47:15,553][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:47:15,577][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:47:15,602][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:47:15,603][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:47:15,603][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:47:16,466][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:47:16,934][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:47:17,443][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:47:17,944][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:47:18,444][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:47:18,945][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:47:19,450][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:47:19,963][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:47:20,473][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:47:20,975][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:47:21,477][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:47:21,979][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:47:22,483][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:47:22,986][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:47:23,485][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:47:23,992][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:47:24,499][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:47:25,001][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:47:25,501][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:47:26,001][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:47:26,501][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:47:27,005][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:47:27,503][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:47:28,006][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:47:28,506][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:47:29,004][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:47:29,502][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:47:30,001][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:47:30,505][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:47:31,003][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:47:31,508][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:47:32,011][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:47:32,517][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:47:33,023][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:47:33,523][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:47:34,025][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:47:34,525][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:47:35,025][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:47:35,527][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:47:36,028][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:47:36,530][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:47:37,028][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:47:37,524][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:47:38,020][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:47:38,516][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:47:39,012][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:47:39,509][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:47:40,006][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:47:40,502][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:47:41,002][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:47:41,499][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:47:41,996][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:47:42,493][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:47:42,990][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:47:43,486][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:47:43,982][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:47:44,479][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:47:44,976][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:47:45,473][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:47:45,971][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:47:46,470][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:47:46,968][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:47:47,468][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:47:47,968][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:47:48,463][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10013 tokens. +[2026-03-26 07:47:49,335][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.05%, Current % of VRAM taken: 60.48%, Block Peak % of device VRAM: 62.07%, ΔTime: 00:00:32 +[2026-03-26 07:47:50,122][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:47:50,125][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:47:50,126][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:47:50,855][__main__][INFO] - Iteration 684 took 50s (27.91% Gen, 70.63% Train). Generation: 13s, Training: 35s. Estimated remaining time: 31h 1m 53s. Estimated total time: 41h 44m 3s. Time estimates for 10 more iterations: 8m 20s, 100 more iterations: 1h 23m 28s, 500 more iterations: 6h 57m 20s. +[2026-03-26 07:47:50,858][__main__][INFO] - Starting iteration 684. +[2026-03-26 07:47:51,272][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 68 and human policies 1. +[2026-03-26 07:47:51,272][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:48:04,442][__main__][INFO] - Number of regex retries in iteration 684: 0 +[2026-03-26 07:48:04,443][__main__][INFO] - agents played in iteration 684 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:48:05,219][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:48:05,243][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:48:05,267][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:48:05,291][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:48:05,291][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:48:05,292][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:48:06,147][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:48:06,603][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:48:07,112][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:48:07,618][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:48:08,117][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:48:08,616][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:48:09,116][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:48:09,620][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:48:10,125][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:48:10,638][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:48:11,146][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:48:11,649][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:48:12,152][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:48:12,658][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:48:13,161][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:48:13,662][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:48:14,164][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:48:14,662][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:48:15,162][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:48:15,662][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:48:16,165][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:48:16,664][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:48:17,163][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:48:17,662][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:48:18,163][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:48:18,662][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:48:19,164][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:48:19,669][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:48:20,174][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:48:20,679][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:48:21,180][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:48:21,681][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:48:22,181][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:48:22,681][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:48:23,181][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:48:23,682][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:48:24,181][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:48:24,679][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:48:25,179][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:48:25,681][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:48:26,179][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:48:26,678][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:48:27,176][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:48:27,678][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:48:28,180][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:48:28,677][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:48:29,175][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:48:29,677][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:48:30,175][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:48:30,676][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:48:31,176][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:48:31,677][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:48:32,179][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:48:32,682][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:48:33,182][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:48:33,681][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:48:34,181][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:48:34,679][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:48:35,177][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:48:35,675][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:48:36,174][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:48:36,675][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:48:37,173][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:48:37,671][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:48:38,174][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10062 tokens. +[2026-03-26 07:48:39,058][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 60.63%, Block Peak % of device VRAM: 62.08%, ΔTime: 00:00:32 +[2026-03-26 07:48:39,815][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:48:39,817][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:48:39,819][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:48:40,553][__main__][INFO] - Iteration 685 took 49s (26.72% Gen, 71.78% Train). Generation: 13s, Training: 35s. Estimated remaining time: 30h 21m 5s. Estimated total time: 41h 4m 5s. Time estimates for 10 more iterations: 8m 12s, 100 more iterations: 1h 22m 8s, 500 more iterations: 6h 50m 40s. +[2026-03-26 07:48:40,555][__main__][INFO] - Starting iteration 685. +[2026-03-26 07:48:40,963][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 68 and human policies 1. +[2026-03-26 07:48:40,964][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:48:55,913][__main__][INFO] - Number of regex retries in iteration 685: 0 +[2026-03-26 07:48:55,914][__main__][INFO] - agents played in iteration 685 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:48:56,697][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:48:56,721][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:48:56,746][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:48:56,771][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:48:56,771][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:48:56,772][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:48:57,613][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:48:58,079][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:48:58,593][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:48:59,096][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:48:59,600][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:49:00,102][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:49:00,605][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:49:01,106][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:49:01,606][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:49:02,109][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:49:02,609][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:49:03,108][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:49:03,608][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:49:04,113][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:49:04,624][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:49:05,125][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:49:05,626][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:49:06,126][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:49:06,626][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:49:07,126][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:49:07,628][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:49:08,142][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:49:08,644][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:49:09,149][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:49:09,650][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:49:10,151][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:49:10,656][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:49:11,157][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:49:11,661][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:49:12,170][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:49:12,678][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:49:13,178][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:49:13,681][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:49:14,182][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:49:14,682][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:49:15,183][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:49:15,683][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:49:16,183][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:49:16,685][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:49:17,200][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:49:17,705][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:49:18,205][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:49:18,708][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:49:19,209][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:49:19,710][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:49:20,209][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:49:20,710][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:49:21,213][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:49:21,715][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:49:22,221][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:49:22,728][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:49:23,234][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:49:23,739][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:49:24,240][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:49:24,744][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:49:25,248][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:49:25,748][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:49:26,245][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:49:26,744][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:49:27,241][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:49:27,743][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:49:28,241][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:49:28,739][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:49:29,237][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:49:29,734][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10162 tokens. +[2026-03-26 07:49:30,613][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 60.51%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:33 +[2026-03-26 07:49:31,363][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:49:31,365][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:49:31,367][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:49:32,085][__main__][INFO] - Iteration 686 took 51s (29.24% Gen, 69.35% Train). Generation: 14s, Training: 35s. Estimated remaining time: 31h 52m 15s. Estimated total time: 42h 36m 7s. Time estimates for 10 more iterations: 8m 31s, 100 more iterations: 1h 25m 12s, 500 more iterations: 7h 6m 1s. +[2026-03-26 07:49:32,087][__main__][INFO] - Starting iteration 686. +[2026-03-26 07:49:32,502][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 68 and human policies 1. +[2026-03-26 07:49:32,503][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:49:44,967][__main__][INFO] - Number of regex retries in iteration 686: 0 +[2026-03-26 07:49:44,968][__main__][INFO] - agents played in iteration 686 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:49:45,737][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:49:45,762][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:49:45,787][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:49:45,812][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:49:45,815][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:49:45,816][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:49:46,660][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:49:47,117][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:49:47,622][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:49:48,122][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:49:48,620][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:49:49,119][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:49:49,617][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:49:50,116][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:49:50,614][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:49:51,114][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:49:51,613][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:49:52,112][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:49:52,612][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:49:53,113][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:49:53,619][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:49:54,123][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:49:54,622][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:49:55,124][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:49:55,625][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:49:56,125][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:49:56,625][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:49:57,123][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:49:57,623][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:49:58,122][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:49:58,621][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:49:59,119][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:49:59,616][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:50:00,139][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:50:00,641][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:50:01,138][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:50:01,635][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:50:02,137][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:50:02,645][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:50:03,145][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:50:03,649][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:50:04,154][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:50:04,654][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:50:05,155][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:50:05,656][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:50:06,156][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:50:06,656][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:50:07,156][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:50:07,656][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:50:08,156][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:50:08,656][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:50:09,156][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:50:09,656][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:50:10,155][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:50:10,655][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:50:11,152][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:50:11,653][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:50:12,150][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:50:12,647][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:50:13,146][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:50:13,643][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:50:14,146][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:50:14,650][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:50:15,151][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:50:15,656][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:50:16,161][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:50:16,664][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:50:17,166][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:50:17,667][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:50:18,167][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:50:18,668][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9949 tokens. +[2026-03-26 07:50:19,547][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 60.52%, Block Peak % of device VRAM: 62.09%, ΔTime: 00:00:32 +[2026-03-26 07:50:20,304][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:50:20,307][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:50:20,309][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:50:21,043][__main__][INFO] - Iteration 687 took 48s (25.68% Gen, 72.80% Train). Generation: 12s, Training: 35s. Estimated remaining time: 29h 42m 23s. Estimated total time: 40h 27m 4s. Time estimates for 10 more iterations: 8m 5s, 100 more iterations: 1h 20m 54s, 500 more iterations: 6h 44m 30s. +[2026-03-26 07:50:21,045][__main__][INFO] - Starting iteration 687. +[2026-03-26 07:50:21,459][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 68 and human policies 1. +[2026-03-26 07:50:21,460][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:50:37,500][__main__][INFO] - Number of regex retries in iteration 687: 0 +[2026-03-26 07:50:37,501][__main__][INFO] - agents played in iteration 687 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:50:38,267][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:50:38,291][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:50:38,315][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:50:38,338][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:50:38,339][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:50:38,339][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:50:39,160][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:50:39,616][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:50:40,117][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:50:40,613][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:50:41,110][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:50:41,609][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:50:42,106][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:50:42,603][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:50:43,098][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:50:43,597][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:50:44,100][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:50:44,598][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:50:45,095][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:50:45,593][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:50:46,090][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:50:46,597][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:50:47,096][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:50:47,599][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:50:48,098][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:50:48,601][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:50:49,101][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:50:49,600][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:50:50,100][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:50:50,599][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:50:51,098][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:50:51,600][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:50:52,106][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:50:52,608][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:50:53,110][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:50:53,610][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:50:54,112][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:50:54,613][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:50:55,113][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:50:55,615][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:50:56,115][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:50:56,615][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:50:57,115][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:50:57,615][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:50:58,115][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:50:58,617][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:50:59,117][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:50:59,615][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:51:00,112][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:51:00,608][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:51:01,105][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:51:01,601][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:51:02,100][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:51:02,598][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:51:03,099][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:51:03,601][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:51:04,103][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:51:04,610][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:51:05,113][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:51:05,627][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:51:06,132][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:51:06,637][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:51:07,164][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:51:07,666][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:51:08,165][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:51:08,667][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:51:09,166][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:51:09,666][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:51:10,168][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:51:10,670][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:51:11,169][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10090 tokens. +[2026-03-26 07:51:12,037][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.42%, ΔTime: 00:00:32 +[2026-03-26 07:51:12,792][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:51:12,794][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:51:12,796][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:51:13,524][__main__][INFO] - Iteration 688 took 52s (30.81% Gen, 67.79% Train). Generation: 16s, Training: 35s. Estimated remaining time: 32h 37m 42s. Estimated total time: 43h 23m 15s. Time estimates for 10 more iterations: 8m 40s, 100 more iterations: 1h 26m 46s, 500 more iterations: 7h 13m 52s. +[2026-03-26 07:51:13,526][__main__][INFO] - Starting iteration 688. +[2026-03-26 07:51:13,934][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 68 and human policies 1. +[2026-03-26 07:51:13,935][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:51:23,963][mllm.models.large_language_model_local][WARNING] - Response Proposal: x hats, y books, 10 balls -By taking all 10 hats, I ensure that I receive the maximum value per item from this round. Alice values books more highly, so she is likely to take all of those, and balls will be divided between us since both of us value them less. Therefore, proposing to take all 10 hats will secure the highest points for me in this round. did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 02:48:25,502][__main__][INFO] - Number of regex retries in iteration 545: 1 -[2025-11-10 02:48:25,503][__main__][INFO] - agents played in iteration 545 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:48:26,338][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:48:26,365][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:48:26,390][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:48:26,412][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:48:26,413][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:48:26,414][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:48:27,153][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:48:27,607][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:48:28,113][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:48:28,632][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:48:29,133][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:48:29,636][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:48:30,135][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:48:30,637][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:48:31,146][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:48:31,648][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:48:32,155][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:48:32,667][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:48:33,170][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:48:33,672][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:48:34,177][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:48:34,680][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:48:35,185][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:48:35,687][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:48:36,189][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:48:36,690][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:48:37,192][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:48:37,695][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:48:38,197][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:48:38,701][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:48:39,218][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:48:39,723][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:48:40,237][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:48:40,740][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:48:41,247][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:48:41,755][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:48:42,255][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:48:42,759][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:48:43,263][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:48:43,765][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:48:44,268][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:48:44,765][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:48:45,265][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:48:45,763][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:48:46,263][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:48:46,758][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:48:47,258][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:48:47,756][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:48:48,259][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:48:48,754][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:48:49,250][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:48:49,746][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:48:50,242][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:48:50,745][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:48:51,242][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:48:51,739][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:48:52,253][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:48:52,751][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:48:53,250][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:48:53,753][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:48:54,255][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:48:54,753][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:48:55,251][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:48:55,751][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:48:56,262][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:48:56,763][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:48:57,265][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:48:57,763][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:48:58,263][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:48:58,767][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:48:59,268][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10088 tokens. -[2025-11-10 02:49:00,069][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 55.66%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 -[2025-11-10 02:49:00,799][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:49:00,800][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:49:00,802][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:49:01,724][__main__][INFO] - Iteration 546 took 1m 1s (41.43% Gen, 57.08% Train). Generation: 25s, Training: 35s. Estimated remaining time: 42h 42m 1s. Estimated total time: 51h 32m 6s. Time estimates for 10 more iterations: 10m 18s, 100 more iterations: 1h 43m 4s, 500 more iterations: 8h 35m 21s. -[2025-11-10 02:49:01,726][__main__][INFO] - Starting iteration 546. -[2025-11-10 02:49:02,187][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 54 and human policies 1. -[2025-11-10 02:49:02,187][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:49:21,269][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 02:49:29,600][__main__][INFO] - Number of regex retries in iteration 546: 1 -[2025-11-10 02:49:29,601][__main__][INFO] - agents played in iteration 546 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:49:30,578][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:49:30,629][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:49:30,663][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:49:30,694][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:49:30,694][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:49:30,695][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:49:31,514][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:49:31,972][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:49:32,480][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:49:32,982][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:49:33,484][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:49:33,985][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:49:34,485][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:49:34,989][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:49:35,492][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:49:35,999][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:49:36,502][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:49:37,004][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:49:37,510][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:49:38,025][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:49:38,529][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:49:39,029][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:49:39,538][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:49:40,045][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:49:40,553][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:49:41,055][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:49:41,555][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:49:42,058][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:49:42,562][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:49:43,066][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:49:43,574][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:49:44,074][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:49:44,576][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:49:45,077][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:49:45,578][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:49:46,080][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:49:46,581][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:49:47,086][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:49:47,593][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:49:48,092][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:49:48,596][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:49:49,096][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:49:49,599][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:49:50,103][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:49:50,604][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:49:51,110][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:49:51,610][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:49:52,114][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:49:52,634][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:49:53,135][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:49:53,650][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:49:54,148][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:49:54,652][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:49:55,150][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:49:55,646][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:49:56,144][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:49:56,641][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:49:57,141][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:49:57,658][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:49:58,154][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:49:58,650][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:49:59,147][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:49:59,645][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:50:00,151][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:50:00,649][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:50:01,145][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:50:01,658][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:50:02,159][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:50:02,658][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:50:03,157][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:50:03,657][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10031 tokens. -[2025-11-10 02:50:04,463][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 55.61%, Block Peak % of device VRAM: 62.17%, ΔTime: 00:00:32 -[2025-11-10 02:50:05,100][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:50:05,101][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:50:05,103][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:50:06,131][__main__][INFO] - Iteration 547 took 1m 3s (42.87% Gen, 55.52% Train). Generation: 27s, Training: 35s. Estimated remaining time: 44h 26m 6s. Estimated total time: 53h 17m 15s. Time estimates for 10 more iterations: 10m 39s, 100 more iterations: 1h 46m 34s, 500 more iterations: 8h 52m 52s. -[2025-11-10 02:50:06,133][__main__][INFO] - Starting iteration 547. -[2025-11-10 02:50:06,629][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 54 and human policies 1. -[2025-11-10 02:50:06,629][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:50:34,715][__main__][INFO] - Number of regex retries in iteration 547: 0 -[2025-11-10 02:50:34,715][__main__][INFO] - agents played in iteration 547 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:50:35,496][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:50:35,523][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:50:35,549][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:50:35,572][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:50:35,572][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:50:35,573][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:50:36,307][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:50:36,765][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:50:37,271][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:50:37,769][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:50:38,270][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:50:38,778][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:50:39,288][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:50:39,792][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:50:40,293][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:50:40,797][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:50:41,298][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:50:41,798][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:50:42,299][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:50:42,800][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:50:43,303][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:50:43,806][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:50:44,307][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:50:44,813][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:50:45,313][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:50:45,814][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:50:46,319][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:50:46,820][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:50:47,322][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:50:47,823][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:50:48,329][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:50:48,840][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:50:49,341][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:50:49,856][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:50:50,357][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:50:50,857][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:50:51,361][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:50:51,865][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:50:52,365][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:50:52,869][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:50:53,370][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:50:53,878][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:50:54,380][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:50:54,881][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:50:55,382][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:50:55,881][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:50:56,383][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:50:56,881][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:50:57,380][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:50:57,880][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:50:58,380][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:50:58,879][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:50:59,378][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:50:59,877][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:51:00,378][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:51:00,876][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:51:01,374][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:51:01,872][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:51:02,368][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:51:02,864][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:51:03,364][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:51:03,859][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:51:04,364][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:51:04,863][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:51:05,364][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:51:05,863][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:51:06,361][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:51:06,867][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:51:07,368][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:51:07,869][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:51:08,380][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10085 tokens. -[2025-11-10 02:51:09,163][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.25%, Current % of VRAM taken: 55.74%, Block Peak % of device VRAM: 62.21%, ΔTime: 00:00:32 -[2025-11-10 02:51:09,930][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:51:09,932][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:51:09,933][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:51:10,872][__main__][INFO] - Iteration 548 took 1m 4s (43.72% Gen, 54.82% Train). Generation: 28s, Training: 35s. Estimated remaining time: 44h 39m 58s. Estimated total time: 53h 32m 12s. Time estimates for 10 more iterations: 10m 42s, 100 more iterations: 1h 47m 4s, 500 more iterations: 8h 55m 22s. -[2025-11-10 02:51:10,874][__main__][INFO] - Starting iteration 548. -[2025-11-10 02:51:11,359][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 54 and human policies 1. -[2025-11-10 02:51:11,360][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:51:34,223][__main__][INFO] - Number of regex retries in iteration 548: 0 -[2025-11-10 02:51:34,224][__main__][INFO] - agents played in iteration 548 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:51:35,031][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:51:35,078][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:51:35,110][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:51:35,138][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:51:35,139][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:51:35,140][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:51:35,895][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:51:36,347][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:51:36,847][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:51:37,358][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:51:37,860][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:51:38,359][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:51:38,856][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:51:39,351][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:51:39,846][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:51:40,341][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:51:40,834][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:51:41,339][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:51:41,841][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:51:42,339][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:51:42,837][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:51:43,336][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:51:43,843][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:51:44,345][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:51:44,848][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:51:45,361][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:51:45,862][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:51:46,363][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:51:46,867][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:51:47,368][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:51:47,884][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:51:48,386][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:51:48,888][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:51:49,388][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:51:49,895][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:51:50,395][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:51:50,898][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:51:51,404][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:51:51,910][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:51:52,414][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:51:52,916][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:51:53,418][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:51:53,922][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:51:54,425][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:51:54,929][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:51:55,431][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:51:55,938][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:51:56,444][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:51:56,953][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:51:57,459][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:51:57,963][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:51:58,480][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:51:58,984][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:51:59,488][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:51:59,987][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:52:00,491][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:52:01,002][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:52:01,506][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:52:02,010][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:52:02,516][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:52:03,019][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:52:03,523][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:52:04,029][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:52:04,531][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:52:05,033][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:52:05,541][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:52:06,046][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:52:06,555][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:52:07,053][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:52:07,566][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:52:08,068][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10140 tokens. -[2025-11-10 02:52:08,893][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 55.57%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:33 -[2025-11-10 02:52:09,540][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:52:09,541][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:52:09,543][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:52:10,402][__main__][INFO] - Iteration 549 took 59s (38.72% Gen, 59.82% Train). Generation: 22s, Training: 35s. Estimated remaining time: 40h 18m 58s. Estimated total time: 49h 12m 12s. Time estimates for 10 more iterations: 9m 50s, 100 more iterations: 1h 38m 24s, 500 more iterations: 8h 12m 2s. -[2025-11-10 02:52:10,404][__main__][INFO] - Starting iteration 549. -[2025-11-10 02:52:10,867][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 54 and human policies 1. -[2025-11-10 02:52:10,867][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:52:39,647][__main__][INFO] - Number of regex retries in iteration 549: 0 -[2025-11-10 02:52:39,647][__main__][INFO] - agents played in iteration 549 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:52:40,421][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:52:40,448][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:52:40,474][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:52:40,497][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:52:40,498][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:52:40,498][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:52:41,252][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:52:41,718][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:52:42,218][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:52:42,718][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:52:43,214][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:52:43,711][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:52:44,211][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:52:44,712][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:52:45,208][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:52:45,708][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:52:46,213][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:52:46,711][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:52:47,208][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:52:47,737][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:52:48,235][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:52:48,736][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:52:49,238][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:52:49,740][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:52:50,245][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:52:50,748][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:52:51,252][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:52:51,758][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:52:52,276][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:52:52,781][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:52:53,294][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:52:53,800][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:52:54,308][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:52:54,817][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:52:55,321][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:52:55,828][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:52:56,335][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:52:56,841][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:52:57,352][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:52:57,858][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:52:58,360][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:52:58,874][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:52:59,375][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:52:59,899][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:53:00,403][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:53:00,907][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:53:01,414][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:53:01,916][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:53:02,424][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:53:02,925][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:53:03,429][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:53:03,933][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:53:04,439][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:53:04,940][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:53:05,444][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:53:05,950][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:53:06,450][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:53:06,949][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:53:07,450][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:53:07,969][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:53:08,474][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:53:08,979][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:53:09,483][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:53:09,985][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:53:10,487][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:53:10,985][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:53:11,484][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:53:11,997][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:53:12,497][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:53:13,001][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:53:13,499][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10262 tokens. -[2025-11-10 02:53:14,328][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 55.66%, Block Peak % of device VRAM: 62.40%, ΔTime: 00:00:33 -[2025-11-10 02:53:15,046][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:53:15,048][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:53:15,050][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:53:16,010][__main__][INFO] - Iteration 550 took 1m 5s (44.18% Gen, 54.34% Train). Generation: 28s, Training: 35s. Estimated remaining time: 45h 22m 52s. Estimated total time: 54h 17m 11s. Time estimates for 10 more iterations: 10m 51s, 100 more iterations: 1h 48m 34s, 500 more iterations: 9h 2m 51s. -[2025-11-10 02:53:16,012][__main__][INFO] - Starting iteration 550. -[2025-11-10 02:53:16,482][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 54 and human policies 1. -[2025-11-10 02:53:16,483][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:53:40,524][__main__][INFO] - Number of regex retries in iteration 550: 0 -[2025-11-10 02:53:40,526][__main__][INFO] - agents played in iteration 550 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:53:41,380][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:53:41,417][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:53:41,446][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:53:41,471][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:53:41,472][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:53:41,472][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:53:42,229][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:53:42,683][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:53:43,186][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:53:43,692][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:53:44,187][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:53:44,684][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:53:45,187][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:53:45,681][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:53:46,176][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:53:46,676][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:53:47,173][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:53:47,684][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:53:48,181][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:53:48,678][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:53:49,179][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:53:49,678][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:53:50,180][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:53:50,702][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:53:51,209][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:53:51,715][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:53:52,216][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:53:52,720][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:53:53,221][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:53:53,724][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:53:54,227][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:53:54,728][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:53:55,231][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:53:55,732][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:53:56,234][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:53:56,738][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:53:57,237][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:53:57,737][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:53:58,245][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:53:58,749][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:53:59,251][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:53:59,754][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:54:00,256][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:54:00,758][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:54:01,260][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:54:01,763][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:54:02,288][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:54:02,790][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:54:03,297][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:54:03,806][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:54:04,309][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:54:04,814][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:54:05,321][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:54:05,824][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:54:06,330][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:54:06,831][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:54:07,333][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:54:07,834][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:54:08,339][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:54:08,846][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:54:09,348][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:54:09,851][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:54:10,360][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:54:10,863][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:54:11,366][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:54:11,883][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:54:12,384][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:54:12,897][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:54:13,402][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:54:13,912][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:54:14,420][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9980 tokens. -[2025-11-10 02:54:15,334][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 55.57%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:33 -[2025-11-10 02:54:15,982][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:54:15,984][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:54:15,986][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:54:17,793][__main__][INFO] - Iteration 551 took 1m 1s (39.21% Gen, 57.84% Train). Generation: 24s, Training: 35s. Estimated remaining time: 42h 10m 15s. Estimated total time: 51h 5m 36s. Time estimates for 10 more iterations: 10m 13s, 100 more iterations: 1h 42m 11s, 500 more iterations: 8h 30m 56s. -[2025-11-10 02:54:17,796][__main__][INFO] - Starting iteration 551. -[2025-11-10 02:54:18,283][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 55 and human policies 1. -[2025-11-10 02:54:18,283][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:54:44,105][__main__][INFO] - Number of regex retries in iteration 551: 0 -[2025-11-10 02:54:44,106][__main__][INFO] - agents played in iteration 551 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:54:44,873][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:54:44,899][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:54:44,925][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:54:44,948][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:54:44,949][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:54:44,950][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:54:45,710][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:54:46,165][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:54:46,666][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:54:47,176][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:54:47,679][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:54:48,177][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:54:48,699][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:54:49,198][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:54:49,696][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:54:50,199][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:54:50,697][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:54:51,204][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:54:51,702][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:54:52,199][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:54:52,710][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:54:53,207][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:54:53,705][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:54:54,207][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:54:54,705][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:54:55,215][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:54:55,714][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:54:56,214][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:54:56,712][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:54:57,210][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:54:57,718][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:54:58,218][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:54:58,715][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:54:59,217][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:54:59,717][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:55:00,217][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:55:00,724][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:55:01,227][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:55:01,727][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:55:02,228][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:55:02,728][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:55:03,230][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:55:03,733][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:55:04,235][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:55:04,741][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:55:05,243][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:55:05,744][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:55:06,250][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:55:06,776][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:55:07,286][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:55:07,792][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:55:08,294][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:55:08,801][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:55:09,307][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:55:09,822][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:55:10,324][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:55:10,826][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:55:11,328][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:55:11,831][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:55:12,335][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:55:12,842][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:55:13,344][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:55:13,848][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:55:14,351][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:55:14,853][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:55:15,356][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:55:15,858][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:55:16,364][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:55:16,864][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:55:17,365][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:55:17,866][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10068 tokens. -[2025-11-10 02:55:18,739][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 55.63%, Block Peak % of device VRAM: 62.27%, ΔTime: 00:00:33 -[2025-11-10 02:55:19,474][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:55:19,476][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:55:19,478][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:55:20,471][__main__][INFO] - Iteration 552 took 1m 2s (41.52% Gen, 56.88% Train). Generation: 25s, Training: 35s. Estimated remaining time: 42h 53m 3s. Estimated total time: 51h 49m 26s. Time estimates for 10 more iterations: 10m 21s, 100 more iterations: 1h 43m 38s, 500 more iterations: 8h 38m 14s. -[2025-11-10 02:55:20,473][__main__][INFO] - Starting iteration 552. -[2025-11-10 02:55:20,938][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 55 and human policies 1. -[2025-11-10 02:55:20,938][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:55:47,936][__main__][INFO] - Number of regex retries in iteration 552: 0 -[2025-11-10 02:55:47,938][__main__][INFO] - agents played in iteration 552 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:55:48,755][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:55:48,806][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:55:48,839][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:55:48,869][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:55:48,869][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:55:48,870][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:55:49,624][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:55:50,092][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:55:50,594][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:55:51,091][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:55:51,596][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:55:52,092][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:55:52,603][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:55:53,099][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:55:53,595][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:55:54,099][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:55:54,599][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:55:55,098][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:55:55,596][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:55:56,097][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:55:56,606][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:55:57,105][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:55:57,600][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:55:58,099][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:55:58,597][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:55:59,117][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:55:59,612][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:56:00,116][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:56:00,624][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:56:01,119][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:56:01,618][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:56:02,116][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:56:02,616][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:56:03,128][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:56:03,629][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:56:04,129][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:56:04,627][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:56:05,127][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:56:05,627][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:56:06,153][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:56:06,658][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:56:07,164][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:56:07,671][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:56:08,174][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:56:08,680][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:56:09,185][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:56:09,701][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:56:10,203][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:56:10,704][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:56:11,212][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:56:11,713][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:56:12,224][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:56:12,724][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:56:13,225][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:56:13,729][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:56:14,230][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:56:14,732][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:56:15,233][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:56:15,735][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:56:16,240][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:56:16,742][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:56:17,243][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:56:17,746][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:56:18,247][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:56:18,748][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:56:19,249][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:56:19,749][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:56:20,255][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:56:20,755][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:56:21,261][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:56:21,766][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10090 tokens. -[2025-11-10 02:56:22,690][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 55.59%, Block Peak % of device VRAM: 62.10%, ΔTime: 00:00:33 -[2025-11-10 02:56:23,340][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:56:23,342][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:56:23,344][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:56:24,226][__main__][INFO] - Iteration 553 took 1m 3s (42.66% Gen, 55.94% Train). Generation: 26s, Training: 35s. Estimated remaining time: 43h 46m 59s. Estimated total time: 52h 44m 26s. Time estimates for 10 more iterations: 10m 32s, 100 more iterations: 1h 45m 28s, 500 more iterations: 8h 47m 24s. -[2025-11-10 02:56:24,228][__main__][INFO] - Starting iteration 553. -[2025-11-10 02:56:24,692][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 55 and human policies 1. -[2025-11-10 02:56:24,693][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:56:53,523][__main__][INFO] - Number of regex retries in iteration 553: 0 -[2025-11-10 02:56:53,524][__main__][INFO] - agents played in iteration 553 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:56:54,288][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:56:54,315][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:56:54,352][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:56:54,377][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:56:54,378][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:56:54,379][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:56:55,127][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:56:55,585][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:56:56,103][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:56:56,601][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:56:57,112][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:56:57,610][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:56:58,108][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:56:58,624][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:56:59,123][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:56:59,622][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:57:00,125][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:57:00,625][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:57:01,132][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:57:01,630][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:57:02,129][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:57:02,637][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:57:03,137][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:57:03,634][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:57:04,134][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:57:04,629][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:57:05,137][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:57:05,633][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:57:06,129][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:57:06,637][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:57:07,133][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:57:07,631][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:57:08,130][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:57:08,627][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:57:09,127][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:57:09,628][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:57:10,130][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:57:10,642][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:57:11,146][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:57:11,644][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:57:12,146][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:57:12,651][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:57:13,153][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:57:13,654][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:57:14,154][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:57:14,657][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:57:15,158][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:57:15,670][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:57:16,172][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:57:16,676][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:57:17,185][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:57:17,687][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:57:18,190][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:57:18,694][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:57:19,196][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:57:19,713][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:57:20,215][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:57:20,722][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:57:21,237][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:57:21,743][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:57:22,246][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:57:22,751][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:57:23,253][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:57:23,757][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:57:24,257][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:57:24,763][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:57:25,270][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:57:25,770][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:57:26,270][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:57:26,773][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:57:27,273][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10071 tokens. -[2025-11-10 02:57:28,170][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 55.56%, Block Peak % of device VRAM: 62.34%, ΔTime: 00:00:33 -[2025-11-10 02:57:28,923][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:57:28,925][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:57:28,926][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:57:29,941][__main__][INFO] - Iteration 554 took 1m 5s (44.18% Gen, 54.26% Train). Generation: 28s, Training: 35s. Estimated remaining time: 45h 23m 55s. Estimated total time: 54h 22m 28s. Time estimates for 10 more iterations: 10m 52s, 100 more iterations: 1h 48m 44s, 500 more iterations: 9h 3m 44s. -[2025-11-10 02:57:29,944][__main__][INFO] - Starting iteration 554. -[2025-11-10 02:57:30,428][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 55 and human policies 1. -[2025-11-10 02:57:30,429][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:57:53,518][__main__][INFO] - Number of regex retries in iteration 554: 0 -[2025-11-10 02:57:53,520][__main__][INFO] - agents played in iteration 554 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:57:54,285][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:57:54,313][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:57:54,339][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:57:54,362][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:57:54,363][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:57:54,364][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:57:55,102][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:57:55,558][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:57:56,059][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:57:56,557][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:57:57,053][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:57:57,550][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:57:58,047][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:57:58,543][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:57:59,041][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:57:59,537][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:58:00,031][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:58:00,532][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:58:01,033][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:58:01,531][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:58:02,030][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:58:02,531][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:58:03,035][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:58:03,532][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:58:04,026][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:58:04,525][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:58:05,021][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:58:05,517][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:58:06,012][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:58:06,507][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:58:07,015][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:58:07,513][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:58:08,010][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:58:08,517][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:58:09,016][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:58:09,513][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:58:10,014][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:58:10,512][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:58:11,033][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:58:11,531][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:58:12,029][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:58:12,562][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:58:13,059][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:58:13,559][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:58:14,065][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:58:14,565][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:58:15,070][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:58:15,575][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:58:16,078][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:58:16,585][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:58:17,090][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:58:17,596][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:58:18,103][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:58:18,614][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:58:19,121][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:58:19,623][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:58:20,146][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:58:20,651][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:58:21,154][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:58:21,655][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:58:22,159][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:58:22,662][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:58:23,165][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:58:23,670][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:58:24,176][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:58:24,677][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:58:25,181][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:58:25,685][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:58:26,185][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:58:26,688][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:58:27,188][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10127 tokens. -[2025-11-10 02:58:28,093][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 55.60%, Block Peak % of device VRAM: 62.48%, ΔTime: 00:00:32 -[2025-11-10 02:58:28,745][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:58:28,747][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:58:28,749][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:58:29,608][__main__][INFO] - Iteration 555 took 59s (39.01% Gen, 59.53% Train). Generation: 23s, Training: 35s. Estimated remaining time: 40h 19m 30s. Estimated total time: 49h 19m 3s. Time estimates for 10 more iterations: 9m 51s, 100 more iterations: 1h 38m 38s, 500 more iterations: 8h 13m 10s. -[2025-11-10 02:58:29,611][__main__][INFO] - Starting iteration 555. -[2025-11-10 02:58:30,074][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 55 and human policies 1. -[2025-11-10 02:58:30,074][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:58:53,121][__main__][INFO] - Number of regex retries in iteration 555: 0 -[2025-11-10 02:58:53,122][__main__][INFO] - agents played in iteration 555 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:58:54,012][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:58:54,037][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:58:54,062][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:58:54,085][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:58:54,085][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:58:54,087][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:58:54,856][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:58:55,330][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:58:55,831][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:58:56,330][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:58:56,831][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:58:57,331][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:58:57,838][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:58:58,337][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:58:58,837][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 02:58:59,336][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 02:58:59,833][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 02:59:00,331][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 02:59:00,828][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 02:59:01,330][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 02:59:01,849][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 02:59:02,346][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 02:59:02,843][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 02:59:03,340][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 02:59:03,837][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 02:59:04,332][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 02:59:04,830][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 02:59:05,324][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 02:59:05,836][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 02:59:06,330][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 02:59:06,830][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 02:59:07,332][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 02:59:07,829][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 02:59:08,340][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 02:59:08,836][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 02:59:09,334][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 02:59:09,832][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 02:59:10,329][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 02:59:10,827][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 02:59:11,326][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 02:59:11,831][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 02:59:12,330][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 02:59:12,826][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 02:59:13,324][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 02:59:13,824][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 02:59:14,321][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 02:59:14,821][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 02:59:15,320][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 02:59:15,817][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 02:59:16,317][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 02:59:16,819][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 02:59:17,330][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 02:59:17,835][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 02:59:18,337][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 02:59:18,854][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 02:59:19,353][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 02:59:19,855][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 02:59:20,359][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 02:59:20,861][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 02:59:21,365][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 02:59:21,866][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 02:59:22,371][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 02:59:22,878][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 02:59:23,380][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 02:59:23,880][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 02:59:24,381][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 02:59:24,881][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 02:59:25,387][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 02:59:25,887][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 02:59:26,389][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 02:59:26,891][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10093 tokens. -[2025-11-10 02:59:27,768][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 55.57%, Block Peak % of device VRAM: 62.27%, ΔTime: 00:00:32 -[2025-11-10 02:59:28,540][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 02:59:28,542][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 02:59:28,544][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 02:59:29,533][__main__][INFO] - Iteration 556 took 59s (38.76% Gen, 59.57% Train). Generation: 23s, Training: 35s. Estimated remaining time: 40h 32m 28s. Estimated total time: 49h 33m 0s. Time estimates for 10 more iterations: 9m 54s, 100 more iterations: 1h 39m 6s, 500 more iterations: 8h 15m 30s. -[2025-11-10 02:59:29,535][__main__][INFO] - Starting iteration 556. -[2025-11-10 02:59:30,083][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 55 and human policies 1. -[2025-11-10 02:59:30,084][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 02:59:54,139][__main__][INFO] - Number of regex retries in iteration 556: 0 -[2025-11-10 02:59:54,141][__main__][INFO] - agents played in iteration 556 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 02:59:54,937][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:59:54,965][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:59:54,992][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:59:55,016][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 02:59:55,016][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 02:59:55,017][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 02:59:55,774][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 02:59:56,232][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 02:59:56,733][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 02:59:57,240][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 02:59:57,737][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 02:59:58,233][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 02:59:58,735][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 02:59:59,230][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 02:59:59,728][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:00:00,222][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:00:00,720][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:00:01,217][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:00:01,713][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:00:02,210][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:00:02,705][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:00:03,200][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:00:03,700][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:00:04,194][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:00:04,688][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:00:05,182][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:00:05,681][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:00:06,180][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:00:06,676][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:00:07,170][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:00:07,676][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:00:08,170][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:00:08,667][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:00:09,162][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:00:09,656][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:00:10,171][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:00:10,665][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:00:11,161][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:00:11,656][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:00:12,152][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:00:12,661][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:00:13,159][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:00:13,657][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:00:14,169][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:00:14,666][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:00:15,163][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:00:15,658][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:00:16,153][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:00:16,666][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:00:17,159][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:00:17,657][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:00:18,152][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:00:18,646][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:00:19,152][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:00:19,645][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:00:20,140][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:00:20,645][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:00:21,146][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:00:21,646][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:00:22,155][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:00:22,654][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:00:23,160][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:00:23,662][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:00:24,162][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:00:24,666][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:00:25,165][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:00:25,665][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:00:26,164][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:00:26,667][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:00:27,183][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:00:27,683][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10037 tokens. -[2025-11-10 03:00:28,586][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.09%, Current % of VRAM taken: 55.59%, Block Peak % of device VRAM: 62.23%, ΔTime: 00:00:32 -[2025-11-10 03:00:29,222][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:00:29,224][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:00:29,226][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:00:30,160][__main__][INFO] - Iteration 557 took 1m 0s (40.04% Gen, 58.40% Train). Generation: 24s, Training: 35s. Estimated remaining time: 41h 2m 22s. Estimated total time: 50h 3m 55s. Time estimates for 10 more iterations: 10m 0s, 100 more iterations: 1h 40m 7s, 500 more iterations: 8h 20m 39s. -[2025-11-10 03:00:30,163][__main__][INFO] - Starting iteration 557. -[2025-11-10 03:00:30,625][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 55 and human policies 1. -[2025-11-10 03:00:30,626][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:00:48,939][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 03:01:00,008][__main__][INFO] - Number of regex retries in iteration 557: 1 -[2025-11-10 03:01:00,008][__main__][INFO] - agents played in iteration 557 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:01:00,774][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:01:00,802][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:01:00,828][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:01:00,851][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:01:00,852][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:01:00,853][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:01:01,594][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:01:02,047][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:01:02,577][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:01:03,072][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:01:03,573][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:01:04,069][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:01:04,564][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:01:05,071][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:01:05,567][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:01:06,062][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:01:06,557][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:01:07,052][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:01:07,546][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:01:08,042][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:01:08,538][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:01:09,037][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:01:09,532][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:01:10,029][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:01:10,528][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:01:11,024][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:01:11,519][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:01:12,019][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:01:12,514][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:01:13,011][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:01:13,510][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:01:14,006][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:01:14,503][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:01:15,003][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:01:15,504][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:01:16,004][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:01:16,502][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:01:17,001][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:01:17,499][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:01:17,998][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:01:18,494][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:01:18,989][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:01:19,495][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:01:19,990][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:01:20,494][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:01:20,996][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:01:21,492][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:01:22,000][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:01:22,501][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:01:22,999][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:01:23,496][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:01:23,997][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:01:24,498][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:01:24,996][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:01:25,497][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:01:26,005][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:01:26,505][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:01:27,006][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:01:27,510][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:01:28,007][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:01:28,514][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:01:29,023][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:01:29,526][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:01:30,029][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:01:30,531][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:01:31,040][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:01:31,554][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:01:32,056][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:01:32,570][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:01:33,071][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:01:33,585][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10124 tokens. -[2025-11-10 03:01:34,521][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 55.56%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:32 -[2025-11-10 03:01:35,282][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:01:35,284][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:01:35,285][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:01:36,287][__main__][INFO] - Iteration 558 took 1m 5s (44.75% Gen, 53.72% Train). Generation: 29s, Training: 35s. Estimated remaining time: 45h 40m 29s. Estimated total time: 54h 43m 8s. Time estimates for 10 more iterations: 10m 56s, 100 more iterations: 1h 49m 26s, 500 more iterations: 9h 7m 11s. -[2025-11-10 03:01:36,290][__main__][INFO] - Starting iteration 558. -[2025-11-10 03:01:36,753][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 55 and human policies 1. -[2025-11-10 03:01:36,753][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:01:53,584][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 03:01:59,239][__main__][INFO] - Number of regex retries in iteration 558: 1 -[2025-11-10 03:01:59,240][__main__][INFO] - agents played in iteration 558 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:02:00,024][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:02:00,051][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:02:00,077][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:02:00,101][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:02:00,101][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:02:00,102][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:02:00,843][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:02:01,303][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:02:01,802][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:02:02,299][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:02:02,796][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:02:03,294][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:02:03,800][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:02:04,296][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:02:04,793][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:02:05,297][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:02:05,798][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:02:06,301][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:02:06,800][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:02:07,300][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:02:07,798][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:02:08,299][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:02:08,800][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:02:09,298][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:02:09,795][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:02:10,307][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:02:10,803][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:02:11,303][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:02:11,799][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:02:12,297][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:02:12,801][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:02:13,299][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:02:13,796][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:02:14,303][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:02:14,803][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:02:15,300][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:02:15,794][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:02:16,289][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:02:16,802][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:02:17,299][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:02:17,794][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:02:18,292][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:02:18,787][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:02:19,290][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:02:19,784][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:02:20,278][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:02:20,789][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:02:21,287][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:02:21,782][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:02:22,298][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:02:22,793][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:02:23,288][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:02:23,784][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:02:24,279][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:02:24,801][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:02:25,301][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:02:25,802][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:02:26,303][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:02:26,802][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:02:27,305][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:02:27,808][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:02:28,308][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:02:28,810][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:02:29,311][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:02:29,817][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:02:30,325][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:02:30,829][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:02:31,334][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:02:31,835][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:02:32,342][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:02:32,852][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10044 tokens. -[2025-11-10 03:02:33,760][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.29%, Current % of VRAM taken: 55.79%, Block Peak % of device VRAM: 62.48%, ΔTime: 00:00:32 -[2025-11-10 03:02:35,748][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:02:35,750][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:02:35,752][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:02:36,647][__main__][INFO] - Iteration 559 took 59s (37.54% Gen, 60.96% Train). Generation: 22s, Training: 36s. Estimated remaining time: 40h 51m 6s. Estimated total time: 49h 54m 45s. Time estimates for 10 more iterations: 9m 58s, 100 more iterations: 1h 39m 49s, 500 more iterations: 8h 19m 7s. -[2025-11-10 03:02:36,652][__main__][INFO] - Starting iteration 559. -[2025-11-10 03:02:37,130][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 55 and human policies 1. -[2025-11-10 03:02:37,131][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:03:07,915][__main__][INFO] - Number of regex retries in iteration 559: 0 -[2025-11-10 03:03:07,916][__main__][INFO] - agents played in iteration 559 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:03:08,684][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.67%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:03:08,708][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.67%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:03:08,734][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.67%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:03:08,756][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.67%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:03:08,756][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:03:08,757][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:03:09,533][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:03:09,991][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:03:10,492][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:03:11,005][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:03:11,501][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:03:11,999][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:03:12,501][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:03:13,000][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:03:13,506][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:03:14,011][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:03:14,507][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:03:15,013][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:03:15,510][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:03:16,006][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:03:16,509][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:03:17,011][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:03:17,511][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:03:18,008][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:03:18,539][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:03:19,037][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:03:19,539][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:03:20,036][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:03:20,532][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:03:21,028][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:03:21,528][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:03:22,033][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:03:22,533][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:03:23,033][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:03:23,534][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:03:24,032][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:03:24,535][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:03:25,040][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:03:25,548][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:03:26,055][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:03:26,555][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:03:27,056][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:03:27,553][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:03:28,057][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:03:28,559][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:03:29,064][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:03:29,576][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:03:30,082][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:03:30,591][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:03:31,099][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:03:31,602][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:03:32,111][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:03:32,614][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:03:33,118][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:03:33,623][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:03:34,125][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:03:34,630][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:03:35,140][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:03:35,644][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:03:36,152][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:03:36,654][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:03:37,158][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:03:37,662][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:03:38,165][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:03:38,669][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:03:39,176][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:03:39,680][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:03:40,193][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:03:40,708][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:03:41,223][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:03:41,731][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10128 tokens. -[2025-11-10 03:03:42,610][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.27%, Current % of VRAM taken: 55.77%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:33 -[2025-11-10 03:03:43,380][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:03:43,381][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:03:43,383][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:03:44,408][__main__][INFO] - Iteration 560 took 1m 7s (45.76% Gen, 52.72% Train). Generation: 30s, Training: 35s. Estimated remaining time: 46h 59m 8s. Estimated total time: 56h 3m 56s. Time estimates for 10 more iterations: 11m 12s, 100 more iterations: 1h 52m 7s, 500 more iterations: 9h 20m 39s. -[2025-11-10 03:03:44,410][__main__][INFO] - Starting iteration 560. -[2025-11-10 03:03:44,884][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 55 and human policies 1. -[2025-11-10 03:03:44,885][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:03:59,213][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 03:04:00,188][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 03:04:07,923][__main__][INFO] - Number of regex retries in iteration 560: 2 -[2025-11-10 03:04:07,924][__main__][INFO] - agents played in iteration 560 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:04:08,702][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:04:08,730][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:04:08,756][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:04:08,780][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:04:08,781][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:04:08,782][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:04:09,525][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:04:09,979][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:04:10,494][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:04:10,991][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:04:11,504][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:04:12,001][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:04:12,500][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:04:13,005][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:04:13,507][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:04:14,015][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:04:14,512][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:04:15,008][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:04:15,523][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:04:16,024][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:04:16,523][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:04:17,020][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:04:17,518][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:04:18,024][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:04:18,529][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:04:19,028][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:04:19,529][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:04:20,048][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:04:20,544][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:04:21,047][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:04:21,545][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:04:22,041][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:04:22,541][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:04:23,037][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:04:23,535][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:04:24,030][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:04:24,525][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:04:25,020][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:04:25,516][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:04:26,023][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:04:26,528][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:04:27,029][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:04:27,539][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:04:28,041][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:04:28,546][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:04:29,043][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:04:29,547][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:04:30,047][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:04:30,546][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:04:31,046][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:04:31,550][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:04:32,053][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:04:32,557][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:04:33,056][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:04:33,557][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:04:34,064][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:04:34,566][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:04:35,068][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:04:35,573][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:04:36,081][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:04:36,876][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:04:37,550][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:04:38,091][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:04:38,593][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:04:39,094][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:04:39,597][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:04:40,100][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:04:40,624][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:04:41,126][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:04:41,633][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:04:42,133][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10084 tokens. -[2025-11-10 03:04:43,060][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 55.62%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:33 -[2025-11-10 03:04:43,715][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:04:43,717][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:04:43,719][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:04:45,493][__main__][INFO] - Iteration 561 took 1m 0s (38.01% Gen, 59.06% Train). Generation: 23s, Training: 35s. Estimated remaining time: 41h 24m 40s. Estimated total time: 50h 30m 29s. Time estimates for 10 more iterations: 10m 6s, 100 more iterations: 1h 41m 0s, 500 more iterations: 8h 25m 4s. -[2025-11-10 03:04:45,497][__main__][INFO] - Starting iteration 561. -[2025-11-10 03:04:45,976][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 56 and human policies 1. -[2025-11-10 03:04:45,976][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:05:14,395][__main__][INFO] - Number of regex retries in iteration 561: 0 -[2025-11-10 03:05:14,396][__main__][INFO] - agents played in iteration 561 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:05:15,211][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:05:15,234][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:05:15,257][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:05:15,279][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:05:15,279][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:05:15,280][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:05:16,033][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:05:16,488][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:05:16,989][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:05:17,486][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:05:17,981][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:05:18,488][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:05:18,982][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:05:19,477][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:05:19,979][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:05:20,475][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:05:20,971][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:05:21,468][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:05:21,963][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:05:22,464][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:05:22,959][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:05:23,453][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:05:23,951][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:05:24,445][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:05:24,941][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:05:25,435][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:05:25,933][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:05:26,440][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:05:26,939][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:05:27,437][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:05:27,940][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:05:28,434][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:05:28,935][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:05:29,434][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:05:29,935][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:05:30,443][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:05:30,943][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:05:31,442][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:05:31,942][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:05:32,442][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:05:32,949][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:05:33,452][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:05:33,955][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:05:34,461][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:05:34,961][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:05:35,463][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:05:35,965][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:05:36,468][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:05:36,974][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:05:37,480][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:05:37,981][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:05:38,488][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:05:38,990][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:05:39,490][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:05:39,991][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:05:40,495][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:05:41,005][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:05:41,509][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:05:42,009][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:05:42,517][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:05:43,017][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:05:43,528][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:05:44,031][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:05:44,529][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:05:45,030][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:05:45,532][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:05:46,043][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:05:46,544][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:05:47,045][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:05:47,547][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:05:48,050][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10019 tokens. -[2025-11-10 03:05:48,893][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 55.63%, Block Peak % of device VRAM: 62.16%, ΔTime: 00:00:32 -[2025-11-10 03:05:49,625][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:05:49,626][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:05:49,628][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:05:50,573][__main__][INFO] - Iteration 562 took 1m 4s (43.99% Gen, 54.54% Train). Generation: 28s, Training: 35s. Estimated remaining time: 44h 43m 0s. Estimated total time: 53h 49m 54s. Time estimates for 10 more iterations: 10m 45s, 100 more iterations: 1h 47m 39s, 500 more iterations: 8h 58m 19s. -[2025-11-10 03:05:50,575][__main__][INFO] - Starting iteration 562. -[2025-11-10 03:05:51,049][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 56 and human policies 1. -[2025-11-10 03:05:51,049][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:06:13,051][__main__][INFO] - Number of regex retries in iteration 562: 0 -[2025-11-10 03:06:13,051][__main__][INFO] - agents played in iteration 562 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:06:13,823][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:06:13,851][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:06:13,877][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:06:13,900][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:06:13,900][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:06:13,902][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:06:14,643][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:06:15,114][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:06:15,613][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:06:16,114][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:06:16,622][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:06:17,121][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:06:17,619][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:06:18,120][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:06:18,620][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:06:19,143][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:06:19,640][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:06:20,141][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:06:20,639][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:06:21,137][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:06:21,647][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:06:22,142][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:06:22,640][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:06:23,154][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:06:23,651][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:06:24,149][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:06:24,654][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:06:25,152][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:06:25,664][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:06:26,162][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:06:26,662][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:06:27,155][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:06:27,650][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:06:28,154][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:06:28,653][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:06:29,159][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:06:29,660][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:06:30,163][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:06:30,663][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:06:31,166][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:06:31,675][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:06:32,179][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:06:32,681][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:06:33,185][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:06:33,689][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:06:34,194][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:06:34,703][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:06:35,208][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:06:35,713][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:06:36,231][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:06:36,736][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:06:37,252][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:06:37,759][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:06:38,266][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:06:38,772][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:06:39,273][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:06:39,775][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:06:40,277][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:06:40,778][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:06:41,282][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:06:41,783][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:06:42,285][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:06:42,784][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:06:43,284][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:06:43,787][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:06:44,292][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:06:44,794][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:06:45,296][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:06:45,803][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:06:46,332][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:06:48,119][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10105 tokens. -[2025-11-10 03:06:49,040][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.03%, Current % of VRAM taken: 55.53%, Block Peak % of device VRAM: 62.22%, ΔTime: 00:00:34 -[2025-11-10 03:06:49,699][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:06:49,700][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:06:49,702][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:06:50,629][__main__][INFO] - Iteration 563 took 59s (36.93% Gen, 61.51% Train). Generation: 22s, Training: 36s. Estimated remaining time: 40h 31m 7s. Estimated total time: 49h 39m 1s. Time estimates for 10 more iterations: 9m 55s, 100 more iterations: 1h 39m 18s, 500 more iterations: 8h 16m 30s. -[2025-11-10 03:06:50,631][__main__][INFO] - Starting iteration 563. -[2025-11-10 03:06:51,103][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 56 and human policies 1. -[2025-11-10 03:06:51,104][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:07:20,924][__main__][INFO] - Number of regex retries in iteration 563: 0 -[2025-11-10 03:07:20,925][__main__][INFO] - agents played in iteration 563 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:07:21,701][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:07:21,726][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:07:21,750][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:07:21,772][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:07:21,773][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:07:21,773][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:07:22,514][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:07:22,970][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:07:23,470][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:07:23,964][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:07:24,467][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:07:24,963][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:07:25,466][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:07:25,959][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:07:26,454][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:07:26,957][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:07:27,451][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:07:27,945][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:07:28,451][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:07:28,947][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:07:29,443][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:07:29,942][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:07:30,442][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:07:30,940][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:07:31,438][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:07:31,936][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:07:32,461][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:07:32,963][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:07:33,466][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:07:33,965][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:07:34,470][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:07:34,978][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:07:35,479][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:07:35,976][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:07:36,478][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:07:36,977][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:07:37,477][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:07:37,978][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:07:38,479][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:07:38,986][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:07:39,493][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:07:39,997][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:07:40,501][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:07:41,006][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:07:41,507][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:07:42,009][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:07:42,511][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:07:43,012][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:07:43,514][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:07:44,017][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:07:44,518][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:07:45,021][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:07:45,539][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:07:46,042][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:07:46,545][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:07:47,068][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:07:47,572][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:07:48,076][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:07:48,582][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:07:49,083][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:07:49,586][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:07:50,086][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:07:50,589][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:07:51,091][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:07:51,595][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:07:52,099][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:07:52,603][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:07:53,106][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:07:53,610][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:07:54,116][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:07:54,621][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10057 tokens. -[2025-11-10 03:07:55,506][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 55.64%, Block Peak % of device VRAM: 62.10%, ΔTime: 00:00:33 -[2025-11-10 03:07:56,287][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:07:56,289][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:07:56,291][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:07:57,265][__main__][INFO] - Iteration 564 took 1m 6s (45.07% Gen, 53.45% Train). Generation: 29s, Training: 35s. Estimated remaining time: 45h 59m 8s. Estimated total time: 55h 8m 8s. Time estimates for 10 more iterations: 11m 1s, 100 more iterations: 1h 50m 16s, 500 more iterations: 9h 11m 21s. -[2025-11-10 03:07:57,267][__main__][INFO] - Starting iteration 564. -[2025-11-10 03:07:57,730][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 56 and human policies 1. -[2025-11-10 03:07:57,731][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:08:22,616][__main__][INFO] - Number of regex retries in iteration 564: 0 -[2025-11-10 03:08:22,616][__main__][INFO] - agents played in iteration 564 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:08:23,405][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:08:23,435][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:08:23,465][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:08:23,489][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:08:23,490][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:08:23,491][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:08:24,234][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:08:24,688][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:08:25,194][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:08:25,691][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:08:26,191][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:08:26,686][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:08:27,186][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:08:27,688][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:08:28,184][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:08:28,682][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:08:29,178][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:08:29,678][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:08:30,179][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:08:30,676][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:08:31,172][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:08:31,668][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:08:32,164][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:08:32,669][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:08:33,163][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:08:33,660][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:08:34,156][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:08:34,652][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:08:35,147][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:08:35,643][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:08:36,139][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:08:36,661][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:08:37,158][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:08:37,660][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:08:38,160][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:08:38,657][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:08:39,169][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:08:39,665][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:08:40,160][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:08:40,660][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:08:41,160][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:08:41,662][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:08:42,161][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:08:42,661][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:08:43,174][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:08:43,671][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:08:44,174][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:08:44,687][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:08:45,187][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:08:45,693][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:08:46,200][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:08:46,701][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:08:47,206][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:08:47,708][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:08:48,212][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:08:48,718][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:08:49,247][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:08:49,755][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:08:50,263][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:08:50,769][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:08:51,277][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:08:51,784][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:08:52,291][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:08:52,792][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:08:53,295][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:08:53,797][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:08:54,297][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:08:54,803][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:08:55,305][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:08:55,807][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:08:57,212][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10185 tokens. -[2025-11-10 03:08:58,239][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 55.71%, Block Peak % of device VRAM: 62.38%, ΔTime: 00:00:34 -[2025-11-10 03:08:59,328][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:08:59,331][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:08:59,333][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:09:00,177][__main__][INFO] - Iteration 565 took 1m 2s (39.85% Gen, 58.80% Train). Generation: 24s, Training: 36s. Estimated remaining time: 42h 52m 20s. Estimated total time: 52h 2m 23s. Time estimates for 10 more iterations: 10m 24s, 100 more iterations: 1h 44m 4s, 500 more iterations: 8h 40m 23s. -[2025-11-10 03:09:00,182][__main__][INFO] - Starting iteration 565. -[2025-11-10 03:09:00,654][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 56 and human policies 1. -[2025-11-10 03:09:00,655][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:09:31,672][__main__][INFO] - Number of regex retries in iteration 565: 0 -[2025-11-10 03:09:31,673][__main__][INFO] - agents played in iteration 565 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:09:32,442][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:09:32,468][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:09:32,494][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:09:32,516][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:09:32,517][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:09:32,517][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:09:33,258][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:09:33,715][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:09:34,218][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:09:34,716][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:09:35,213][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:09:35,713][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:09:36,210][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:09:36,706][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:09:37,206][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:09:37,709][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:09:38,205][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:09:38,702][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:09:39,200][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:09:39,699][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:09:40,199][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:09:40,696][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:09:41,199][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:09:41,694][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:09:42,190][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:09:42,684][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:09:43,180][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:09:43,682][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:09:44,177][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:09:44,673][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:09:45,185][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:09:45,680][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:09:46,180][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:09:46,681][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:09:47,183][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:09:47,692][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:09:48,193][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:09:48,697][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:09:49,208][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:09:49,713][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:09:50,220][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:09:50,725][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:09:51,231][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:09:51,738][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:09:52,242][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:09:52,745][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:09:53,258][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:09:53,762][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:09:54,276][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:09:54,781][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:09:55,282][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:09:55,793][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:09:56,296][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:09:56,802][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:09:57,306][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:09:57,807][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:09:58,316][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:09:58,824][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:09:59,333][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:09:59,834][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:10:00,334][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:10:00,839][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:10:01,340][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:10:01,841][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:10:02,351][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:10:02,850][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:10:03,354][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:10:03,854][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:10:04,356][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:10:04,861][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:10:05,363][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10214 tokens. -[2025-11-10 03:10:06,266][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 55.57%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:33 -[2025-11-10 03:10:07,044][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:10:07,047][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:10:07,049][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:10:08,022][__main__][INFO] - Iteration 566 took 1m 7s (46.04% Gen, 52.51% Train). Generation: 31s, Training: 35s. Estimated remaining time: 46h 57m 13s. Estimated total time: 56h 8m 24s. Time estimates for 10 more iterations: 11m 13s, 100 more iterations: 1h 52m 16s, 500 more iterations: 9h 21m 24s. -[2025-11-10 03:10:08,024][__main__][INFO] - Starting iteration 566. -[2025-11-10 03:10:08,491][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 56 and human policies 1. -[2025-11-10 03:10:08,492][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:10:34,909][__main__][INFO] - Number of regex retries in iteration 566: 0 -[2025-11-10 03:10:34,910][__main__][INFO] - agents played in iteration 566 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:10:35,683][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:10:35,708][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:10:35,733][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:10:35,756][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:10:35,756][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:10:35,757][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:10:36,498][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:10:36,955][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:10:37,457][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:10:37,954][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:10:38,449][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:10:38,944][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:10:39,441][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:10:39,937][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:10:40,436][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:10:40,934][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:10:41,430][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:10:41,930][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:10:42,434][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:10:42,952][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:10:43,460][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:10:43,958][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:10:44,454][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:10:44,951][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:10:45,448][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:10:45,950][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:10:46,448][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:10:46,946][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:10:47,447][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:10:47,943][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:10:48,441][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:10:48,938][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:10:49,436][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:10:49,934][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:10:50,430][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:10:50,930][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:10:51,425][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:10:51,921][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:10:52,420][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:10:52,918][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:10:53,415][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:10:53,915][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:10:54,417][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:10:54,925][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:10:55,426][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:10:55,928][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:10:56,434][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:10:56,934][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:10:57,436][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:10:57,937][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:10:58,441][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:10:58,943][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:10:59,445][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:10:59,945][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:11:00,444][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:11:00,949][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:11:01,455][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:11:01,959][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:11:02,462][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:11:02,964][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:11:03,467][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:11:03,980][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:11:04,488][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:11:04,992][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:11:05,496][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:11:06,871][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:11:07,434][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:11:07,936][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:11:08,441][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:11:08,945][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:11:09,454][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10102 tokens. -[2025-11-10 03:11:10,383][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 55.65%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:33 -[2025-11-10 03:11:11,027][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:11:11,029][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:11:11,031][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:11:11,888][__main__][INFO] - Iteration 567 took 1m 3s (41.67% Gen, 56.98% Train). Generation: 26s, Training: 36s. Estimated remaining time: 43h 37m 37s. Estimated total time: 52h 49m 51s. Time estimates for 10 more iterations: 10m 33s, 100 more iterations: 1h 45m 39s, 500 more iterations: 8h 48m 18s. -[2025-11-10 03:11:11,892][__main__][INFO] - Starting iteration 567. -[2025-11-10 03:11:12,361][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 56 and human policies 1. -[2025-11-10 03:11:12,362][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:11:45,009][__main__][INFO] - Number of regex retries in iteration 567: 0 -[2025-11-10 03:11:45,010][__main__][INFO] - agents played in iteration 567 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:11:45,824][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:11:45,851][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:11:45,877][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:11:45,900][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:11:45,901][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:11:45,901][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:11:46,656][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:11:47,108][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:11:47,628][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:11:48,127][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:11:48,631][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:11:49,128][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:11:49,624][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:11:50,122][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:11:50,628][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:11:51,127][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:11:51,629][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:11:52,126][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:11:52,628][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:11:53,130][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:11:53,627][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:11:54,124][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:11:54,622][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:11:55,123][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:11:55,620][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:11:56,123][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:11:56,627][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:11:57,125][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:11:57,621][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:11:58,121][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:11:58,620][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:11:59,120][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:11:59,628][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:12:00,141][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:12:00,644][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:12:01,150][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:12:01,656][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:12:02,163][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:12:02,669][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:12:03,171][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:12:03,680][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:12:04,185][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:12:04,699][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:12:05,202][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:12:05,703][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:12:06,207][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:12:06,713][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:12:07,225][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:12:07,731][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:12:08,235][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:12:08,740][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:12:09,246][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:12:09,747][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:12:10,251][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:12:10,757][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:12:11,261][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:12:11,763][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:12:12,265][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:12:12,769][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:12:13,273][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:12:13,780][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:12:14,291][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:12:14,795][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:12:15,303][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:12:15,810][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:12:16,313][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:12:16,819][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:12:17,324][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:12:17,832][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:12:18,334][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:12:18,841][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10242 tokens. -[2025-11-10 03:12:19,749][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 55.71%, Block Peak % of device VRAM: 62.20%, ΔTime: 00:00:33 -[2025-11-10 03:12:20,524][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:12:20,526][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:12:20,528][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:12:21,456][__main__][INFO] - Iteration 568 took 1m 9s (47.25% Gen, 51.40% Train). Generation: 32s, Training: 35s. Estimated remaining time: 48h 21m 21s. Estimated total time: 57h 34m 46s. Time estimates for 10 more iterations: 11m 30s, 100 more iterations: 1h 55m 9s, 500 more iterations: 9h 35m 47s. -[2025-11-10 03:12:21,458][__main__][INFO] - Starting iteration 568. -[2025-11-10 03:12:21,937][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 56 and human policies 1. -[2025-11-10 03:12:21,937][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:12:37,374][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 03:12:41,465][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 03:12:48,023][__main__][INFO] - Number of regex retries in iteration 568: 2 -[2025-11-10 03:12:48,024][__main__][INFO] - agents played in iteration 568 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:12:48,811][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:12:48,834][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:12:48,859][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:12:48,882][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:12:48,883][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:12:48,883][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:12:49,643][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:12:50,098][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:12:50,607][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:12:51,106][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:12:51,605][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:12:52,122][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:12:52,623][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:12:53,128][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:12:53,627][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:12:54,127][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:12:54,630][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:12:55,130][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:12:55,631][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:12:56,135][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:12:56,631][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:12:57,129][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:12:57,626][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:12:58,130][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:12:58,631][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:12:59,132][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:12:59,634][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:13:00,133][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:13:00,640][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:13:01,139][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:13:01,639][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:13:02,147][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:13:02,654][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:13:03,158][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:13:03,661][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:13:04,168][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:13:04,672][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:13:05,176][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:13:05,683][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:13:06,186][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:13:06,688][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:13:07,193][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:13:07,704][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:13:08,218][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:13:08,722][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:13:09,241][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:13:09,745][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:13:10,248][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:13:10,750][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:13:11,253][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:13:11,753][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:13:12,254][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:13:12,754][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:13:13,254][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:13:13,755][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:13:15,133][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:13:15,822][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:13:16,326][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:13:16,828][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:13:17,330][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:13:17,833][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:13:18,351][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:13:18,851][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:13:19,353][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:13:19,861][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:13:20,361][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:13:20,863][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:13:21,364][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:13:21,864][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:13:22,366][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:13:22,869][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10077 tokens. -[2025-11-10 03:13:23,782][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 55.69%, Block Peak % of device VRAM: 62.12%, ΔTime: 00:00:34 -[2025-11-10 03:13:24,435][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:13:24,437][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:13:24,439][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:13:25,312][__main__][INFO] - Iteration 569 took 1m 3s (41.16% Gen, 57.46% Train). Generation: 26s, Training: 36s. Estimated remaining time: 43h 34m 20s. Estimated total time: 52h 48m 49s. Time estimates for 10 more iterations: 10m 33s, 100 more iterations: 1h 45m 37s, 500 more iterations: 8h 48m 8s. -[2025-11-10 03:13:25,316][__main__][INFO] - Starting iteration 569. -[2025-11-10 03:13:25,780][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 56 and human policies 1. -[2025-11-10 03:13:25,781][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:13:54,201][__main__][INFO] - Number of regex retries in iteration 569: 0 -[2025-11-10 03:13:54,201][__main__][INFO] - agents played in iteration 569 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:13:54,976][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:13:54,999][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:13:55,021][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:13:55,043][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:13:55,044][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:13:55,045][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:13:55,792][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:13:56,249][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:13:56,755][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:13:57,251][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:13:57,746][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:13:58,240][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:13:58,736][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:13:59,231][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:13:59,726][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:14:00,226][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:14:00,722][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:14:01,218][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:14:01,712][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:14:02,211][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:14:02,706][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:14:03,202][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:14:03,700][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:14:04,197][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:14:04,697][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:14:05,201][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:14:05,703][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:14:06,209][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:14:06,708][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:14:07,206][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:14:07,714][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:14:08,215][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:14:08,728][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:14:09,229][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:14:09,731][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:14:10,236][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:14:10,736][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:14:11,249][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:14:11,750][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:14:12,256][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:14:12,757][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:14:13,259][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:14:13,762][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:14:14,266][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:14:14,776][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:14:15,282][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:14:15,787][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:14:16,288][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:14:16,791][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:14:17,291][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:14:17,794][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:14:18,295][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:14:18,796][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:14:19,301][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:14:19,805][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:14:20,305][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:14:20,807][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:14:21,309][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:14:21,817][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:14:22,316][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:14:22,818][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:14:23,336][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:14:23,838][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:14:24,341][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:14:24,845][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:14:25,347][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:14:25,858][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:14:26,362][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:14:26,869][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:14:27,379][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:14:27,883][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10083 tokens. -[2025-11-10 03:14:28,741][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.25%, Current % of VRAM taken: 55.75%, Block Peak % of device VRAM: 62.23%, ΔTime: 00:00:32 -[2025-11-10 03:14:29,488][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:14:29,490][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:14:29,492][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:14:30,394][__main__][INFO] - Iteration 570 took 1m 4s (43.99% Gen, 54.62% Train). Generation: 28s, Training: 35s. Estimated remaining time: 44h 35m 9s. Estimated total time: 53h 50m 42s. Time estimates for 10 more iterations: 10m 46s, 100 more iterations: 1h 47m 41s, 500 more iterations: 8h 58m 27s. -[2025-11-10 03:14:30,396][__main__][INFO] - Starting iteration 570. -[2025-11-10 03:14:30,864][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 56 and human policies 1. -[2025-11-10 03:14:30,865][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:14:55,973][__main__][INFO] - Number of regex retries in iteration 570: 0 -[2025-11-10 03:14:55,974][__main__][INFO] - agents played in iteration 570 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:14:56,740][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:14:56,762][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:14:56,786][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:14:56,808][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:14:56,809][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:14:56,809][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:14:57,582][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:14:58,042][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:14:58,546][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:14:59,047][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:14:59,546][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:15:00,048][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:15:00,548][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:15:01,047][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:15:01,557][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:15:02,059][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:15:02,558][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:15:03,061][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:15:03,558][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:15:04,058][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:15:04,554][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:15:05,051][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:15:05,565][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:15:06,061][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:15:06,560][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:15:07,057][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:15:07,553][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:15:08,061][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:15:08,562][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:15:09,064][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:15:09,568][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:15:10,074][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:15:10,577][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:15:11,078][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:15:11,582][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:15:12,113][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:15:12,623][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:15:13,129][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:15:13,647][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:15:14,148][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:15:14,665][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:15:15,174][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:15:15,674][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:15:16,186][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:15:16,688][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:15:17,193][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:15:17,697][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:15:18,201][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:15:18,709][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:15:19,211][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:15:19,714][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:15:20,223][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:15:20,724][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:15:21,227][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:15:21,733][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:15:22,235][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:15:22,737][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:15:23,237][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:15:23,747][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:15:24,989][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:15:25,496][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:15:26,000][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:15:26,507][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:15:27,012][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:15:27,517][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:15:28,022][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:15:28,533][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:15:29,047][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:15:29,548][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:15:30,049][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:15:30,556][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10192 tokens. -[2025-11-10 03:15:31,463][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.25%, Current % of VRAM taken: 55.75%, Block Peak % of device VRAM: 62.21%, ΔTime: 00:00:33 -[2025-11-10 03:15:32,136][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:15:32,138][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:15:32,140][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:15:33,885][__main__][INFO] - Iteration 571 took 1m 3s (39.84% Gen, 57.39% Train). Generation: 25s, Training: 36s. Estimated remaining time: 43h 14m 26s. Estimated total time: 52h 31m 3s. Time estimates for 10 more iterations: 10m 30s, 100 more iterations: 1h 45m 2s, 500 more iterations: 8h 45m 10s. -[2025-11-10 03:15:33,887][__main__][INFO] - Starting iteration 571. -[2025-11-10 03:15:34,355][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 57 and human policies 1. -[2025-11-10 03:15:34,355][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:16:03,579][__main__][INFO] - Number of regex retries in iteration 571: 0 -[2025-11-10 03:16:03,580][__main__][INFO] - agents played in iteration 571 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:16:04,413][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:16:04,442][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:16:04,468][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:16:04,491][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:16:04,492][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:16:04,493][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:16:05,239][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:16:05,703][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:16:06,209][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:16:06,708][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:16:07,212][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:16:07,714][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:16:08,214][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:16:08,711][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:16:09,209][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:16:09,711][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:16:10,209][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:16:10,706][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:16:11,209][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:16:11,709][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:16:12,210][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:16:12,712][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:16:13,210][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:16:13,713][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:16:14,220][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:16:14,725][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:16:15,232][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:16:15,732][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:16:16,247][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:16:16,751][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:16:17,255][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:16:17,755][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:16:18,254][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:16:18,777][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:16:19,278][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:16:19,780][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:16:20,285][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:16:20,786][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:16:21,290][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:16:21,792][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:16:22,295][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:16:22,802][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:16:23,305][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:16:23,809][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:16:24,311][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:16:24,813][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:16:25,315][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:16:25,823][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:16:26,326][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:16:26,834][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:16:27,339][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:16:27,849][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:16:28,354][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:16:28,861][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:16:29,386][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:16:29,887][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:16:30,389][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:16:30,886][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:16:31,386][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:16:31,896][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:16:32,395][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:16:32,893][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:16:33,395][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:16:33,901][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:16:34,405][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:16:34,906][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:16:35,407][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:16:35,912][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:16:36,415][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:16:36,916][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:16:37,421][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10192 tokens. -[2025-11-10 03:16:38,330][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 55.66%, Block Peak % of device VRAM: 62.42%, ΔTime: 00:00:33 -[2025-11-10 03:16:39,094][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:16:39,096][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:16:39,098][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:16:40,005][__main__][INFO] - Iteration 572 took 1m 5s (44.52% Gen, 54.10% Train). Generation: 29s, Training: 35s. Estimated remaining time: 45h 24m 49s. Estimated total time: 54h 42m 32s. Time estimates for 10 more iterations: 10m 56s, 100 more iterations: 1h 49m 25s, 500 more iterations: 9h 7m 5s. -[2025-11-10 03:16:40,007][__main__][INFO] - Starting iteration 572. -[2025-11-10 03:16:40,480][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 57 and human policies 1. -[2025-11-10 03:16:40,480][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:17:07,970][__main__][INFO] - Number of regex retries in iteration 572: 0 -[2025-11-10 03:17:07,971][__main__][INFO] - agents played in iteration 572 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:17:08,741][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:17:08,769][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:17:08,796][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:17:08,819][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:17:08,819][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:17:08,820][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:17:09,563][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:17:10,024][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:17:10,532][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:17:11,033][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:17:11,537][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:17:12,038][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:17:12,540][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:17:13,036][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:17:13,534][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:17:14,032][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:17:14,537][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:17:15,046][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:17:15,555][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:17:16,058][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:17:16,562][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:17:17,063][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:17:17,565][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:17:18,070][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:17:18,570][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:17:19,071][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:17:19,573][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:17:20,077][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:17:20,597][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:17:21,097][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:17:21,599][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:17:22,110][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:17:22,616][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:17:23,121][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:17:23,625][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:17:24,128][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:17:24,639][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:17:25,146][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:17:25,654][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:17:26,156][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:17:26,662][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:17:27,176][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:17:27,679][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:17:28,179][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:17:28,684][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:17:29,186][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:17:29,689][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:17:30,189][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:17:30,695][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:17:31,208][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:17:31,709][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:17:32,215][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:17:32,717][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:17:33,222][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:17:33,726][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:17:34,716][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:17:35,303][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:17:35,804][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:17:36,311][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:17:36,815][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:17:37,317][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:17:37,822][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:17:38,327][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:17:38,829][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:17:39,330][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:17:39,832][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:17:40,344][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:17:40,846][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:17:41,350][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:17:41,861][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:17:42,363][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10150 tokens. -[2025-11-10 03:17:43,258][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.01%, Current % of VRAM taken: 55.51%, Block Peak % of device VRAM: 62.12%, ΔTime: 00:00:33 -[2025-11-10 03:17:43,909][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:17:43,910][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:17:43,913][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:17:44,795][__main__][INFO] - Iteration 573 took 1m 4s (42.74% Gen, 55.88% Train). Generation: 27s, Training: 35s. Estimated remaining time: 44h 17m 0s. Estimated total time: 53h 35m 48s. Time estimates for 10 more iterations: 10m 43s, 100 more iterations: 1h 47m 11s, 500 more iterations: 8h 55m 58s. -[2025-11-10 03:17:44,798][__main__][INFO] - Starting iteration 573. -[2025-11-10 03:17:45,263][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 57 and human policies 1. -[2025-11-10 03:17:45,263][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:18:13,898][__main__][INFO] - Number of regex retries in iteration 573: 0 -[2025-11-10 03:18:13,898][__main__][INFO] - agents played in iteration 573 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:18:14,688][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:18:14,711][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:18:14,734][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:18:14,756][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:18:14,757][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:18:14,757][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:18:15,503][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:18:15,975][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:18:16,478][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:18:16,974][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:18:17,470][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:18:17,966][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:18:18,490][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:18:18,986][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:18:19,483][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:18:19,984][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:18:20,487][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:18:21,003][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:18:21,503][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:18:22,007][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:18:22,512][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:18:23,010][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:18:23,513][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:18:24,012][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:18:24,514][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:18:25,019][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:18:25,522][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:18:26,025][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:18:26,526][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:18:27,029][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:18:27,529][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:18:28,058][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:18:28,563][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:18:29,077][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:18:29,579][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:18:30,084][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:18:30,592][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:18:31,097][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:18:31,607][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:18:32,105][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:18:32,609][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:18:33,116][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:18:33,623][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:18:34,127][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:18:34,629][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:18:35,131][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:18:35,633][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:18:36,134][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:18:36,632][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:18:37,135][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:18:37,636][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:18:38,138][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:18:38,638][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:18:39,138][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:18:39,641][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:18:40,147][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:18:40,650][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:18:41,158][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:18:41,665][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:18:42,182][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:18:42,686][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:18:43,189][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:18:43,693][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:18:44,192][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:18:44,707][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:18:45,207][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:18:45,713][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:18:46,216][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:18:46,719][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:18:47,223][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:18:47,725][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10095 tokens. -[2025-11-10 03:18:48,648][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 55.56%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:33 -[2025-11-10 03:18:49,414][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:18:49,415][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:18:49,417][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:18:50,332][__main__][INFO] - Iteration 574 took 1m 5s (44.01% Gen, 54.59% Train). Generation: 28s, Training: 35s. Estimated remaining time: 44h 53m 35s. Estimated total time: 54h 13m 29s. Time estimates for 10 more iterations: 10m 50s, 100 more iterations: 1h 48m 26s, 500 more iterations: 9h 2m 14s. -[2025-11-10 03:18:50,334][__main__][INFO] - Starting iteration 574. -[2025-11-10 03:18:50,801][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 57 and human policies 1. -[2025-11-10 03:18:50,802][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:19:07,602][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 03:19:09,476][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 03:19:17,003][__main__][INFO] - Number of regex retries in iteration 574: 2 -[2025-11-10 03:19:17,004][__main__][INFO] - agents played in iteration 574 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:19:17,804][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:19:17,832][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:19:17,859][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:19:17,883][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:19:17,883][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:19:17,884][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:19:18,641][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:19:19,096][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:19:19,598][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:19:20,110][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:19:20,608][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:19:21,105][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:19:21,601][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:19:22,096][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:19:22,603][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:19:23,098][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:19:23,600][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:19:24,098][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:19:24,594][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:19:25,090][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:19:25,590][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:19:26,088][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:19:26,602][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:19:27,109][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:19:27,608][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:19:28,109][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:19:28,610][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:19:29,119][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:19:29,624][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:19:30,126][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:19:30,630][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:19:31,133][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:19:31,638][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:19:32,141][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:19:32,644][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:19:33,148][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:19:33,652][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:19:34,156][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:19:34,659][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:19:35,160][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:19:35,665][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:19:36,171][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:19:36,672][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:19:37,187][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:19:37,723][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:19:38,231][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:19:38,735][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:19:39,238][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:19:39,740][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:19:40,244][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:19:40,742][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:19:41,244][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:19:41,742][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:19:42,244][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:19:42,744][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:19:43,248][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:19:43,752][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:19:44,253][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:19:44,755][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:19:45,857][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:19:46,722][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:19:47,226][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:19:47,729][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:19:48,234][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:19:48,738][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:19:49,240][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:19:49,740][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:19:50,239][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:19:50,739][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:19:51,239][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:19:51,738][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10082 tokens. -[2025-11-10 03:19:52,637][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 55.60%, Block Peak % of device VRAM: 62.18%, ΔTime: 00:00:34 -[2025-11-10 03:19:53,296][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:19:53,297][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:19:53,299][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:19:54,119][__main__][INFO] - Iteration 575 took 1m 3s (41.38% Gen, 57.32% Train). Generation: 26s, Training: 36s. Estimated remaining time: 43h 24m 56s. Estimated total time: 52h 45m 53s. Time estimates for 10 more iterations: 10m 33s, 100 more iterations: 1h 45m 31s, 500 more iterations: 8h 47m 38s. -[2025-11-10 03:19:54,124][__main__][INFO] - Starting iteration 575. -[2025-11-10 03:19:54,600][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 57 and human policies 1. -[2025-11-10 03:19:54,600][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:20:22,704][__main__][INFO] - Number of regex retries in iteration 575: 0 -[2025-11-10 03:20:22,705][__main__][INFO] - agents played in iteration 575 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:20:23,472][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:20:23,499][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:20:23,524][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:20:23,547][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:20:23,548][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:20:23,548][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:20:24,291][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:20:24,748][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:20:25,249][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:20:25,747][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:20:26,246][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:20:26,745][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:20:27,246][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:20:27,742][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:20:28,240][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:20:28,739][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:20:29,241][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:20:29,744][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:20:30,244][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:20:30,745][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:20:31,248][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:20:31,747][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:20:32,245][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:20:32,744][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:20:33,245][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:20:33,747][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:20:34,250][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:20:34,754][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:20:35,255][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:20:35,758][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:20:36,259][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:20:36,768][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:20:37,269][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:20:37,771][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:20:38,277][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:20:38,778][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:20:39,279][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:20:39,783][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:20:40,288][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:20:40,790][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:20:41,301][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:20:41,803][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:20:42,319][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:20:42,821][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:20:43,321][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:20:43,827][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:20:44,328][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:20:44,831][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:20:45,337][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:20:45,841][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:20:46,348][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:20:46,849][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:20:47,357][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:20:47,861][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:20:48,362][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:20:48,866][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:20:49,369][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:20:49,875][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:20:50,376][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:20:50,877][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:20:51,377][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:20:51,880][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:20:52,383][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:20:52,887][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:20:53,387][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:20:53,885][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:20:54,383][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:20:54,883][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:20:55,383][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:20:55,881][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:20:56,383][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10128 tokens. -[2025-11-10 03:20:57,249][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 55.60%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:32 -[2025-11-10 03:20:58,003][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:20:58,005][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:20:58,006][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:20:58,888][__main__][INFO] - Iteration 576 took 1m 4s (43.72% Gen, 54.91% Train). Generation: 28s, Training: 35s. Estimated remaining time: 44h 12m 25s. Estimated total time: 53h 34m 27s. Time estimates for 10 more iterations: 10m 42s, 100 more iterations: 1h 47m 8s, 500 more iterations: 8h 55m 44s. -[2025-11-10 03:20:58,890][__main__][INFO] - Starting iteration 576. -[2025-11-10 03:20:59,377][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 57 and human policies 1. -[2025-11-10 03:20:59,378][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:21:24,946][__main__][INFO] - Number of regex retries in iteration 576: 0 -[2025-11-10 03:21:24,947][__main__][INFO] - agents played in iteration 576 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:21:25,767][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:21:25,792][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:21:25,819][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:21:25,842][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:21:25,843][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:21:25,843][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:21:26,577][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:21:27,036][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:21:27,536][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:21:28,037][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:21:28,531][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:21:29,026][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:21:29,530][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:21:30,035][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:21:30,536][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:21:31,044][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:21:31,542][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:21:32,043][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:21:32,544][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:21:33,044][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:21:33,546][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:21:34,046][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:21:34,546][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:21:35,048][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:21:35,548][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:21:36,048][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:21:36,547][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:21:37,046][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:21:37,548][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:21:38,049][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:21:38,552][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:21:39,054][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:21:39,562][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:21:40,071][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:21:40,577][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:21:41,082][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:21:41,590][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:21:42,092][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:21:42,596][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:21:43,097][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:21:43,596][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:21:44,101][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:21:44,603][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:21:45,105][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:21:45,605][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:21:46,105][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:21:46,607][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:21:47,108][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:21:47,613][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:21:48,122][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:21:48,624][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:21:49,140][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:21:49,641][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:21:50,142][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:21:50,646][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:21:51,149][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:21:51,656][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:21:52,158][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:21:52,661][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:21:53,165][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:21:53,669][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:21:54,171][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:21:54,671][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:21:55,167][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:21:55,666][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:21:56,163][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:21:56,664][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:21:57,166][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:21:57,663][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:21:58,163][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:21:58,662][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10000 tokens. -[2025-11-10 03:21:59,545][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.00%, Current % of VRAM taken: 55.50%, Block Peak % of device VRAM: 62.07%, ΔTime: 00:00:32 -[2025-11-10 03:22:00,314][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:22:00,316][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:22:00,317][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:22:01,249][__main__][INFO] - Iteration 577 took 1m 1s (41.33% Gen, 57.17% Train). Generation: 25s, Training: 35s. Estimated remaining time: 42h 10m 31s. Estimated total time: 51h 33m 35s. Time estimates for 10 more iterations: 10m 18s, 100 more iterations: 1h 43m 7s, 500 more iterations: 8h 35m 35s. -[2025-11-10 03:22:01,251][__main__][INFO] - Starting iteration 577. -[2025-11-10 03:22:01,740][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 57 and human policies 1. -[2025-11-10 03:22:01,740][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:22:32,765][__main__][INFO] - Number of regex retries in iteration 577: 0 -[2025-11-10 03:22:32,767][__main__][INFO] - agents played in iteration 577 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:22:33,540][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:22:33,568][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:22:33,595][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:22:33,618][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:22:33,618][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:22:33,620][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:22:34,355][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:22:34,810][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:22:35,323][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:22:35,823][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:22:36,326][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:22:36,825][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:22:37,325][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:22:37,832][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:22:38,331][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:22:38,835][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:22:39,338][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:22:39,839][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:22:40,356][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:22:40,859][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:22:41,361][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:22:41,868][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:22:42,369][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:22:42,876][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:22:43,384][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:22:43,887][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:22:44,391][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:22:44,892][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:22:45,397][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:22:45,900][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:22:46,402][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:22:46,908][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:22:47,412][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:22:47,915][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:22:48,428][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:22:48,930][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:22:49,443][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:22:49,946][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:22:50,452][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:22:50,954][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:22:51,454][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:22:51,960][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:22:52,462][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:22:52,962][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:22:53,470][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:22:53,970][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:22:54,473][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:22:54,976][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:22:55,478][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:22:55,982][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:22:56,485][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:22:56,988][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:22:57,494][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:22:57,997][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:22:58,501][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:22:59,004][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:22:59,505][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:23:00,012][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:23:00,513][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:23:01,015][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:23:01,519][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:23:02,020][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:23:02,536][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:23:03,039][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:23:03,539][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:23:04,040][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:23:04,540][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:23:05,047][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:23:05,546][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:23:06,045][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:23:06,554][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10095 tokens. -[2025-11-10 03:23:07,372][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 55.60%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:33 -[2025-11-10 03:23:08,008][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:23:08,010][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:23:08,012][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:23:08,796][__main__][INFO] - Iteration 578 took 1m 7s (46.27% Gen, 52.56% Train). Generation: 31s, Training: 35s. Estimated remaining time: 46h 28m 40s. Estimated total time: 55h 52m 52s. Time estimates for 10 more iterations: 11m 10s, 100 more iterations: 1h 51m 45s, 500 more iterations: 9h 18m 48s. -[2025-11-10 03:23:08,799][__main__][INFO] - Starting iteration 578. -[2025-11-10 03:23:09,291][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 57 and human policies 1. -[2025-11-10 03:23:09,291][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:23:34,118][__main__][INFO] - Number of regex retries in iteration 578: 0 -[2025-11-10 03:23:34,119][__main__][INFO] - agents played in iteration 578 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:23:34,883][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:23:34,910][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:23:34,938][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:23:34,961][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:23:34,962][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:23:34,963][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:23:35,710][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:23:36,163][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:23:36,665][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:23:37,161][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:23:37,659][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:23:38,170][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:23:38,665][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:23:39,164][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:23:39,659][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:23:40,162][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:23:40,664][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:23:41,160][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:23:41,658][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:23:42,164][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:23:42,660][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:23:43,154][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:23:43,653][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:23:44,153][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:23:44,665][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:23:45,165][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:23:45,666][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:23:46,166][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:23:46,664][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:23:47,166][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:23:47,667][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:23:48,172][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:23:48,680][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:23:49,183][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:23:49,686][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:23:50,194][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:23:50,696][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:23:51,198][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:23:51,699][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:23:52,200][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:23:52,705][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:23:53,208][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:23:53,712][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:23:54,215][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:23:54,717][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:23:55,223][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:23:55,733][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:23:56,235][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:23:56,738][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:23:57,243][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:23:57,750][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:23:58,251][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:23:58,751][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:23:59,257][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:23:59,758][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:24:00,268][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:24:00,772][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:24:01,274][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:24:01,775][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:24:02,278][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:24:03,602][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:24:04,130][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:24:04,631][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:24:05,137][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:24:05,644][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:24:06,145][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:24:06,648][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:24:07,153][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:24:07,653][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:24:08,152][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:24:08,653][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10028 tokens. -[2025-11-10 03:24:09,585][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.04%, Current % of VRAM taken: 55.54%, Block Peak % of device VRAM: 62.16%, ΔTime: 00:00:33 -[2025-11-10 03:24:10,234][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:24:10,236][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:24:10,238][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:24:11,050][__main__][INFO] - Iteration 579 took 1m 1s (40.20% Gen, 58.48% Train). Generation: 24s, Training: 36s. Estimated remaining time: 42h 2m 45s. Estimated total time: 51h 27m 59s. Time estimates for 10 more iterations: 10m 17s, 100 more iterations: 1h 42m 55s, 500 more iterations: 8h 34m 39s. -[2025-11-10 03:24:11,054][__main__][INFO] - Starting iteration 579. -[2025-11-10 03:24:11,519][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 57 and human policies 1. -[2025-11-10 03:24:11,519][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:24:43,647][__main__][INFO] - Number of regex retries in iteration 579: 0 -[2025-11-10 03:24:43,648][__main__][INFO] - agents played in iteration 579 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:24:44,431][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:24:44,459][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:24:44,485][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:24:44,508][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:24:44,509][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:24:44,509][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:24:45,259][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:24:45,711][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:24:46,213][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:24:46,707][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:24:47,208][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:24:47,704][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:24:48,201][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:24:48,702][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:24:49,202][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:24:49,705][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:24:50,217][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:24:50,717][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:24:51,228][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:24:51,744][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:24:52,247][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:24:52,761][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:24:53,266][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:24:53,769][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:24:54,281][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:24:54,783][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:24:55,285][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:24:55,785][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:24:56,286][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:24:56,793][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:24:57,293][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:24:57,799][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:24:58,303][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:24:58,810][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:24:59,313][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:24:59,819][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:25:00,342][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:25:00,853][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:25:01,356][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:25:01,866][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:25:02,367][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:25:02,867][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:25:03,368][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:25:03,873][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:25:04,380][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:25:04,880][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:25:05,381][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:25:05,888][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:25:06,389][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:25:06,893][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:25:07,396][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:25:07,897][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:25:08,403][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:25:08,906][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:25:09,407][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:25:09,914][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:25:10,414][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:25:10,920][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:25:11,420][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:25:11,921][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:25:12,427][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:25:12,928][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:25:13,430][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:25:13,943][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:25:14,444][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:25:14,953][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:25:15,459][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:25:15,965][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:25:16,481][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:25:16,979][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:25:17,480][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10103 tokens. -[2025-11-10 03:25:18,270][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 55.66%, Block Peak % of device VRAM: 62.37%, ΔTime: 00:00:33 -[2025-11-10 03:25:19,036][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:25:19,038][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:25:19,040][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:25:19,934][__main__][INFO] - Iteration 580 took 1m 8s (46.96% Gen, 51.73% Train). Generation: 32s, Training: 35s. Estimated remaining time: 47h 34m 25s. Estimated total time: 57h 0m 48s. Time estimates for 10 more iterations: 11m 24s, 100 more iterations: 1h 54m 1s, 500 more iterations: 9h 30m 8s. -[2025-11-10 03:25:19,936][__main__][INFO] - Starting iteration 580. -[2025-11-10 03:25:20,403][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 57 and human policies 1. -[2025-11-10 03:25:20,403][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:25:45,455][__main__][INFO] - Number of regex retries in iteration 580: 0 -[2025-11-10 03:25:45,456][__main__][INFO] - agents played in iteration 580 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:25:46,224][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:25:46,252][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:25:46,279][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:25:46,302][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:25:46,302][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:25:46,303][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:25:47,063][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:25:47,517][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:25:48,023][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:25:48,519][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:25:49,013][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:25:49,510][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:25:50,004][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:25:50,500][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:25:50,993][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:25:51,494][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:25:52,000][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:25:52,500][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:25:52,998][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:25:53,494][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:25:53,991][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:25:54,491][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:25:54,985][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:25:55,481][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:25:55,983][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:25:56,481][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:25:56,976][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:25:57,474][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:25:57,973][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:25:58,473][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:25:58,968][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:25:59,464][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:25:59,975][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:26:00,470][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:26:00,965][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:26:01,466][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:26:01,964][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:26:02,476][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:26:02,977][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:26:03,480][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:26:03,985][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:26:04,488][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:26:04,994][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:26:05,498][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:26:06,001][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:26:06,509][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:26:07,016][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:26:07,523][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:26:08,027][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:26:08,532][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:26:09,033][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:26:09,538][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:26:10,038][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:26:10,542][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:26:11,045][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:26:11,554][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:26:13,357][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:26:13,934][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:26:14,459][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:26:14,964][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:26:15,470][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:26:15,976][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:26:16,479][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:26:16,986][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:26:17,491][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:26:17,999][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:26:18,507][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:26:19,008][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:26:19,510][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:26:20,013][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:26:20,513][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10173 tokens. -[2025-11-10 03:26:21,433][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 55.63%, Block Peak % of device VRAM: 62.35%, ΔTime: 00:00:34 -[2025-11-10 03:26:22,105][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:26:22,107][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:26:22,109][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:26:23,818][__main__][INFO] - Iteration 581 took 1m 3s (39.50% Gen, 57.80% Train). Generation: 25s, Training: 36s. Estimated remaining time: 43h 23m 21s. Estimated total time: 52h 50m 48s. Time estimates for 10 more iterations: 10m 34s, 100 more iterations: 1h 45m 41s, 500 more iterations: 8h 48m 28s. -[2025-11-10 03:26:23,821][__main__][INFO] - Starting iteration 581. -[2025-11-10 03:26:24,287][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 58 and human policies 1. -[2025-11-10 03:26:24,288][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:26:52,090][__main__][INFO] - Number of regex retries in iteration 581: 0 -[2025-11-10 03:26:52,090][__main__][INFO] - agents played in iteration 581 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:26:52,862][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:26:52,890][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:26:52,916][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:26:52,939][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:26:52,940][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:26:52,940][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:26:53,703][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:26:54,159][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:26:54,659][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:26:55,163][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:26:55,660][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:26:56,158][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:26:56,689][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:26:57,188][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:26:57,684][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:26:58,186][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:26:58,684][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:26:59,196][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:26:59,696][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:27:00,195][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:27:00,694][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:27:01,192][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:27:01,691][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:27:02,186][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:27:02,683][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:27:03,187][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:27:03,681][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:27:04,181][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:27:04,685][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:27:05,186][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:27:05,688][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:27:06,186][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:27:06,686][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:27:07,210][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:27:07,710][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:27:08,215][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:27:08,719][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:27:09,221][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:27:09,724][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:27:10,230][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:27:10,731][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:27:11,239][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:27:11,745][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:27:12,247][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:27:12,747][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:27:13,251][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:27:13,756][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:27:14,258][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:27:14,759][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:27:15,260][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:27:15,762][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:27:16,271][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:27:16,771][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:27:17,271][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:27:17,785][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:27:18,285][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:27:18,791][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:27:19,294][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:27:19,793][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:27:20,299][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:27:20,803][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:27:21,306][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:27:21,807][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:27:22,308][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:27:22,806][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:27:23,305][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:27:23,805][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:27:24,307][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:27:24,810][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:27:25,317][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:27:25,819][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10023 tokens. -[2025-11-10 03:27:26,713][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 13.98%, Current % of VRAM taken: 55.48%, Block Peak % of device VRAM: 62.03%, ΔTime: 00:00:33 -[2025-11-10 03:27:27,449][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:27:27,451][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:27:27,453][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:27:28,386][__main__][INFO] - Iteration 582 took 1m 4s (43.37% Gen, 55.17% Train). Generation: 27s, Training: 35s. Estimated remaining time: 43h 56m 28s. Estimated total time: 53h 25m 0s. Time estimates for 10 more iterations: 10m 41s, 100 more iterations: 1h 46m 50s, 500 more iterations: 8h 54m 10s. -[2025-11-10 03:27:28,389][__main__][INFO] - Starting iteration 582. -[2025-11-10 03:27:28,853][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 58 and human policies 1. -[2025-11-10 03:27:28,854][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:27:58,528][__main__][INFO] - Number of regex retries in iteration 582: 0 -[2025-11-10 03:27:58,528][__main__][INFO] - agents played in iteration 582 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:27:59,302][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:27:59,325][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:27:59,347][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:27:59,369][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:27:59,370][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:27:59,370][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:28:00,134][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:28:00,589][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:28:01,088][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:28:01,583][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:28:02,093][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:28:02,587][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:28:03,082][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:28:03,580][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:28:04,077][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:28:04,579][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:28:05,077][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:28:05,574][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:28:06,073][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:28:06,570][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:28:07,066][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:28:07,566][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:28:08,063][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:28:08,560][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:28:09,057][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:28:09,559][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:28:10,056][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:28:10,552][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:28:11,048][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:28:11,546][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:28:12,043][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:28:12,546][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:28:13,050][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:28:13,552][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:28:14,057][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:28:14,556][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:28:15,055][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:28:15,555][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:28:16,056][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:28:16,561][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:28:17,063][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:28:17,561][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:28:18,065][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:28:18,568][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:28:19,071][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:28:19,571][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:28:20,978][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:28:21,551][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:28:22,052][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:28:22,554][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:28:23,061][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:28:23,563][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:28:24,067][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:28:24,568][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:28:25,071][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:28:25,573][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:28:26,077][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:28:26,582][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:28:27,088][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:28:27,589][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:28:28,090][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:28:28,600][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:28:29,103][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:28:29,620][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:28:30,123][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:28:30,625][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:28:31,129][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:28:31,631][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:28:32,132][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:28:32,634][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:28:33,162][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10071 tokens. -[2025-11-10 03:28:34,108][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.25%, Current % of VRAM taken: 55.75%, Block Peak % of device VRAM: 62.16%, ΔTime: 00:00:33 -[2025-11-10 03:28:34,771][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:28:34,773][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:28:34,775][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:28:35,583][__main__][INFO] - Iteration 583 took 1m 6s (44.47% Gen, 54.32% Train). Generation: 29s, Training: 36s. Estimated remaining time: 46h 6m 53s. Estimated total time: 55h 36m 31s. Time estimates for 10 more iterations: 11m 7s, 100 more iterations: 1h 51m 13s, 500 more iterations: 9h 16m 5s. -[2025-11-10 03:28:35,586][__main__][INFO] - Starting iteration 583. -[2025-11-10 03:28:36,069][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 58 and human policies 1. -[2025-11-10 03:28:36,070][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:29:08,490][__main__][INFO] - Number of regex retries in iteration 583: 0 -[2025-11-10 03:29:08,491][__main__][INFO] - agents played in iteration 583 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:29:09,264][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:29:09,292][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:29:09,319][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:29:09,342][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:29:09,343][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:29:09,344][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:29:10,092][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:29:10,546][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:29:11,053][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:29:11,553][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:29:12,048][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:29:12,543][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:29:13,040][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:29:13,535][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:29:14,032][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:29:14,529][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:29:15,029][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:29:15,532][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:29:16,028][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:29:16,524][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:29:17,024][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:29:17,522][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:29:18,019][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:29:18,515][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:29:19,015][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:29:19,536][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:29:20,036][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:29:20,538][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:29:21,040][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:29:21,539][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:29:22,041][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:29:22,545][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:29:23,047][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:29:23,551][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:29:24,053][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:29:24,554][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:29:25,054][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:29:25,554][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:29:26,062][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:29:26,569][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:29:27,067][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:29:27,573][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:29:28,076][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:29:28,583][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:29:29,085][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:29:29,588][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:29:30,091][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:29:30,587][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:29:31,088][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:29:31,598][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:29:32,099][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:29:32,598][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:29:33,096][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:29:33,599][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:29:34,110][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:29:34,613][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:29:35,116][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:29:35,626][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:29:36,130][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:29:36,654][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:29:37,156][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:29:37,659][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:29:38,168][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:29:38,667][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:29:39,168][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:29:39,670][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:29:40,170][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:29:40,672][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:29:41,176][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:29:41,677][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:29:42,181][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10065 tokens. -[2025-11-10 03:29:43,079][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 55.60%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:32 -[2025-11-10 03:29:43,798][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:29:43,800][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:29:43,802][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:29:44,683][__main__][INFO] - Iteration 584 took 1m 8s (47.25% Gen, 51.46% Train). Generation: 32s, Training: 35s. Estimated remaining time: 47h 39m 54s. Estimated total time: 57h 10m 42s. Time estimates for 10 more iterations: 11m 26s, 100 more iterations: 1h 54m 21s, 500 more iterations: 9h 31m 47s. -[2025-11-10 03:29:44,685][__main__][INFO] - Starting iteration 584. -[2025-11-10 03:29:45,158][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 58 and human policies 1. -[2025-11-10 03:29:45,159][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:29:54,420][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 03:30:04,665][__main__][INFO] - Number of regex retries in iteration 584: 1 -[2025-11-10 03:30:04,666][__main__][INFO] - agents played in iteration 584 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:30:05,438][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:30:05,466][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:30:05,492][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:30:05,516][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:30:05,516][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:30:05,517][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:30:06,279][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:30:06,736][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:30:07,240][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:30:07,737][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:30:08,233][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:30:08,736][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:30:09,233][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:30:09,744][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:30:10,239][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:30:10,738][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:30:11,234][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:30:11,728][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:30:12,225][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:30:12,721][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:30:13,215][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:30:13,721][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:30:14,221][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:30:14,714][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:30:15,223][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:30:15,717][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:30:16,211][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:30:16,705][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:30:17,201][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:30:17,699][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:30:18,198][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:30:18,692][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:30:19,187][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:30:19,684][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:30:20,181][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:30:20,676][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:30:21,173][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:30:21,674][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:30:22,170][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:30:22,668][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:30:23,178][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:30:23,681][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:30:24,183][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:30:24,686][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:30:25,186][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:30:25,693][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:30:26,195][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:30:26,693][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:30:27,198][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:30:27,697][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:30:28,205][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:30:28,706][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:30:29,206][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:30:29,708][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:30:30,207][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:30:30,706][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:30:31,266][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:30:32,283][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:30:32,784][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:30:33,285][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:30:33,787][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:30:34,289][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:30:34,796][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:30:35,301][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:30:35,803][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:30:36,308][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:30:36,808][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:30:37,309][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:30:37,810][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:30:38,319][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:30:38,820][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10051 tokens. -[2025-11-10 03:30:39,696][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.09%, Current % of VRAM taken: 55.58%, Block Peak % of device VRAM: 62.17%, ΔTime: 00:00:33 -[2025-11-10 03:30:40,347][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:30:40,349][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:30:40,351][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:30:41,164][__main__][INFO] - Iteration 585 took 56s (34.83% Gen, 63.71% Train). Generation: 19s, Training: 35s. Estimated remaining time: 37h 8m 37s. Estimated total time: 46h 40m 21s. Time estimates for 10 more iterations: 9m 20s, 100 more iterations: 1h 33m 20s, 500 more iterations: 7h 46m 43s. -[2025-11-10 03:30:41,167][__main__][INFO] - Starting iteration 585. -[2025-11-10 03:30:41,630][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 58 and human policies 1. -[2025-11-10 03:30:41,631][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:31:08,327][__main__][INFO] - Number of regex retries in iteration 585: 0 -[2025-11-10 03:31:08,328][__main__][INFO] - agents played in iteration 585 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:31:09,177][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:31:09,203][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:31:09,229][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:31:09,251][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:31:09,252][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:31:09,252][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:31:10,011][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:31:10,468][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:31:10,971][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:31:11,477][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:31:11,973][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:31:12,470][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:31:12,968][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:31:13,465][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:31:13,962][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:31:14,466][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:31:14,964][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:31:15,482][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:31:15,980][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:31:16,483][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:31:16,981][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:31:17,478][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:31:17,982][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:31:18,480][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:31:18,979][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:31:19,487][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:31:19,984][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:31:20,480][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:31:20,980][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:31:21,476][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:31:21,986][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:31:22,484][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:31:22,986][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:31:23,490][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:31:23,991][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:31:24,498][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:31:25,003][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:31:25,506][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:31:26,005][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:31:26,505][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:31:27,007][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:31:27,509][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:31:28,010][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:31:28,510][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:31:29,010][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:31:29,510][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:31:30,012][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:31:30,517][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:31:31,023][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:31:31,526][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:31:32,027][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:31:32,527][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:31:33,028][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:31:33,527][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:31:34,031][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:31:34,538][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:31:35,039][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:31:35,543][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:31:36,043][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:31:36,557][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:31:37,059][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:31:37,560][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:31:38,062][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:31:38,565][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:31:39,066][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:31:39,567][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:31:40,069][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:31:40,577][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:31:41,082][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:31:41,585][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:31:42,086][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10052 tokens. -[2025-11-10 03:31:42,980][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 55.56%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:32 -[2025-11-10 03:31:43,724][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:31:43,726][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:31:43,728][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:31:44,657][__main__][INFO] - Iteration 586 took 1m 3s (42.36% Gen, 56.17% Train). Generation: 26s, Training: 35s. Estimated remaining time: 42h 58m 36s. Estimated total time: 52h 31m 24s. Time estimates for 10 more iterations: 10m 30s, 100 more iterations: 1h 45m 2s, 500 more iterations: 8h 45m 14s. -[2025-11-10 03:31:44,660][__main__][INFO] - Starting iteration 586. -[2025-11-10 03:31:45,125][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 58 and human policies 1. -[2025-11-10 03:31:45,125][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:32:08,198][__main__][INFO] - Number of regex retries in iteration 586: 0 -[2025-11-10 03:32:08,198][__main__][INFO] - agents played in iteration 586 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:32:08,975][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:32:09,002][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:32:09,028][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:32:09,050][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:32:09,051][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:32:09,052][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:32:09,815][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:32:10,267][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:32:10,767][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:32:11,269][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:32:11,764][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:32:12,282][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:32:12,777][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:32:13,272][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:32:13,773][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:32:14,270][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:32:14,777][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:32:15,270][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:32:15,763][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:32:16,258][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:32:16,751][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:32:17,245][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:32:17,747][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:32:18,245][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:32:18,742][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:32:19,244][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:32:19,764][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:32:20,270][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:32:20,765][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:32:21,262][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:32:21,769][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:32:22,265][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:32:22,770][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:32:23,268][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:32:23,764][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:32:24,276][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:32:24,776][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:32:25,271][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:32:25,771][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:32:26,267][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:32:26,775][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:32:27,273][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:32:27,776][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:32:28,293][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:32:28,791][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:32:29,290][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:32:29,786][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:32:30,285][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:32:30,801][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:32:31,299][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:32:31,803][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:32:32,305][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:32:32,804][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:32:33,308][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:32:33,812][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:32:34,311][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:32:34,814][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:32:35,317][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:32:35,817][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:32:36,319][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:32:36,824][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:32:37,327][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:32:37,828][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:32:38,327][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:32:38,827][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:32:39,328][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:32:39,828][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:32:40,331][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:32:40,832][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:32:41,330][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:32:41,831][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9950 tokens. -[2025-11-10 03:32:42,713][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 55.57%, Block Peak % of device VRAM: 62.06%, ΔTime: 00:00:32 -[2025-11-10 03:32:43,453][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:32:43,455][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:32:43,456][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:32:44,365][__main__][INFO] - Iteration 587 took 59s (38.95% Gen, 59.52% Train). Generation: 23s, Training: 35s. Estimated remaining time: 39h 48m 16s. Estimated total time: 49h 22m 3s. Time estimates for 10 more iterations: 9m 52s, 100 more iterations: 1h 38m 44s, 500 more iterations: 8h 13m 40s. -[2025-11-10 03:32:44,367][__main__][INFO] - Starting iteration 587. -[2025-11-10 03:32:44,829][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 58 and human policies 1. -[2025-11-10 03:32:44,830][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:33:11,983][__main__][INFO] - Number of regex retries in iteration 587: 0 -[2025-11-10 03:33:11,985][__main__][INFO] - agents played in iteration 587 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:33:12,767][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:33:12,795][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:33:12,821][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:33:12,845][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:33:12,845][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:33:12,846][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:33:13,600][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:33:14,054][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:33:14,556][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:33:15,052][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:33:15,550][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:33:16,045][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:33:16,542][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:33:17,043][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:33:17,541][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:33:18,046][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:33:18,548][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:33:19,044][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:33:19,548][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:33:20,043][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:33:20,540][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:33:21,041][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:33:21,539][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:33:22,042][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:33:22,543][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:33:23,040][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:33:23,541][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:33:24,042][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:33:24,542][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:33:25,040][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:33:25,539][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:33:26,033][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:33:26,525][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:33:27,021][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:33:27,526][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:33:28,024][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:33:28,524][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:33:29,025][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:33:29,524][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:33:30,021][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:33:30,524][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:33:31,028][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:33:31,534][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:33:32,036][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:33:32,540][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:33:33,040][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:33:33,543][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:33:34,049][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:33:34,552][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:33:35,052][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:33:35,557][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:33:36,060][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:33:36,567][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:33:37,066][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:33:37,566][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:33:38,070][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:33:38,572][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:33:39,073][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:33:39,575][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:33:40,075][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:33:40,591][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:33:41,093][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:33:41,596][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:33:42,101][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:33:42,603][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:33:43,115][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:33:43,616][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:33:44,117][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:33:44,624][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:33:45,126][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:33:45,634][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10059 tokens. -[2025-11-10 03:33:46,554][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 55.69%, Block Peak % of device VRAM: 62.33%, ΔTime: 00:00:32 -[2025-11-10 03:33:47,197][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:33:47,199][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:33:47,202][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:33:48,031][__main__][INFO] - Iteration 588 took 1m 3s (42.96% Gen, 55.72% Train). Generation: 27s, Training: 35s. Estimated remaining time: 43h 5m 18s. Estimated total time: 52h 40m 9s. Time estimates for 10 more iterations: 10m 32s, 100 more iterations: 1h 45m 20s, 500 more iterations: 8h 46m 41s. -[2025-11-10 03:33:48,034][__main__][INFO] - Starting iteration 588. -[2025-11-10 03:33:48,505][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 58 and human policies 1. -[2025-11-10 03:33:48,506][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:34:08,989][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 03:34:14,987][__main__][INFO] - Number of regex retries in iteration 588: 1 -[2025-11-10 03:34:14,988][__main__][INFO] - agents played in iteration 588 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:34:15,793][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:34:15,821][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:34:15,849][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:34:15,873][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:34:15,873][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:34:15,874][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:34:16,669][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:34:17,125][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:34:17,625][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:34:18,121][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:34:18,622][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:34:19,123][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:34:19,624][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:34:20,121][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:34:20,619][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:34:21,116][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:34:21,613][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:34:22,113][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:34:22,607][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:34:23,103][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:34:23,615][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:34:24,111][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:34:24,607][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:34:25,108][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:34:25,603][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:34:26,114][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:34:26,616][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:34:27,117][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:34:27,615][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:34:28,110][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:34:28,604][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:34:29,103][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:34:29,599][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:34:30,101][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:34:30,598][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:34:31,094][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:34:31,601][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:34:32,096][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:34:32,598][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:34:33,099][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:34:33,601][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:34:34,121][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:34:34,624][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:34:35,121][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:34:35,624][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:34:36,127][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:34:36,633][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:34:37,137][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:34:37,638][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:34:38,141][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:34:38,646][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:34:39,151][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:34:39,652][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:34:40,156][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:34:40,663][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:34:41,166][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:34:41,666][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:34:42,169][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:34:42,671][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:34:43,173][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:34:43,676][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:34:44,177][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:34:44,682][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:34:45,183][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:34:45,685][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:34:46,187][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:34:46,687][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:34:47,192][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:34:47,706][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:34:48,209][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:34:48,724][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10066 tokens. -[2025-11-10 03:34:49,619][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.02%, Current % of VRAM taken: 55.51%, Block Peak % of device VRAM: 62.10%, ΔTime: 00:00:32 -[2025-11-10 03:34:50,382][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:34:50,384][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:34:50,386][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:34:51,301][__main__][INFO] - Iteration 589 took 1m 2s (42.17% Gen, 56.37% Train). Generation: 26s, Training: 35s. Estimated remaining time: 42h 43m 55s. Estimated total time: 52h 19m 50s. Time estimates for 10 more iterations: 10m 27s, 100 more iterations: 1h 44m 39s, 500 more iterations: 8h 43m 18s. -[2025-11-10 03:34:51,303][__main__][INFO] - Starting iteration 589. -[2025-11-10 03:34:51,797][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 58 and human policies 1. -[2025-11-10 03:34:51,798][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:35:17,361][__main__][INFO] - Number of regex retries in iteration 589: 0 -[2025-11-10 03:35:17,364][__main__][INFO] - agents played in iteration 589 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:35:18,127][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:35:18,150][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:35:18,172][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:35:18,195][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:35:18,195][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:35:18,197][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:35:18,959][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:35:19,417][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:35:19,916][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:35:20,433][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:35:20,927][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:35:21,423][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:35:21,921][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:35:22,417][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:35:22,913][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:35:23,407][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:35:23,905][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:35:24,401][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:35:24,896][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:35:25,391][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:35:25,887][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:35:26,384][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:35:26,882][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:35:27,384][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:35:27,882][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:35:28,379][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:35:28,878][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:35:29,378][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:35:29,875][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:35:30,373][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:35:30,870][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:35:31,365][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:35:31,859][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:35:32,357][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:35:32,852][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:35:33,347][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:35:33,845][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:35:34,341][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:35:34,837][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:35:35,332][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:35:35,829][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:35:36,329][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:35:36,826][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:35:37,331][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:35:37,832][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:35:38,332][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:35:38,852][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:35:39,349][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:35:39,848][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:35:40,349][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:35:40,851][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:35:41,366][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:35:41,868][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:35:42,369][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:35:42,876][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:35:43,379][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:35:43,877][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:35:44,377][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:35:44,875][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:35:45,374][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:35:45,876][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:35:46,378][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:35:46,880][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:35:47,384][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:35:47,889][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:35:48,391][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:35:48,901][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:35:49,407][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:35:49,909][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:35:50,416][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:35:50,927][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10092 tokens. -[2025-11-10 03:35:51,828][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 55.68%, Block Peak % of device VRAM: 62.23%, ΔTime: 00:00:32 -[2025-11-10 03:35:52,479][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:35:52,481][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:35:52,483][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:35:53,247][__main__][INFO] - Iteration 590 took 1m 1s (41.60% Gen, 57.15% Train). Generation: 25s, Training: 35s. Estimated remaining time: 41h 35m 34s. Estimated total time: 51h 12m 30s. Time estimates for 10 more iterations: 10m 14s, 100 more iterations: 1h 42m 25s, 500 more iterations: 8h 32m 5s. -[2025-11-10 03:35:53,249][__main__][INFO] - Starting iteration 590. -[2025-11-10 03:35:53,722][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 58 and human policies 1. -[2025-11-10 03:35:53,723][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:36:08,841][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 03:36:19,796][__main__][INFO] - Number of regex retries in iteration 590: 1 -[2025-11-10 03:36:19,796][__main__][INFO] - agents played in iteration 590 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:36:20,603][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:36:20,627][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:36:20,650][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:36:20,672][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:36:20,673][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:36:20,674][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:36:21,449][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:36:21,903][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:36:22,406][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:36:22,903][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:36:23,401][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:36:23,899][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:36:24,396][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:36:24,892][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:36:25,388][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:36:25,888][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:36:26,385][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:36:26,885][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:36:27,391][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:36:27,892][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:36:28,388][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:36:28,885][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:36:29,381][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:36:29,882][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:36:30,380][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:36:30,877][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:36:31,371][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:36:31,868][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:36:32,370][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:36:32,868][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:36:33,364][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:36:33,859][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:36:34,358][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:36:34,853][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:36:35,348][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:36:35,847][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:36:36,345][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:36:36,843][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:36:37,343][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:36:37,861][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:36:38,354][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:36:38,850][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:36:39,347][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:36:39,844][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:36:40,358][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:36:40,859][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:36:41,356][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:36:41,852][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:36:42,349][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:36:42,857][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:36:43,359][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:36:43,858][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:36:44,371][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:36:44,882][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:36:45,394][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:36:45,893][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:36:46,390][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:36:46,893][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:36:47,397][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:36:47,901][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:36:48,410][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:36:48,917][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:36:49,425][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:36:49,928][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:36:50,428][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:36:50,933][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:36:51,436][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:36:51,938][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:36:52,438][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:36:52,939][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:36:53,459][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10128 tokens. -[2025-11-10 03:36:54,362][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 55.66%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:32 -[2025-11-10 03:36:55,121][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:36:55,122][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:36:55,124][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:36:56,947][__main__][INFO] - Iteration 591 took 1m 3s (41.24% Gen, 55.88% Train). Generation: 26s, Training: 35s. Estimated remaining time: 43h 3m 17s. Estimated total time: 52h 41m 17s. Time estimates for 10 more iterations: 10m 32s, 100 more iterations: 1h 45m 22s, 500 more iterations: 8h 46m 52s. -[2025-11-10 03:36:56,950][__main__][INFO] - Starting iteration 591. -[2025-11-10 03:36:57,439][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 59 and human policies 1. -[2025-11-10 03:36:57,440][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:37:26,227][__main__][INFO] - Number of regex retries in iteration 591: 0 -[2025-11-10 03:37:26,229][__main__][INFO] - agents played in iteration 591 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:37:26,990][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:37:27,012][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:37:27,035][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:37:27,057][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:37:27,057][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:37:27,058][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:37:27,815][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:37:28,274][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:37:28,781][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:37:29,278][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:37:29,775][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:37:30,274][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:37:30,771][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:37:31,268][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:37:31,765][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:37:32,263][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:37:32,781][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:37:33,279][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:37:33,775][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:37:34,271][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:37:34,768][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:37:35,273][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:37:35,769][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:37:36,265][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:37:36,778][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:37:37,274][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:37:37,769][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:37:38,268][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:37:38,765][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:37:39,273][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:37:39,768][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:37:40,264][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:37:40,778][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:37:41,275][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:37:41,770][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:37:42,267][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:37:42,762][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:37:43,276][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:37:43,777][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:37:44,278][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:37:44,787][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:37:45,287][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:37:45,792][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:37:46,285][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:37:46,785][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:37:47,294][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:37:47,794][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:37:48,294][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:37:48,794][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:37:49,299][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:37:49,803][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:37:50,308][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:37:50,810][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:37:51,317][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:37:51,821][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:37:52,325][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:37:52,829][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:37:53,331][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:37:53,835][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:37:54,337][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:37:54,840][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:37:55,346][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:37:55,849][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:37:56,363][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:37:56,870][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:37:57,370][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:37:57,886][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:37:58,386][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:37:58,886][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:37:59,386][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:37:59,886][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10018 tokens. -[2025-11-10 03:38:00,792][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 55.61%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:32 -[2025-11-10 03:38:01,446][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:38:01,447][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:38:01,449][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:38:02,278][__main__][INFO] - Iteration 592 took 1m 4s (44.40% Gen, 54.32% Train). Generation: 28s, Training: 35s. Estimated remaining time: 44h 22m 53s. Estimated total time: 54h 1m 58s. Time estimates for 10 more iterations: 10m 48s, 100 more iterations: 1h 48m 3s, 500 more iterations: 9h 0m 19s. -[2025-11-10 03:38:02,281][__main__][INFO] - Starting iteration 592. -[2025-11-10 03:38:02,846][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 59 and human policies 1. -[2025-11-10 03:38:02,847][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:38:28,453][__main__][INFO] - Number of regex retries in iteration 592: 0 -[2025-11-10 03:38:28,454][__main__][INFO] - agents played in iteration 592 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:38:29,340][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:38:29,367][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:38:29,393][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:38:29,416][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:38:29,416][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:38:29,417][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:38:30,200][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:38:30,660][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:38:31,167][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:38:31,679][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:38:32,182][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:38:32,681][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:38:33,180][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:38:33,680][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:38:34,181][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:38:34,675][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:38:35,173][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:38:35,684][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:38:36,183][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:38:36,680][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:38:37,179][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:38:37,673][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:38:38,169][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:38:38,665][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:38:39,161][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:38:39,676][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:38:40,170][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:38:40,665][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:38:41,171][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:38:41,669][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:38:42,175][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:38:42,670][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:38:43,166][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:38:43,677][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:38:44,174][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:38:44,671][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:38:45,169][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:38:45,664][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:38:46,175][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:38:46,674][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:38:47,172][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:38:47,676][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:38:48,175][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:38:48,681][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:38:49,182][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:38:49,681][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:38:50,196][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:38:50,691][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:38:51,188][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:38:51,693][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:38:52,192][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:38:52,690][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:38:53,192][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:38:53,696][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:38:54,189][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:38:54,689][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:38:55,196][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:38:55,696][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:38:56,199][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:38:56,707][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:38:57,208][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:38:57,710][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:38:58,210][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:38:58,714][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:38:59,221][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:38:59,724][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:39:00,226][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:39:00,728][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:39:01,229][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:39:01,733][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:39:02,234][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10111 tokens. -[2025-11-10 03:39:03,138][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 55.65%, Block Peak % of device VRAM: 62.45%, ΔTime: 00:00:32 -[2025-11-10 03:39:03,844][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:39:03,847][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:39:03,849][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:39:04,764][__main__][INFO] - Iteration 593 took 1m 1s (41.36% Gen, 57.16% Train). Generation: 25s, Training: 35s. Estimated remaining time: 41h 55m 48s. Estimated total time: 51h 35m 56s. Time estimates for 10 more iterations: 10m 19s, 100 more iterations: 1h 43m 11s, 500 more iterations: 8h 35m 59s. -[2025-11-10 03:39:04,766][__main__][INFO] - Starting iteration 593. -[2025-11-10 03:39:05,234][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 59 and human policies 1. -[2025-11-10 03:39:05,235][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:39:31,801][__main__][INFO] - Number of regex retries in iteration 593: 0 -[2025-11-10 03:39:31,803][__main__][INFO] - agents played in iteration 593 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:39:32,585][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:39:32,608][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:39:32,630][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:39:32,652][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:39:32,652][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:39:32,653][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:39:33,420][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:39:33,875][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:39:34,373][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:39:34,874][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:39:35,368][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:39:35,863][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:39:36,363][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:39:36,855][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:39:37,350][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:39:37,846][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:39:38,344][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:39:38,840][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:39:39,336][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:39:39,837][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:39:40,332][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:39:40,834][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:39:41,331][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:39:41,826][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:39:42,323][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:39:42,820][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:39:43,315][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:39:43,819][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:39:44,319][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:39:44,813][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:39:45,313][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:39:45,809][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:39:46,305][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:39:46,801][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:39:47,301][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:39:47,805][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:39:48,301][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:39:48,796][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:39:49,294][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:39:49,789][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:39:50,283][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:39:50,779][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:39:51,279][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:39:51,800][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:39:52,298][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:39:52,796][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:39:53,295][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:39:53,795][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:39:54,308][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:39:54,813][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:39:55,315][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:39:55,817][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:39:56,317][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:39:56,816][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:39:57,320][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:39:57,823][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:39:58,357][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:39:58,861][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:39:59,363][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:39:59,874][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:40:00,376][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:40:00,882][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:40:01,385][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:40:01,886][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:40:02,404][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:40:02,904][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:40:03,407][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:40:03,912][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:40:04,411][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:40:04,922][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:40:05,423][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10077 tokens. -[2025-11-10 03:40:06,331][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 55.61%, Block Peak % of device VRAM: 62.13%, ΔTime: 00:00:32 -[2025-11-10 03:40:06,978][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:40:06,980][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:40:06,983][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:40:07,810][__main__][INFO] - Iteration 594 took 1m 2s (42.46% Gen, 56.22% Train). Generation: 26s, Training: 35s. Estimated remaining time: 42h 27m 38s. Estimated total time: 52h 8m 49s. Time estimates for 10 more iterations: 10m 25s, 100 more iterations: 1h 44m 17s, 500 more iterations: 8h 41m 28s. -[2025-11-10 03:40:07,812][__main__][INFO] - Starting iteration 594. -[2025-11-10 03:40:08,276][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 59 and human policies 1. -[2025-11-10 03:40:08,277][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:40:31,231][__main__][INFO] - Number of regex retries in iteration 594: 0 -[2025-11-10 03:40:31,232][__main__][INFO] - agents played in iteration 594 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:40:32,115][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:40:32,142][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:40:32,169][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:40:32,192][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:40:32,193][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:40:32,194][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:40:32,973][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:40:33,430][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:40:33,937][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:40:34,444][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:40:34,947][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:40:35,447][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:40:35,944][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:40:36,443][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:40:36,944][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:40:37,440][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:40:37,937][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:40:38,431][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:40:38,932][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:40:39,430][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:40:39,925][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:40:40,424][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:40:40,922][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:40:41,420][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:40:41,927][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:40:42,425][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:40:42,924][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:40:43,418][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:40:43,913][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:40:44,409][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:40:44,906][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:40:45,403][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:40:45,915][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:40:46,410][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:40:46,905][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:40:47,404][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:40:47,902][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:40:48,409][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:40:48,906][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:40:49,406][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:40:49,923][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:40:50,421][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:40:50,924][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:40:51,423][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:40:51,919][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:40:52,426][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:40:52,922][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:40:53,425][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:40:53,916][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:40:54,410][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:40:54,915][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:40:55,414][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:40:55,912][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:40:56,423][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:40:56,921][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:40:57,420][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:40:57,919][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:40:58,418][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:40:58,924][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:40:59,428][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:40:59,927][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:41:00,436][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:41:00,940][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:41:01,443][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:41:01,945][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:41:02,447][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:41:02,948][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:41:03,450][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:41:03,948][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:41:04,451][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:41:04,951][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10061 tokens. -[2025-11-10 03:41:05,838][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 55.58%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 -[2025-11-10 03:41:06,546][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:41:06,548][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:41:06,549][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:41:07,458][__main__][INFO] - Iteration 595 took 59s (38.79% Gen, 59.68% Train). Generation: 22s, Training: 35s. Estimated remaining time: 39h 36m 56s. Estimated total time: 49h 19m 7s. Time estimates for 10 more iterations: 9m 51s, 100 more iterations: 1h 38m 38s, 500 more iterations: 8h 13m 11s. -[2025-11-10 03:41:07,460][__main__][INFO] - Starting iteration 595. -[2025-11-10 03:41:07,951][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 59 and human policies 1. -[2025-11-10 03:41:07,952][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:41:31,183][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 1 y book, 10 z balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 03:41:32,104][__main__][INFO] - Number of regex retries in iteration 595: 1 -[2025-11-10 03:41:32,105][__main__][INFO] - agents played in iteration 595 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:41:32,882][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:41:32,910][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:41:32,936][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:41:32,959][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:41:32,960][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:41:32,960][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:41:33,740][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:41:34,194][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:41:34,696][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:41:35,194][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:41:35,699][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:41:36,196][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:41:36,691][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:41:37,187][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:41:37,684][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:41:38,186][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:41:38,700][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:41:39,200][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:41:39,702][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:41:40,211][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:41:40,710][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:41:41,233][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:41:41,731][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:41:42,225][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:41:42,728][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:41:43,224][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:41:43,730][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:41:44,226][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:41:44,722][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:41:45,236][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:41:45,729][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:41:46,224][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:41:46,723][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:41:47,219][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:41:47,734][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:41:48,229][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:41:48,727][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:41:49,228][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:41:49,724][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:41:50,231][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:41:50,726][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:41:51,222][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:41:51,731][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:41:52,227][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:41:52,724][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:41:53,223][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:41:53,722][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:41:54,222][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:41:54,716][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:41:55,212][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:41:55,706][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:41:56,201][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:41:56,694][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:41:57,192][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:41:57,689][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:41:58,188][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:41:58,686][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:41:59,185][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:41:59,687][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:42:00,717][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:42:01,236][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:42:01,733][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:42:02,249][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:42:02,748][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:42:03,249][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:42:03,748][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:42:04,250][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:42:04,757][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:42:05,257][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:42:05,757][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:42:06,259][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10084 tokens. -[2025-11-10 03:42:07,169][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 55.60%, Block Peak % of device VRAM: 62.33%, ΔTime: 00:00:33 -[2025-11-10 03:42:07,822][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:42:07,825][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:42:07,827][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:42:08,708][__main__][INFO] - Iteration 596 took 1m 0s (39.75% Gen, 58.79% Train). Generation: 24s, Training: 35s. Estimated remaining time: 40h 54m 39s. Estimated total time: 50h 37m 51s. Time estimates for 10 more iterations: 10m 7s, 100 more iterations: 1h 41m 15s, 500 more iterations: 8h 26m 18s. -[2025-11-10 03:42:08,711][__main__][INFO] - Starting iteration 596. -[2025-11-10 03:42:09,188][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 59 and human policies 1. -[2025-11-10 03:42:09,189][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:42:37,540][__main__][INFO] - Number of regex retries in iteration 596: 0 -[2025-11-10 03:42:37,541][__main__][INFO] - agents played in iteration 596 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:42:38,310][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:42:38,337][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:42:38,363][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:42:38,386][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:42:38,386][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:42:38,387][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:42:39,184][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:42:39,643][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:42:40,149][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:42:40,651][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:42:41,157][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:42:41,658][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:42:42,164][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:42:42,664][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:42:43,164][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:42:43,662][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:42:44,165][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:42:44,666][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:42:45,168][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:42:45,669][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:42:46,166][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:42:46,666][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:42:47,162][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:42:47,658][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:42:48,157][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:42:48,652][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:42:49,148][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:42:49,645][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:42:50,143][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:42:50,642][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:42:51,138][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:42:51,635][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:42:52,143][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:42:52,638][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:42:53,141][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:42:53,640][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:42:54,136][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:42:54,632][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:42:55,130][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:42:55,626][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:42:56,146][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:42:56,645][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:42:57,143][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:42:57,655][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:42:58,150][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:42:58,646][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:42:59,142][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:42:59,639][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:43:00,145][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:43:00,643][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:43:01,139][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:43:01,636][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:43:02,133][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:43:02,629][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:43:03,136][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:43:03,637][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:43:04,145][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:43:04,643][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:43:05,141][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:43:05,656][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:43:06,153][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:43:06,652][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:43:07,149][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:43:07,646][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:43:08,163][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:43:08,664][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:43:09,167][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:43:09,667][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:43:10,168][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:43:10,672][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:43:11,173][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10075 tokens. -[2025-11-10 03:43:12,035][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 55.62%, Block Peak % of device VRAM: 62.36%, ΔTime: 00:00:32 -[2025-11-10 03:43:12,777][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:43:12,778][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:43:12,780][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:43:13,681][__main__][INFO] - Iteration 597 took 1m 4s (43.96% Gen, 54.64% Train). Generation: 28s, Training: 35s. Estimated remaining time: 44h 0m 24s. Estimated total time: 53h 44m 41s. Time estimates for 10 more iterations: 10m 44s, 100 more iterations: 1h 47m 29s, 500 more iterations: 8h 57m 26s. -[2025-11-10 03:43:13,684][__main__][INFO] - Starting iteration 597. -[2025-11-10 03:43:14,151][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 59 and human policies 1. -[2025-11-10 03:43:14,151][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:43:38,030][__main__][INFO] - Number of regex retries in iteration 597: 0 -[2025-11-10 03:43:38,031][__main__][INFO] - agents played in iteration 597 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:43:38,830][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:43:38,859][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:43:38,885][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:43:38,908][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:43:38,909][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:43:38,910][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:43:39,688][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:43:40,140][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:43:40,644][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:43:41,144][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:43:41,640][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:43:42,141][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:43:42,639][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:43:43,135][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:43:43,632][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:43:44,130][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:43:44,626][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:43:45,125][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:43:45,623][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:43:46,124][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:43:46,621][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:43:47,118][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:43:47,627][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:43:48,129][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:43:48,635][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:43:49,131][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:43:49,627][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:43:50,123][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:43:50,621][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:43:51,116][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:43:51,612][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:43:52,109][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:43:52,605][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:43:53,101][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:43:53,596][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:43:54,093][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:43:54,588][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:43:55,090][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:43:55,585][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:43:56,081][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:43:56,581][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:43:57,075][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:43:57,570][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:43:58,077][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:43:58,571][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:43:59,065][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:43:59,559][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:44:00,054][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:44:00,566][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:44:01,063][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:44:01,559][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:44:02,057][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:44:02,552][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:44:03,062][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:44:03,557][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:44:04,051][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:44:04,562][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:44:05,057][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:44:05,551][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:44:06,046][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:44:06,545][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:44:07,055][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:44:07,555][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:44:08,054][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:44:08,555][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:44:09,054][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:44:09,552][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:44:10,055][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:44:10,558][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:44:11,072][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:44:11,572][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9980 tokens. -[2025-11-10 03:44:12,441][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 55.58%, Block Peak % of device VRAM: 62.11%, ΔTime: 00:00:32 -[2025-11-10 03:44:13,175][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:44:13,177][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:44:13,179][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:44:14,186][__main__][INFO] - Iteration 598 took 1m 0s (39.77% Gen, 58.55% Train). Generation: 23s, Training: 35s. Estimated remaining time: 40h 16m 32s. Estimated total time: 50h 1m 49s. Time estimates for 10 more iterations: 10m 0s, 100 more iterations: 1h 40m 3s, 500 more iterations: 8h 20m 18s. -[2025-11-10 03:44:14,212][__main__][INFO] - Starting iteration 598. -[2025-11-10 03:44:14,959][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 59 and human policies 1. -[2025-11-10 03:44:14,961][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:44:41,773][__main__][INFO] - Number of regex retries in iteration 598: 0 -[2025-11-10 03:44:41,774][__main__][INFO] - agents played in iteration 598 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:44:42,658][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:44:42,682][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:44:42,707][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:44:42,729][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:44:42,729][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:44:42,730][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:44:43,520][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:44:43,975][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:44:44,478][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:44:44,987][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:44:45,485][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:44:45,985][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:44:46,481][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:44:46,982][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:44:47,482][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:44:47,983][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:44:48,484][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:44:48,980][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:44:49,474][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:44:49,976][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:44:50,474][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:44:50,969][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:44:51,472][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:44:51,970][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:44:52,470][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:44:52,975][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:44:53,473][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:44:53,977][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:44:54,475][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:44:54,977][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:44:55,482][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:44:55,979][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:44:56,476][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:44:56,970][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:44:57,463][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:44:57,958][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:44:58,452][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:44:58,948][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:44:59,443][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:44:59,940][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:45:00,436][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:45:00,932][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:45:01,429][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:45:01,928][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:45:02,425][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:45:02,924][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:45:03,425][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:45:03,922][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:45:04,419][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:45:04,917][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:45:05,413][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:45:05,911][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:45:06,408][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:45:06,905][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:45:07,404][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:45:07,902][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:45:08,398][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:45:08,895][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:45:09,391][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:45:09,892][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:45:10,393][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:45:10,893][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:45:11,395][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:45:11,896][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:45:12,396][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:45:12,894][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:45:13,395][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:45:13,900][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:45:14,400][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:45:14,904][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:45:15,405][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10054 tokens. -[2025-11-10 03:45:16,273][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 55.55%, Block Peak % of device VRAM: 62.20%, ΔTime: 00:00:32 -[2025-11-10 03:45:16,912][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:45:16,914][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:45:16,916][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:45:17,919][__main__][INFO] - Iteration 599 took 1m 2s (42.59% Gen, 55.82% Train). Generation: 26s, Training: 35s. Estimated remaining time: 42h 41m 39s. Estimated total time: 52h 27m 59s. Time estimates for 10 more iterations: 10m 29s, 100 more iterations: 1h 44m 55s, 500 more iterations: 8h 44m 39s. -[2025-11-10 03:45:17,921][__main__][INFO] - Starting iteration 599. -[2025-11-10 03:45:18,389][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 59 and human policies 1. -[2025-11-10 03:45:18,391][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:45:43,393][__main__][INFO] - Number of regex retries in iteration 599: 0 -[2025-11-10 03:45:43,393][__main__][INFO] - agents played in iteration 599 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:45:44,198][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:45:44,228][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:45:44,254][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:45:44,277][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:45:44,277][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:45:44,278][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:45:45,051][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:45:45,511][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:45:46,015][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:45:46,514][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:45:47,017][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:45:47,514][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:45:48,016][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:45:48,513][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:45:49,015][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:45:49,514][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:45:50,017][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:45:50,516][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:45:51,013][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:45:51,516][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:45:52,018][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:45:52,517][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:45:53,014][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:45:53,519][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:45:54,016][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:45:54,514][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:45:55,011][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:45:55,512][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:45:56,013][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:45:56,517][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:45:57,017][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:45:57,520][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:45:58,020][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:45:58,522][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:45:59,022][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:45:59,521][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:46:00,021][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:46:00,522][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:46:01,022][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:46:01,519][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:46:02,016][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:46:02,514][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:46:03,014][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:46:03,510][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:46:04,009][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:46:04,506][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:46:05,004][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:46:05,500][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:46:05,995][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:46:06,500][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:46:06,995][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:46:07,493][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:46:07,987][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:46:08,482][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:46:08,987][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:46:09,483][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:46:09,990][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:46:10,515][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:46:11,012][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:46:11,514][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:46:12,011][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:46:12,507][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:46:13,006][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:46:13,508][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:46:14,008][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:46:14,509][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:46:15,016][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:46:15,520][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:46:16,025][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:46:16,529][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:46:17,030][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10100 tokens. -[2025-11-10 03:46:17,899][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 55.66%, Block Peak % of device VRAM: 62.23%, ΔTime: 00:00:32 -[2025-11-10 03:46:18,669][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:46:18,671][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:46:18,672][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:46:19,572][__main__][INFO] - Iteration 600 took 1m 1s (40.86% Gen, 57.66% Train). Generation: 25s, Training: 35s. Estimated remaining time: 41h 11m 45s. Estimated total time: 50h 59m 7s. Time estimates for 10 more iterations: 10m 11s, 100 more iterations: 1h 41m 58s, 500 more iterations: 8h 29m 51s. -[2025-11-10 03:46:19,574][__main__][INFO] - Starting iteration 600. -[2025-11-10 03:46:20,058][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 59 and human policies 1. -[2025-11-10 03:46:20,059][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:46:48,620][__main__][INFO] - Number of regex retries in iteration 600: 0 -[2025-11-10 03:46:48,622][__main__][INFO] - agents played in iteration 600 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:46:49,422][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:46:49,445][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:46:49,468][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:46:49,490][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:46:49,491][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:46:49,492][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:46:50,358][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:46:50,817][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:46:51,324][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:46:51,828][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:46:52,327][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:46:52,826][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:46:53,325][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:46:53,827][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:46:54,327][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:46:54,824][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:46:55,322][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:46:55,822][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:46:56,323][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:46:56,822][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:46:57,318][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:46:57,818][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:46:58,315][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:46:58,814][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:46:59,315][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:46:59,813][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:47:00,326][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:47:00,827][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:47:01,327][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:47:01,833][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:47:02,338][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:47:02,845][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:47:03,350][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:47:03,846][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:47:04,347][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:47:04,846][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:47:05,347][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:47:05,847][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:47:06,349][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:47:06,848][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:47:07,350][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:47:07,848][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:47:08,348][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:47:08,846][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:47:09,342][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:47:09,842][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:47:10,338][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:47:10,835][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:47:11,336][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:47:11,835][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:47:12,335][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:47:12,834][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:47:13,331][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:47:13,835][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:47:14,332][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:47:14,830][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:47:15,328][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:47:15,824][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:47:16,326][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:47:16,824][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:47:17,324][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:47:17,824][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:47:18,324][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:47:18,827][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:47:19,326][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:47:19,828][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:47:20,332][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:47:20,830][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:47:21,331][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:47:21,833][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:47:22,335][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10145 tokens. -[2025-11-10 03:47:23,257][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 55.58%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 -[2025-11-10 03:47:23,893][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:47:23,896][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:47:23,898][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:47:25,746][__main__][INFO] - Iteration 601 took 1m 5s (43.48% Gen, 53.70% Train). Generation: 28s, Training: 35s. Estimated remaining time: 44h 55m 54s. Estimated total time: 54h 44m 23s. Time estimates for 10 more iterations: 10m 56s, 100 more iterations: 1h 49m 28s, 500 more iterations: 9h 7m 23s. -[2025-11-10 03:47:25,748][__main__][INFO] - Starting iteration 601. -[2025-11-10 03:47:26,233][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 60 and human policies 1. -[2025-11-10 03:47:26,234][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:47:57,043][__main__][INFO] - Number of regex retries in iteration 601: 0 -[2025-11-10 03:47:57,044][__main__][INFO] - agents played in iteration 601 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:47:57,852][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:47:57,880][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:47:57,907][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:47:57,931][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:47:57,931][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:47:57,932][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:47:58,718][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:47:59,172][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:47:59,676][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:48:00,184][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:48:00,683][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:48:01,192][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:48:01,694][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:48:02,191][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:48:02,689][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:48:03,190][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:48:03,695][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:48:04,198][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:48:04,694][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:48:05,194][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:48:05,691][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:48:06,196][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:48:06,699][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:48:07,197][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:48:07,706][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:48:08,201][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:48:08,699][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:48:09,204][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:48:09,703][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:48:10,201][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:48:10,697][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:48:11,192][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:48:11,705][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:48:12,201][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:48:12,696][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:48:13,208][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:48:13,704][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:48:14,206][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:48:14,702][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:48:15,196][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:48:15,703][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:48:16,201][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:48:16,695][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:48:17,206][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:48:17,700][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:48:18,211][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:48:18,706][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:48:19,201][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:48:19,707][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:48:20,202][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:48:20,702][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:48:21,203][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:48:21,703][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:48:22,206][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:48:22,706][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:48:23,206][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:48:23,710][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:48:24,211][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:48:24,725][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:48:25,231][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:48:25,734][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:48:26,237][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:48:26,737][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:48:27,238][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:48:27,736][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:48:28,236][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:48:28,736][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:48:29,236][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:48:29,742][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:48:30,246][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:48:30,746][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10065 tokens. -[2025-11-10 03:48:31,629][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 55.55%, Block Peak % of device VRAM: 62.14%, ΔTime: 00:00:32 -[2025-11-10 03:48:32,355][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:48:32,356][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:48:32,359][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:48:33,267][__main__][INFO] - Iteration 602 took 1m 7s (45.96% Gen, 52.68% Train). Generation: 30s, Training: 35s. Estimated remaining time: 46h 2m 7s. Estimated total time: 55h 51m 43s. Time estimates for 10 more iterations: 11m 10s, 100 more iterations: 1h 51m 43s, 500 more iterations: 9h 18m 37s. -[2025-11-10 03:48:33,269][__main__][INFO] - Starting iteration 602. -[2025-11-10 03:48:33,763][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 60 and human policies 1. -[2025-11-10 03:48:33,764][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:49:01,245][__main__][INFO] - Number of regex retries in iteration 602: 0 -[2025-11-10 03:49:01,248][__main__][INFO] - agents played in iteration 602 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:49:02,051][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:49:02,075][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:49:02,098][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:49:02,121][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:49:02,122][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:49:02,123][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:49:03,012][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:49:03,472][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:49:03,976][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:49:04,476][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:49:04,979][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:49:05,482][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:49:05,985][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:49:06,487][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:49:06,993][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:49:07,490][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:49:07,988][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:49:08,486][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:49:08,985][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:49:09,485][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:49:09,983][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:49:10,482][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:49:10,983][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:49:11,480][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:49:11,981][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:49:12,486][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:49:12,983][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:49:13,482][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:49:13,979][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:49:14,475][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:49:14,976][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:49:15,475][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:49:15,980][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:49:16,479][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:49:16,973][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:49:17,472][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:49:17,967][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:49:18,464][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:49:18,960][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:49:19,460][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:49:19,958][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:49:20,457][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:49:20,951][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:49:21,447][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:49:21,946][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:49:22,441][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:49:22,939][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:49:23,438][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:49:23,951][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:49:24,449][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:49:24,949][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:49:25,448][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:49:25,946][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:49:26,452][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:49:26,952][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:49:27,453][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:49:27,958][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:49:28,460][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:49:28,959][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:49:29,463][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:49:29,964][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:49:30,468][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:49:30,967][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:49:31,467][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:49:31,969][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:49:32,467][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:49:32,965][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:49:33,466][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:49:33,972][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:49:34,471][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:49:34,971][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10122 tokens. -[2025-11-10 03:49:35,853][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 55.60%, Block Peak % of device VRAM: 62.44%, ΔTime: 00:00:32 -[2025-11-10 03:49:36,490][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:49:36,492][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:49:36,494][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:49:37,302][__main__][INFO] - Iteration 603 took 1m 3s (43.25% Gen, 55.47% Train). Generation: 27s, Training: 35s. Estimated remaining time: 43h 6m 16s. Estimated total time: 52h 56m 56s. Time estimates for 10 more iterations: 10m 35s, 100 more iterations: 1h 45m 53s, 500 more iterations: 8h 49m 29s. -[2025-11-10 03:49:37,304][__main__][INFO] - Starting iteration 603. -[2025-11-10 03:49:37,768][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 60 and human policies 1. -[2025-11-10 03:49:37,768][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:49:59,314][__main__][INFO] - Number of regex retries in iteration 603: 0 -[2025-11-10 03:49:59,315][__main__][INFO] - agents played in iteration 603 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:50:00,215][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:50:00,238][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:50:00,261][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:50:00,284][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:50:00,285][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:50:00,286][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:50:01,147][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:50:01,604][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:50:02,125][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:50:02,627][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:50:03,128][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:50:03,631][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:50:04,130][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:50:04,640][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:50:05,142][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:50:05,647][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:50:06,153][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:50:06,653][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:50:07,154][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:50:07,654][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:50:08,154][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:50:08,655][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:50:09,152][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:50:09,651][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:50:10,147][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:50:10,646][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:50:11,147][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:50:11,646][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:50:12,143][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:50:12,652][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:50:13,150][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:50:13,652][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:50:14,155][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:50:14,655][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:50:15,157][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:50:15,655][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:50:16,155][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:50:16,653][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:50:17,151][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:50:17,651][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:50:18,151][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:50:18,650][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:50:19,148][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:50:19,645][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:50:20,143][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:50:20,649][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:50:21,150][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:50:21,653][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:50:22,147][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:50:22,645][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:50:23,144][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:50:23,638][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:50:24,135][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:50:24,631][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:50:25,126][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:50:25,639][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:50:26,133][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:50:26,632][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:50:27,131][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:50:27,626][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:50:28,130][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:50:28,624][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:50:29,125][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:50:29,627][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:50:30,130][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:50:30,633][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:50:31,135][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:50:31,633][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:50:32,134][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:50:32,637][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:50:33,139][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10085 tokens. -[2025-11-10 03:50:34,032][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 55.61%, Block Peak % of device VRAM: 62.06%, ΔTime: 00:00:32 -[2025-11-10 03:50:34,744][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:50:34,748][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:50:34,749][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:50:35,675][__main__][INFO] - Iteration 604 took 57s (37.21% Gen, 61.19% Train). Generation: 21s, Training: 35s. Estimated remaining time: 38h 23m 47s. Estimated total time: 48h 15m 25s. Time estimates for 10 more iterations: 9m 39s, 100 more iterations: 1h 36m 30s, 500 more iterations: 8h 2m 34s. -[2025-11-10 03:50:35,677][__main__][INFO] - Starting iteration 604. -[2025-11-10 03:50:36,171][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 60 and human policies 1. -[2025-11-10 03:50:36,172][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:50:58,282][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 03:51:03,673][__main__][INFO] - Number of regex retries in iteration 604: 1 -[2025-11-10 03:51:03,674][__main__][INFO] - agents played in iteration 604 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:51:04,469][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:51:04,493][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:51:04,518][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:51:04,540][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:51:04,541][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:51:04,542][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:51:05,385][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:51:05,844][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:51:06,360][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:51:06,860][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:51:07,364][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:51:07,873][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:51:08,374][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:51:08,882][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:51:09,385][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:51:09,884][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:51:10,386][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:51:10,885][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:51:11,386][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:51:11,887][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:51:12,388][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:51:12,891][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:51:13,394][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:51:13,894][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:51:14,395][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:51:14,896][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:51:15,391][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:51:15,888][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:51:16,385][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:51:16,888][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:51:17,383][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:51:17,881][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:51:18,378][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:51:18,875][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:51:19,371][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:51:19,867][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:51:20,364][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:51:20,869][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:51:21,365][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:51:21,863][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:51:22,361][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:51:22,859][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:51:23,367][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:51:23,865][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:51:24,371][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:51:24,868][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:51:25,365][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:51:25,862][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:51:26,362][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:51:26,862][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:51:27,372][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:51:27,868][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:51:28,367][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:51:28,867][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:51:29,365][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:51:29,867][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:51:30,399][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:51:30,899][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:51:31,406][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:51:31,909][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:51:32,412][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:51:32,914][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:51:33,418][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:51:33,917][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:51:34,417][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:51:34,918][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:51:35,430][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:51:35,929][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:51:36,431][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:51:36,935][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:51:37,435][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10056 tokens. -[2025-11-10 03:51:38,322][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 55.68%, Block Peak % of device VRAM: 62.33%, ΔTime: 00:00:32 -[2025-11-10 03:51:38,961][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:51:38,963][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:51:38,965][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:51:39,804][__main__][INFO] - Iteration 605 took 1m 3s (43.22% Gen, 55.46% Train). Generation: 27s, Training: 35s. Estimated remaining time: 43h 9m 0s. Estimated total time: 53h 1m 43s. Time estimates for 10 more iterations: 10m 36s, 100 more iterations: 1h 46m 3s, 500 more iterations: 8h 50m 17s. -[2025-11-10 03:51:39,807][__main__][INFO] - Starting iteration 605. -[2025-11-10 03:51:40,282][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 60 and human policies 1. -[2025-11-10 03:51:40,283][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:52:06,938][__main__][INFO] - Number of regex retries in iteration 605: 0 -[2025-11-10 03:52:06,939][__main__][INFO] - agents played in iteration 605 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:52:07,736][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:52:07,762][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:52:07,788][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:52:07,812][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:52:07,812][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:52:07,813][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:52:08,691][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:52:09,150][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:52:09,672][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:52:10,170][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:52:10,670][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:52:11,170][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:52:11,671][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:52:12,174][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:52:12,674][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:52:13,172][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:52:13,674][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:52:14,184][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:52:14,685][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:52:15,183][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:52:15,678][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:52:16,181][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:52:16,676][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:52:17,172][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:52:17,668][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:52:18,171][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:52:18,675][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:52:19,171][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:52:19,668][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:52:20,178][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:52:20,673][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:52:21,170][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:52:21,666][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:52:22,164][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:52:22,671][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:52:23,168][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:52:23,665][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:52:24,180][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:52:24,679][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:52:25,183][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:52:25,683][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:52:26,180][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:52:26,686][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:52:27,186][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:52:27,688][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:52:28,188][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:52:28,690][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:52:29,189][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:52:29,682][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:52:30,176][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:52:30,673][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:52:31,169][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:52:31,663][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:52:32,160][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:52:32,654][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:52:33,149][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:52:33,644][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:52:34,141][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:52:34,638][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:52:35,137][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:52:35,632][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:52:36,133][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:52:36,633][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:52:37,133][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:52:37,633][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:52:38,134][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:52:38,646][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:52:39,146][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:52:39,646][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:52:40,146][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:52:40,651][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9984 tokens. -[2025-11-10 03:52:41,562][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 55.59%, Block Peak % of device VRAM: 62.05%, ΔTime: 00:00:32 -[2025-11-10 03:52:42,361][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:52:42,362][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:52:42,364][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:52:43,228][__main__][INFO] - Iteration 606 took 1m 2s (42.34% Gen, 56.28% Train). Generation: 26s, Training: 35s. Estimated remaining time: 42h 33m 32s. Estimated total time: 52h 27m 18s. Time estimates for 10 more iterations: 10m 29s, 100 more iterations: 1h 44m 54s, 500 more iterations: 8h 44m 33s. -[2025-11-10 03:52:43,230][__main__][INFO] - Starting iteration 606. -[2025-11-10 03:52:43,726][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 60 and human policies 1. -[2025-11-10 03:52:43,728][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:53:04,302][__main__][INFO] - Number of regex retries in iteration 606: 0 -[2025-11-10 03:53:04,302][__main__][INFO] - agents played in iteration 606 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:53:05,099][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:53:05,125][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:53:05,151][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:53:05,174][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:53:05,175][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:53:05,175][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:53:06,084][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:53:06,542][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:53:07,055][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:53:07,562][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:53:08,063][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:53:08,575][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:53:09,493][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:53:10,245][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:53:10,750][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:53:11,264][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:53:11,765][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:53:12,267][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:53:12,768][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:53:13,269][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:53:13,774][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:53:14,275][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:53:14,773][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:53:15,286][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:53:15,788][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:53:16,290][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:53:16,794][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:53:17,297][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:53:17,801][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:53:18,302][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:53:18,804][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:53:19,307][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:53:19,808][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:53:20,310][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:53:20,809][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:53:21,313][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:53:21,814][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:53:22,317][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:53:22,819][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:53:23,318][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:53:23,813][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:53:24,313][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:53:24,808][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:53:25,305][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:53:25,805][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:53:26,308][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:53:26,805][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:53:27,303][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:53:27,803][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:53:28,303][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:53:28,803][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:53:29,302][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:53:29,802][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:53:30,300][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:53:30,799][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:53:31,298][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:53:31,802][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:53:32,304][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:53:32,806][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:53:33,307][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:53:33,807][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:53:34,307][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:53:34,811][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:53:35,311][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:53:35,809][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:53:36,309][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:53:36,807][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:53:37,304][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:53:37,803][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:53:38,299][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:53:38,805][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10139 tokens. -[2025-11-10 03:53:39,728][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 55.72%, Block Peak % of device VRAM: 62.20%, ΔTime: 00:00:33 -[2025-11-10 03:53:40,385][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:53:40,387][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:53:40,388][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:53:41,264][__main__][INFO] - Iteration 607 took 57s (35.76% Gen, 62.72% Train). Generation: 20s, Training: 36s. Estimated remaining time: 38h 2m 11s. Estimated total time: 47h 56m 56s. Time estimates for 10 more iterations: 9m 35s, 100 more iterations: 1h 35m 53s, 500 more iterations: 7h 59m 29s. -[2025-11-10 03:53:41,266][__main__][INFO] - Starting iteration 607. -[2025-11-10 03:53:41,760][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 60 and human policies 1. -[2025-11-10 03:53:41,761][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:54:14,728][__main__][INFO] - Number of regex retries in iteration 607: 0 -[2025-11-10 03:54:14,729][__main__][INFO] - agents played in iteration 607 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:54:15,528][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:54:15,551][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:54:15,578][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:54:15,602][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:54:15,602][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:54:15,603][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:54:16,452][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:54:16,913][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:54:17,417][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:54:17,917][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:54:18,424][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:54:18,924][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:54:19,429][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:54:19,938][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:54:20,439][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:54:20,940][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:54:21,441][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:54:21,938][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:54:22,436][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:54:22,937][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:54:23,439][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:54:23,937][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:54:24,433][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:54:24,932][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:54:25,431][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:54:25,929][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:54:26,426][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:54:26,928][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:54:27,426][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:54:27,923][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:54:28,421][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:54:28,923][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:54:29,422][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:54:29,924][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:54:30,424][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:54:30,923][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:54:31,423][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:54:31,922][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:54:32,424][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:54:32,923][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:54:33,425][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:54:33,929][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:54:34,427][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:54:34,925][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:54:35,423][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:54:35,924][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:54:36,432][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:54:36,929][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:54:37,427][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:54:37,923][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:54:38,419][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:54:38,914][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:54:39,410][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:54:39,905][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:54:40,407][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:54:40,904][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:54:41,400][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:54:41,898][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:54:42,400][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:54:42,900][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:54:43,399][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:54:43,905][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:54:44,413][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:54:44,913][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:54:45,414][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:54:45,916][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:54:46,415][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:54:46,918][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:54:47,417][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:54:47,921][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:54:48,420][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10030 tokens. -[2025-11-10 03:54:49,293][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.03%, Current % of VRAM taken: 55.53%, Block Peak % of device VRAM: 62.13%, ΔTime: 00:00:32 -[2025-11-10 03:54:50,030][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:54:50,031][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:54:50,033][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:54:50,981][__main__][INFO] - Iteration 608 took 1m 9s (47.63% Gen, 51.00% Train). Generation: 32s, Training: 35s. Estimated remaining time: 47h 45m 10s. Estimated total time: 57h 41m 4s. Time estimates for 10 more iterations: 11m 32s, 100 more iterations: 1h 55m 22s, 500 more iterations: 9h 36m 50s. -[2025-11-10 03:54:50,984][__main__][INFO] - Starting iteration 608. -[2025-11-10 03:54:51,449][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 60 and human policies 1. -[2025-11-10 03:54:51,450][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:55:16,465][__main__][INFO] - Number of regex retries in iteration 608: 0 -[2025-11-10 03:55:16,466][__main__][INFO] - agents played in iteration 608 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:55:17,270][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:55:17,293][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:55:17,317][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:55:17,340][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:55:17,340][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:55:17,341][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:55:18,224][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:55:18,686][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:55:19,190][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:55:19,709][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:55:20,210][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:55:20,716][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:55:21,216][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:55:21,720][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:55:22,227][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:55:22,730][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:55:23,228][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:55:23,729][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:55:24,228][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:55:24,725][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:55:25,230][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:55:25,727][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:55:26,226][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:55:26,730][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:55:27,231][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:55:27,735][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:55:28,234][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:55:28,733][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:55:29,233][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:55:29,732][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:55:30,232][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:55:30,728][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:55:31,226][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:55:31,730][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:55:32,232][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:55:32,729][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:55:33,228][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:55:33,729][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:55:34,230][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:55:34,733][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:55:35,228][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:55:35,726][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:55:36,223][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:55:36,719][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:55:37,214][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:55:37,710][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:55:38,211][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:55:38,710][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:55:39,207][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:55:39,714][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:55:40,212][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:55:40,711][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:55:41,210][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:55:41,711][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:55:42,218][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:55:42,720][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:55:43,226][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:55:43,727][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:55:44,229][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:55:44,733][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:55:45,234][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:55:45,735][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:55:46,242][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:55:46,746][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:55:47,250][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:55:47,754][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:55:48,260][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:55:48,765][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:55:49,266][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:55:49,767][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:55:50,286][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10164 tokens. -[2025-11-10 03:55:52,394][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 55.68%, Block Peak % of device VRAM: 62.28%, ΔTime: 00:00:34 -[2025-11-10 03:55:53,715][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:55:53,717][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:55:53,719][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:55:54,610][__main__][INFO] - Iteration 609 took 1m 3s (39.61% Gen, 58.98% Train). Generation: 25s, Training: 37s. Estimated remaining time: 42h 41m 7s. Estimated total time: 52h 38m 4s. Time estimates for 10 more iterations: 10m 31s, 100 more iterations: 1h 45m 16s, 500 more iterations: 8h 46m 20s. -[2025-11-10 03:55:54,614][__main__][INFO] - Starting iteration 609. -[2025-11-10 03:55:55,087][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 60 and human policies 1. -[2025-11-10 03:55:55,088][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:56:25,598][__main__][INFO] - Number of regex retries in iteration 609: 0 -[2025-11-10 03:56:25,599][__main__][INFO] - agents played in iteration 609 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:56:26,390][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:56:26,418][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:56:26,444][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:56:26,468][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:56:26,468][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:56:26,469][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:56:27,349][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:56:27,807][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:56:28,318][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:56:28,815][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:56:29,311][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:56:29,809][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:56:30,306][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:56:30,803][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:56:31,303][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:56:31,798][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:56:32,312][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:56:32,809][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:56:33,307][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:56:33,807][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:56:34,305][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:56:34,813][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:56:35,310][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:56:35,806][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:56:36,309][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:56:36,806][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:56:37,306][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:56:37,806][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:56:38,302][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:56:38,815][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:56:39,316][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:56:39,816][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:56:40,316][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:56:40,811][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:56:41,317][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:56:41,819][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:56:42,317][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:56:42,829][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:56:43,326][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:56:43,821][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:56:44,314][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:56:44,809][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:56:45,328][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:56:45,821][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:56:46,318][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:56:46,817][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:56:47,314][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:56:47,831][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:56:48,331][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:56:48,829][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:56:49,334][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:56:49,835][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:56:50,338][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:56:50,841][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:56:51,347][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:56:51,852][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:56:52,354][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:56:52,856][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:56:53,360][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:56:53,864][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:56:54,366][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:56:54,868][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:56:55,369][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:56:55,874][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:56:56,377][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:56:56,878][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:56:57,381][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:56:57,880][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:56:58,382][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:56:58,893][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:56:59,393][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10098 tokens. -[2025-11-10 03:57:00,281][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 55.63%, Block Peak % of device VRAM: 62.22%, ΔTime: 00:00:32 -[2025-11-10 03:57:01,081][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:57:01,084][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:57:01,086][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:57:02,026][__main__][INFO] - Iteration 610 took 1m 6s (45.58% Gen, 53.01% Train). Generation: 30s, Training: 35s. Estimated remaining time: 45h 48m 55s. Estimated total time: 55h 47m 0s. Time estimates for 10 more iterations: 11m 9s, 100 more iterations: 1h 51m 34s, 500 more iterations: 9h 17m 50s. -[2025-11-10 03:57:02,028][__main__][INFO] - Starting iteration 610. -[2025-11-10 03:57:02,620][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 60 and human policies 1. -[2025-11-10 03:57:02,621][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:57:27,667][__main__][INFO] - Number of regex retries in iteration 610: 0 -[2025-11-10 03:57:27,668][__main__][INFO] - agents played in iteration 610 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:57:28,452][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:57:28,481][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:57:28,507][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:57:28,531][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:57:28,532][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:57:28,533][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:57:29,393][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:57:30,022][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:57:30,545][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:57:31,047][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:57:31,550][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:57:32,052][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:57:32,559][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:57:33,060][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:57:33,561][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:57:34,058][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:57:34,559][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:57:35,060][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:57:35,557][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:57:36,053][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:57:36,551][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:57:37,052][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:57:37,550][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:57:38,047][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:57:38,545][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:57:39,042][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:57:39,541][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:57:40,038][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:57:40,535][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:57:41,044][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:57:41,543][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:57:42,046][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:57:42,547][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:57:43,046][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:57:43,544][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:57:44,044][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:57:44,546][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:57:45,052][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:57:45,553][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:57:46,051][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:57:46,551][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:57:47,045][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:57:47,543][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:57:48,039][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:57:48,534][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:57:49,044][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:57:49,541][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:57:50,041][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:57:50,544][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:57:51,044][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:57:51,549][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:57:52,048][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:57:52,549][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:57:53,052][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:57:53,558][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:57:54,060][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:57:54,562][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:57:55,063][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:57:55,567][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:57:56,070][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:57:56,570][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:57:57,072][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:57:57,593][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:57:58,092][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:57:58,591][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:57:59,098][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:57:59,603][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:58:00,811][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:58:01,723][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:58:02,230][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:58:02,734][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9983 tokens. -[2025-11-10 03:58:03,648][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 55.58%, Block Peak % of device VRAM: 62.10%, ΔTime: 00:00:34 -[2025-11-10 03:58:04,316][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:58:04,318][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:58:04,320][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:58:06,078][__main__][INFO] - Iteration 611 took 1m 3s (39.47% Gen, 57.76% Train). Generation: 25s, Training: 36s. Estimated remaining time: 42h 53m 48s. Estimated total time: 52h 52m 57s. Time estimates for 10 more iterations: 10m 34s, 100 more iterations: 1h 45m 45s, 500 more iterations: 8h 48m 49s. -[2025-11-10 03:58:06,083][__main__][INFO] - Starting iteration 611. -[2025-11-10 03:58:06,555][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 61 and human policies 1. -[2025-11-10 03:58:06,556][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:58:25,910][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 03:58:36,389][__main__][INFO] - Number of regex retries in iteration 611: 1 -[2025-11-10 03:58:36,390][__main__][INFO] - agents played in iteration 611 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:58:37,157][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:58:37,180][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:58:37,202][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:58:37,225][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:58:37,225][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:58:37,226][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:58:38,019][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:58:38,474][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:58:38,978][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:58:39,476][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:58:39,975][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:58:40,475][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:58:40,973][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:58:41,471][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:58:41,970][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:58:42,467][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:58:42,966][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:58:43,467][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:58:43,965][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:58:44,468][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:58:44,965][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:58:45,465][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:58:45,971][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:58:46,470][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:58:46,975][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:58:47,472][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:58:47,967][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:58:48,463][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:58:48,959][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:58:49,458][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:58:49,953][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:58:50,450][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:58:50,967][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:58:51,463][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:58:51,967][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:58:52,467][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:58:52,973][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:58:53,473][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:58:53,974][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:58:54,474][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:58:54,970][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:58:55,471][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:58:55,974][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:58:56,476][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:58:56,976][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:58:57,482][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:58:57,982][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:58:58,482][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:58:58,983][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:58:59,483][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:58:59,988][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:59:00,492][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:59:00,998][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 03:59:01,498][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 03:59:01,998][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 03:59:02,501][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 03:59:03,004][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 03:59:03,509][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 03:59:04,027][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 03:59:04,528][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 03:59:05,043][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 03:59:05,545][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 03:59:06,047][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 03:59:06,562][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 03:59:07,067][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 03:59:07,569][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 03:59:08,071][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 03:59:08,569][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 03:59:09,068][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 03:59:09,566][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 03:59:10,064][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10072 tokens. -[2025-11-10 03:59:10,912][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 55.58%, Block Peak % of device VRAM: 62.23%, ΔTime: 00:00:32 -[2025-11-10 03:59:11,691][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 03:59:11,693][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 03:59:11,695][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 03:59:12,594][__main__][INFO] - Iteration 612 took 1m 6s (45.18% Gen, 53.46% Train). Generation: 29s, Training: 35s. Estimated remaining time: 45h 1m 41s. Estimated total time: 55h 1m 57s. Time estimates for 10 more iterations: 11m 0s, 100 more iterations: 1h 50m 3s, 500 more iterations: 9h 10m 19s. -[2025-11-10 03:59:12,596][__main__][INFO] - Starting iteration 612. -[2025-11-10 03:59:13,062][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 61 and human policies 1. -[2025-11-10 03:59:13,062][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 03:59:34,955][__main__][INFO] - Number of regex retries in iteration 612: 0 -[2025-11-10 03:59:34,955][__main__][INFO] - agents played in iteration 612 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 03:59:35,827][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:59:35,855][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:59:35,881][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:59:35,905][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 03:59:35,905][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 03:59:35,906][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 03:59:36,696][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 03:59:37,149][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 03:59:37,654][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 03:59:38,150][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 03:59:38,648][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 03:59:39,146][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 03:59:39,648][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 03:59:40,147][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 03:59:40,662][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 03:59:41,160][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 03:59:41,659][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 03:59:42,162][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 03:59:42,659][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 03:59:43,166][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 03:59:43,667][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 03:59:44,167][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 03:59:44,675][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 03:59:45,173][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 03:59:45,678][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 03:59:46,175][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 03:59:46,675][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 03:59:47,175][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 03:59:47,675][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 03:59:48,172][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 03:59:48,678][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 03:59:49,177][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 03:59:49,673][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 03:59:50,169][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 03:59:50,666][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 03:59:51,171][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 03:59:51,672][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 03:59:52,175][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 03:59:52,675][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 03:59:53,175][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 03:59:53,679][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 03:59:54,178][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 03:59:54,674][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 03:59:55,174][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 03:59:55,671][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 03:59:56,169][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 03:59:56,668][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 03:59:57,170][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 03:59:57,673][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 03:59:58,173][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 03:59:58,678][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 03:59:59,183][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 03:59:59,684][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:00:00,187][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:00:00,688][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:00:01,192][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:00:01,721][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:00:02,223][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:00:02,725][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:00:03,233][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:00:03,736][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:00:04,236][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:00:04,738][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:00:05,245][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:00:05,744][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:00:06,248][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:00:06,748][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:00:07,255][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:00:07,761][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:00:08,282][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:00:08,781][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10094 tokens. -[2025-11-10 04:00:09,638][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 55.57%, Block Peak % of device VRAM: 62.17%, ΔTime: 00:00:32 -[2025-11-10 04:00:10,392][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:00:10,394][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:00:10,395][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:00:11,358][__main__][INFO] - Iteration 613 took 58s (37.55% Gen, 60.79% Train). Generation: 21s, Training: 35s. Estimated remaining time: 38h 33m 36s. Estimated total time: 48h 34m 50s. Time estimates for 10 more iterations: 9m 42s, 100 more iterations: 1h 37m 9s, 500 more iterations: 8h 5m 48s. -[2025-11-10 04:00:11,360][__main__][INFO] - Starting iteration 613. -[2025-11-10 04:00:11,821][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 61 and human policies 1. -[2025-11-10 04:00:11,822][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:00:35,413][__main__][INFO] - Number of regex retries in iteration 613: 0 -[2025-11-10 04:00:35,415][__main__][INFO] - agents played in iteration 613 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:00:36,261][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:00:36,302][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:00:36,333][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:00:36,361][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:00:36,362][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:00:36,363][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:00:37,248][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:00:37,706][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:00:38,213][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:00:38,717][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:00:39,217][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:00:39,718][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:00:40,222][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:00:40,724][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:00:41,224][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:00:41,732][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:00:42,230][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:00:42,732][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:00:43,230][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:00:43,734][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:00:44,233][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:00:44,729][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:00:45,228][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:00:45,725][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:00:46,230][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:00:46,726][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:00:47,252][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:00:47,748][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:00:48,246][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:00:48,743][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:00:49,248][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:00:49,745][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:00:50,245][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:00:50,743][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:00:51,242][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:00:51,741][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:00:52,239][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:00:52,738][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:00:53,250][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:00:53,749][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:00:54,246][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:00:54,746][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:00:55,251][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:00:55,758][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:00:56,262][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:00:56,763][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:00:57,267][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:00:57,767][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:00:58,268][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:00:58,772][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:00:59,270][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:00:59,773][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:01:00,275][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:01:00,774][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:01:01,272][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:01:01,771][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:01:02,287][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:01:02,788][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:01:03,290][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:01:03,791][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:01:04,290][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:01:04,797][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:01:05,299][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:01:05,799][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:01:06,299][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:01:06,799][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:01:07,300][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:01:07,797][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:01:08,301][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:01:08,802][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:01:09,306][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10067 tokens. -[2025-11-10 04:01:10,222][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 55.69%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:32 -[2025-11-10 04:01:10,869][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:01:10,871][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:01:10,874][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:01:11,716][__main__][INFO] - Iteration 614 took 59s (39.39% Gen, 59.20% Train). Generation: 23s, Training: 35s. Estimated remaining time: 39h 52m 29s. Estimated total time: 49h 54m 44s. Time estimates for 10 more iterations: 9m 58s, 100 more iterations: 1h 39m 49s, 500 more iterations: 8h 19m 7s. -[2025-11-10 04:01:11,718][__main__][INFO] - Starting iteration 614. -[2025-11-10 04:01:12,185][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 61 and human policies 1. -[2025-11-10 04:01:12,186][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:01:36,984][__main__][INFO] - Number of regex retries in iteration 614: 0 -[2025-11-10 04:01:36,985][__main__][INFO] - agents played in iteration 614 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:01:37,849][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:01:37,872][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:01:37,897][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:01:37,920][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:01:37,921][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:01:37,922][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:01:38,788][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:01:39,246][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:01:39,758][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:01:40,272][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:01:40,779][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:01:41,282][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:01:41,788][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:01:42,289][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:01:42,799][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:01:43,305][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:01:43,806][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:01:44,310][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:01:44,811][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:01:45,315][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:01:45,816][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:01:46,313][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:01:46,819][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:01:47,317][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:01:47,816][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:01:48,312][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:01:48,808][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:01:49,312][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:01:49,810][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:01:50,314][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:01:50,815][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:01:51,312][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:01:51,814][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:01:52,311][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:01:52,809][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:01:53,308][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:01:53,809][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:01:54,310][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:01:54,810][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:01:55,310][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:01:55,811][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:01:56,313][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:01:56,817][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:01:57,321][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:01:57,825][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:01:58,333][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:01:58,834][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:01:59,336][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:01:59,843][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:02:00,341][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:02:00,850][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:02:01,356][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:02:01,860][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:02:02,367][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:02:02,872][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:02:03,375][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:02:03,877][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:02:04,378][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:02:04,879][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:02:05,386][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:02:05,890][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:02:06,412][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:02:06,920][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:02:07,421][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:02:07,925][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:02:08,427][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:02:08,930][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:02:09,432][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:02:09,933][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:02:10,436][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:02:10,940][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10039 tokens. -[2025-11-10 04:02:11,824][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.03%, Current % of VRAM taken: 55.53%, Block Peak % of device VRAM: 62.08%, ΔTime: 00:00:33 -[2025-11-10 04:02:12,578][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:02:12,579][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:02:12,581][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:02:13,524][__main__][INFO] - Iteration 615 took 1m 1s (40.43% Gen, 58.03% Train). Generation: 24s, Training: 35s. Estimated remaining time: 41h 3m 43s. Estimated total time: 51h 6m 59s. Time estimates for 10 more iterations: 10m 13s, 100 more iterations: 1h 42m 13s, 500 more iterations: 8h 31m 9s. -[2025-11-10 04:02:13,526][__main__][INFO] - Starting iteration 615. -[2025-11-10 04:02:14,016][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 61 and human policies 1. -[2025-11-10 04:02:14,017][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:02:30,301][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 04:02:38,817][__main__][INFO] - Number of regex retries in iteration 615: 1 -[2025-11-10 04:02:38,818][__main__][INFO] - agents played in iteration 615 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:02:39,619][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:02:39,642][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:02:39,666][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:02:39,690][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:02:39,691][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:02:39,691][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:02:40,558][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:02:41,015][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:02:41,527][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:02:42,029][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:02:42,529][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:02:43,035][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:02:43,535][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:02:45,581][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:02:46,137][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:02:46,636][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:02:47,139][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:02:47,640][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:02:48,148][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:02:48,647][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:02:49,147][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:02:49,646][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:02:50,147][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:02:50,643][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:02:51,141][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:02:51,640][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:02:52,140][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:02:52,637][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:02:53,134][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:02:53,631][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:02:54,127][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:02:54,630][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:02:55,128][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:02:55,626][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:02:56,126][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:02:56,624][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:02:57,124][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:02:57,625][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:02:58,129][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:02:58,632][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:02:59,131][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:02:59,632][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:03:00,134][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:03:00,633][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:03:01,136][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:03:01,640][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:03:02,139][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:03:02,642][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:03:03,145][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:03:03,649][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:03:04,151][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:03:04,651][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:03:05,155][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:03:05,658][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:03:06,157][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:03:06,661][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:03:07,170][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:03:07,671][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:03:08,172][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:03:08,678][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:03:09,185][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:03:09,687][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:03:10,190][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:03:10,702][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:03:11,207][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:03:11,723][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:03:12,227][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:03:12,730][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:03:13,232][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:03:13,736][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:03:14,240][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10105 tokens. -[2025-11-10 04:03:15,170][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 55.58%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:34 -[2025-11-10 04:03:15,836][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:03:15,838][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:03:15,840][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:03:16,738][__main__][INFO] - Iteration 616 took 1m 2s (39.54% Gen, 59.03% Train). Generation: 24s, Training: 37s. Estimated remaining time: 42h 11m 51s. Estimated total time: 52h 16m 11s. Time estimates for 10 more iterations: 10m 27s, 100 more iterations: 1h 44m 32s, 500 more iterations: 8h 42m 41s. -[2025-11-10 04:03:16,741][__main__][INFO] - Starting iteration 616. -[2025-11-10 04:03:17,214][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 61 and human policies 1. -[2025-11-10 04:03:17,214][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:03:48,108][__main__][INFO] - Number of regex retries in iteration 616: 0 -[2025-11-10 04:03:48,108][__main__][INFO] - agents played in iteration 616 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:03:48,920][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:03:48,947][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:03:48,975][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:03:48,998][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:03:48,999][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:03:48,999][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:03:49,863][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:03:50,324][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:03:50,832][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:03:51,337][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:03:51,844][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:03:52,350][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:03:52,854][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:03:53,355][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:03:53,858][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:03:54,357][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:03:54,857][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:03:55,354][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:03:55,856][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:03:56,352][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:03:56,849][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:03:57,347][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:03:57,849][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:03:58,348][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:03:58,848][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:03:59,349][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:03:59,848][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:04:00,349][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:04:00,851][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:04:01,350][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:04:01,851][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:04:02,349][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:04:02,847][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:04:03,347][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:04:03,845][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:04:04,343][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:04:04,844][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:04:05,346][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:04:05,850][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:04:06,353][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:04:06,854][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:04:07,362][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:04:07,867][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:04:08,368][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:04:08,872][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:04:09,371][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:04:09,874][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:04:10,372][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:04:10,871][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:04:11,371][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:04:11,871][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:04:12,376][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:04:12,878][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:04:13,380][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:04:13,883][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:04:14,387][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:04:14,889][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:04:15,391][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:04:15,892][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:04:16,408][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:04:16,910][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:04:17,409][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:04:17,915][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:04:18,417][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:04:18,923][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:04:19,422][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:04:19,925][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:04:20,427][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:04:20,930][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:04:21,432][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:04:21,934][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10128 tokens. -[2025-11-10 04:04:22,769][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.09%, Current % of VRAM taken: 55.58%, Block Peak % of device VRAM: 62.20%, ΔTime: 00:00:32 -[2025-11-10 04:04:23,489][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:04:23,491][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:04:23,492][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:04:24,455][__main__][INFO] - Iteration 617 took 1m 7s (45.94% Gen, 52.62% Train). Generation: 30s, Training: 35s. Estimated remaining time: 45h 56m 39s. Estimated total time: 56h 2m 6s. Time estimates for 10 more iterations: 11m 12s, 100 more iterations: 1h 52m 4s, 500 more iterations: 9h 20m 21s. -[2025-11-10 04:04:24,457][__main__][INFO] - Starting iteration 617. -[2025-11-10 04:04:24,935][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 61 and human policies 1. -[2025-11-10 04:04:24,935][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:04:49,612][__main__][INFO] - Number of regex retries in iteration 617: 0 -[2025-11-10 04:04:49,612][__main__][INFO] - agents played in iteration 617 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:04:50,415][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:04:50,441][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:04:50,465][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:04:50,488][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:04:50,489][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:04:50,489][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:04:51,354][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:04:51,817][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:04:52,323][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:04:52,834][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:04:53,335][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:04:53,837][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:04:54,340][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:04:54,841][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:04:55,344][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:04:55,845][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:04:56,459][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:04:56,967][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:04:57,473][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:04:57,973][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:04:58,473][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:04:59,029][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:04:59,555][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:05:00,053][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:05:00,550][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:05:01,049][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:05:01,550][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:05:02,051][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:05:02,552][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:05:03,052][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:05:03,549][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:05:04,048][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:05:04,546][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:05:05,045][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:05:05,546][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:05:06,051][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:05:06,549][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:05:07,050][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:05:07,555][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:05:08,053][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:05:08,558][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:05:09,061][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:05:09,563][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:05:10,063][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:05:10,567][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:05:11,065][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:05:11,567][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:05:12,072][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:05:12,574][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:05:13,083][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:05:13,584][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:05:14,087][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:05:14,589][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:05:15,090][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:05:15,597][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:05:16,097][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:05:16,598][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:05:17,099][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:05:17,601][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:05:18,117][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:05:18,620][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:05:19,137][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:05:19,639][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:05:20,139][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:05:20,650][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:05:21,152][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:05:21,656][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:05:22,160][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:05:22,661][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:05:23,172][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:05:23,677][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10073 tokens. -[2025-11-10 04:05:24,616][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 55.70%, Block Peak % of device VRAM: 62.17%, ΔTime: 00:00:33 -[2025-11-10 04:05:25,251][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:05:25,253][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:05:25,255][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:05:26,210][__main__][INFO] - Iteration 618 took 1m 1s (40.27% Gen, 58.17% Train). Generation: 24s, Training: 35s. Estimated remaining time: 40h 57m 21s. Estimated total time: 51h 3m 50s. Time estimates for 10 more iterations: 10m 12s, 100 more iterations: 1h 42m 7s, 500 more iterations: 8h 30m 38s. -[2025-11-10 04:05:26,213][__main__][INFO] - Starting iteration 618. -[2025-11-10 04:05:26,689][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 61 and human policies 1. -[2025-11-10 04:05:26,689][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:05:46,977][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 04:05:58,117][__main__][INFO] - Number of regex retries in iteration 618: 1 -[2025-11-10 04:05:58,117][__main__][INFO] - agents played in iteration 618 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:05:58,921][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:05:58,948][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:05:58,974][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:05:58,997][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:05:58,997][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:05:58,998][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:05:59,877][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:06:00,335][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:06:00,846][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:06:01,348][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:06:01,849][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:06:02,355][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:06:02,860][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:06:03,363][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:06:03,869][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:06:04,374][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:06:04,876][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:06:05,375][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:06:05,878][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:06:06,387][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:06:06,890][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:06:07,390][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:06:07,895][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:06:08,400][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:06:08,905][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:06:09,407][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:06:09,908][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:06:10,408][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:06:10,912][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:06:11,412][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:06:11,916][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:06:12,417][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:06:12,923][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:06:13,426][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:06:13,930][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:06:14,434][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:06:14,935][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:06:15,437][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:06:15,938][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:06:16,440][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:06:16,951][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:06:17,456][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:06:17,957][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:06:18,460][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:06:18,966][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:06:19,472][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:06:19,978][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:06:20,482][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:06:20,991][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:06:21,493][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:06:21,994][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:06:22,495][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:06:22,997][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:06:23,498][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:06:23,999][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:06:24,503][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:06:25,005][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:06:25,509][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:06:26,010][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:06:26,514][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:06:27,021][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:06:27,526][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:06:28,026][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:06:28,523][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:06:29,020][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:06:29,515][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:06:30,020][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:06:30,515][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:06:31,013][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:06:31,517][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:06:32,014][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10232 tokens. -[2025-11-10 04:06:32,859][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.05%, Current % of VRAM taken: 55.55%, Block Peak % of device VRAM: 62.50%, ΔTime: 00:00:32 -[2025-11-10 04:06:33,590][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:06:33,592][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:06:33,593][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:06:34,520][__main__][INFO] - Iteration 619 took 1m 7s (46.33% Gen, 52.30% Train). Generation: 31s, Training: 35s. Estimated remaining time: 46h 23m 56s. Estimated total time: 56h 31m 34s. Time estimates for 10 more iterations: 11m 18s, 100 more iterations: 1h 53m 3s, 500 more iterations: 9h 25m 15s. -[2025-11-10 04:06:34,522][__main__][INFO] - Starting iteration 619. -[2025-11-10 04:06:35,005][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 61 and human policies 1. -[2025-11-10 04:06:35,006][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:06:58,615][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 04:07:00,882][__main__][INFO] - Number of regex retries in iteration 619: 1 -[2025-11-10 04:07:00,882][__main__][INFO] - agents played in iteration 619 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:07:01,695][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:07:01,718][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:07:01,741][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:07:01,764][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:07:01,764][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:07:01,765][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:07:02,647][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:07:03,116][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:07:03,620][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:07:04,121][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:07:04,622][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:07:05,122][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:07:05,626][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:07:06,126][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:07:06,629][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:07:07,145][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:07:07,650][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:07:08,158][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:07:08,668][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:07:09,175][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:07:09,690][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:07:10,196][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:07:10,704][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:07:11,205][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:07:11,705][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:07:12,236][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:07:12,739][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:07:13,242][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:07:13,765][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:07:14,269][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:07:14,791][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:07:15,296][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:07:15,798][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:07:16,301][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:07:17,497][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:07:19,058][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:07:19,563][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:07:20,083][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:07:20,588][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:07:21,099][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:07:21,605][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:07:22,114][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:07:22,621][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:07:23,131][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:07:23,634][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:07:24,148][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:07:24,653][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:07:25,173][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:07:25,684][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:07:26,197][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:07:26,705][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:07:27,216][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:07:27,725][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:07:28,231][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:07:28,735][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:07:29,239][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:07:29,744][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:07:30,248][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:07:30,751][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:07:31,250][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:07:31,760][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:07:32,265][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:07:32,764][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:07:33,285][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:07:33,783][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:07:34,283][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:07:34,785][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:07:35,284][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:07:35,786][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:07:36,286][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:07:36,785][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10156 tokens. -[2025-11-10 04:07:37,638][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 55.65%, Block Peak % of device VRAM: 62.43%, ΔTime: 00:00:34 -[2025-11-10 04:07:38,280][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:07:38,282][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:07:38,283][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:07:39,139][__main__][INFO] - Iteration 620 took 1m 4s (40.35% Gen, 58.32% Train). Generation: 25s, Training: 37s. Estimated remaining time: 43h 18m 1s. Estimated total time: 53h 26m 43s. Time estimates for 10 more iterations: 10m 41s, 100 more iterations: 1h 46m 53s, 500 more iterations: 8h 54m 27s. -[2025-11-10 04:07:39,142][__main__][INFO] - Starting iteration 620. -[2025-11-10 04:07:39,630][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 61 and human policies 1. -[2025-11-10 04:07:39,630][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:08:06,804][__main__][INFO] - Number of regex retries in iteration 620: 0 -[2025-11-10 04:08:06,805][__main__][INFO] - agents played in iteration 620 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:08:07,591][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:08:07,620][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:08:07,647][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:08:07,670][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:08:07,671][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:08:07,672][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:08:08,539][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:08:09,000][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:08:09,504][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:08:10,005][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:08:10,504][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:08:11,005][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:08:11,515][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:08:12,014][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:08:12,513][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:08:13,014][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:08:13,514][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:08:14,017][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:08:14,519][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:08:15,019][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:08:15,520][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:08:16,022][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:08:16,525][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:08:17,033][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:08:17,533][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:08:18,038][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:08:18,541][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:08:19,042][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:08:19,550][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:08:20,053][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:08:20,555][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:08:21,054][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:08:21,555][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:08:22,072][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:08:22,571][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:08:23,074][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:08:23,579][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:08:24,080][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:08:24,588][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:08:25,089][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:08:25,590][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:08:26,091][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:08:26,599][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:08:27,102][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:08:27,609][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:08:28,112][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:08:28,620][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:08:29,121][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:08:29,657][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:08:30,159][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:08:30,661][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:08:31,185][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:08:31,684][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:08:32,179][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:08:32,690][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:08:33,187][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:08:33,688][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:08:34,200][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:08:34,698][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:08:35,198][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:08:35,694][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:08:36,191][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:08:36,695][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:08:37,192][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:08:37,689][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:08:38,192][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:08:38,689][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:08:39,192][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:08:39,690][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:08:40,188][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:08:40,704][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10105 tokens. -[2025-11-10 04:08:41,563][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 55.63%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:33 -[2025-11-10 04:08:42,309][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:08:42,311][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:08:42,312][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:08:44,192][__main__][INFO] - Iteration 621 took 1m 4s (42.09% Gen, 55.00% Train). Generation: 27s, Training: 35s. Estimated remaining time: 43h 38m 21s. Estimated total time: 53h 48m 8s. Time estimates for 10 more iterations: 10m 45s, 100 more iterations: 1h 47m 36s, 500 more iterations: 8h 58m 1s. -[2025-11-10 04:08:44,194][__main__][INFO] - Starting iteration 621. -[2025-11-10 04:08:44,661][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 62 and human policies 1. -[2025-11-10 04:08:44,662][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:09:08,148][__main__][INFO] - Number of regex retries in iteration 621: 0 -[2025-11-10 04:09:08,148][__main__][INFO] - agents played in iteration 621 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:09:08,937][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:09:08,964][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:09:08,990][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:09:09,014][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:09:09,014][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:09:09,016][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:09:09,872][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:09:10,330][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:09:10,843][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:09:11,344][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:09:11,848][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:09:12,352][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:09:12,854][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:09:13,356][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:09:13,860][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:09:14,367][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:09:14,889][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:09:15,413][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:09:15,914][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:09:16,421][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:09:16,923][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:09:17,429][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:09:17,931][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:09:18,431][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:09:18,929][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:09:19,429][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:09:19,935][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:09:20,433][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:09:20,932][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:09:21,439][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:09:21,942][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:09:22,446][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:09:22,946][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:09:23,446][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:09:23,951][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:09:24,451][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:09:24,951][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:09:25,454][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:09:25,955][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:09:26,459][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:09:26,963][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:09:27,466][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:09:27,969][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:09:28,470][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:09:28,968][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:09:29,473][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:09:29,975][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:09:30,488][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:09:30,989][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:09:31,485][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:09:31,980][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:09:32,476][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:09:32,971][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:09:33,471][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:09:33,968][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:09:34,469][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:09:34,969][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:09:35,469][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:09:35,980][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:09:36,476][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:09:36,973][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:09:37,474][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:09:37,971][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:09:38,470][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:09:38,968][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:09:39,466][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:09:39,964][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:09:40,462][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:09:40,961][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:09:41,468][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:09:41,968][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10031 tokens. -[2025-11-10 04:09:42,795][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 55.61%, Block Peak % of device VRAM: 62.09%, ΔTime: 00:00:32 -[2025-11-10 04:09:43,565][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:09:43,566][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:09:43,568][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:09:44,447][__main__][INFO] - Iteration 622 took 59s (39.28% Gen, 59.24% Train). Generation: 23s, Training: 35s. Estimated remaining time: 39h 38m 33s. Estimated total time: 49h 49m 20s. Time estimates for 10 more iterations: 9m 57s, 100 more iterations: 1h 39m 38s, 500 more iterations: 8h 18m 13s. -[2025-11-10 04:09:44,450][__main__][INFO] - Starting iteration 622. -[2025-11-10 04:09:44,917][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 62 and human policies 1. -[2025-11-10 04:09:44,917][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:10:07,842][__main__][INFO] - Number of regex retries in iteration 622: 0 -[2025-11-10 04:10:07,843][__main__][INFO] - agents played in iteration 622 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:10:08,639][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:10:08,671][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:10:08,696][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:10:08,720][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:10:08,720][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:10:08,722][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:10:09,615][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:10:10,073][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:10:10,578][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:10:11,092][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:10:11,594][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:10:12,097][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:10:12,600][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:10:13,103][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:10:13,608][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:10:14,109][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:10:14,611][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:10:15,113][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:10:15,621][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:10:16,132][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:10:16,634][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:10:17,136][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:10:17,638][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:10:18,148][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:10:18,659][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:10:19,164][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:10:19,695][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:10:20,207][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:10:20,710][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:10:21,221][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:10:21,729][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:10:22,232][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:10:22,738][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:10:23,245][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:10:23,751][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:10:24,253][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:10:24,756][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:10:25,260][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:10:25,767][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:10:26,270][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:10:26,777][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:10:27,285][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:10:27,790][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:10:28,292][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:10:28,793][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:10:29,297][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:10:29,798][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:10:30,300][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:10:30,799][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:10:31,302][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:10:31,827][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:10:32,327][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:10:32,828][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:10:33,331][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:10:33,832][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:10:34,334][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:10:34,839][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:10:35,340][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:10:35,845][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:10:36,347][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:10:36,845][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:10:37,347][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:10:37,847][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:10:38,346][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:10:38,847][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:10:39,345][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:10:39,850][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:10:40,349][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:10:40,844][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:10:41,340][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:10:41,838][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10045 tokens. -[2025-11-10 04:10:42,677][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 55.58%, Block Peak % of device VRAM: 62.21%, ΔTime: 00:00:33 -[2025-11-10 04:10:43,353][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:10:43,355][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:10:43,357][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:10:44,283][__main__][INFO] - Iteration 623 took 59s (38.62% Gen, 59.82% Train). Generation: 22s, Training: 35s. Estimated remaining time: 39h 16m 31s. Estimated total time: 49h 28m 19s. Time estimates for 10 more iterations: 9m 53s, 100 more iterations: 1h 38m 56s, 500 more iterations: 8h 14m 43s. -[2025-11-10 04:10:44,285][__main__][INFO] - Starting iteration 623. -[2025-11-10 04:10:44,765][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 62 and human policies 1. -[2025-11-10 04:10:44,766][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:11:14,692][__main__][INFO] - Number of regex retries in iteration 623: 0 -[2025-11-10 04:11:14,693][__main__][INFO] - agents played in iteration 623 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:11:15,478][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:11:15,506][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:11:15,533][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:11:15,557][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:11:15,558][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:11:15,560][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:11:16,355][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:11:16,818][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:11:17,326][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:11:17,829][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:11:18,330][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:11:18,832][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:11:19,333][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:11:19,833][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:11:20,351][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:11:20,856][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:11:21,362][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:11:21,863][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:11:22,366][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:11:22,867][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:11:23,370][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:11:23,871][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:11:24,376][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:11:24,877][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:11:25,384][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:11:25,885][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:11:26,388][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:11:26,897][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:11:27,397][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:11:27,896][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:11:28,397][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:11:28,898][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:11:29,404][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:11:29,908][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:11:30,409][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:11:30,923][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:11:31,425][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:11:31,941][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:11:32,444][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:11:32,944][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:11:33,459][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:11:33,960][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:11:34,463][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:11:34,963][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:11:35,464][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:11:35,969][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:11:36,472][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:11:36,976][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:11:37,477][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:11:37,984][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:11:38,484][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:11:38,987][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:11:39,492][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:11:39,994][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:11:40,496][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:11:40,994][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:11:41,490][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:11:41,990][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:11:42,492][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:11:42,989][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:11:43,488][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:11:43,987][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:11:44,484][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:11:44,983][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:11:45,488][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:11:45,985][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:11:46,485][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:11:46,983][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:11:47,487][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:11:47,985][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:11:48,489][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10083 tokens. -[2025-11-10 04:11:49,314][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 55.64%, Block Peak % of device VRAM: 62.20%, ΔTime: 00:00:32 -[2025-11-10 04:11:50,068][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:11:50,069][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:11:50,071][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:11:51,048][__main__][INFO] - Iteration 624 took 1m 6s (45.15% Gen, 53.37% Train). Generation: 29s, Training: 35s. Estimated remaining time: 45h 1m 14s. Estimated total time: 55h 14m 9s. Time estimates for 10 more iterations: 11m 2s, 100 more iterations: 1h 50m 28s, 500 more iterations: 9h 12m 21s. -[2025-11-10 04:11:51,050][__main__][INFO] - Starting iteration 624. -[2025-11-10 04:11:51,544][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 62 and human policies 1. -[2025-11-10 04:11:51,544][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:12:20,018][__main__][INFO] - Number of regex retries in iteration 624: 0 -[2025-11-10 04:12:20,021][__main__][INFO] - agents played in iteration 624 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:12:20,836][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:12:20,873][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:12:20,902][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:12:20,928][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:12:20,928][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:12:20,929][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:12:21,792][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:12:22,252][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:12:22,771][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:12:23,273][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:12:23,775][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:12:24,279][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:12:24,784][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:12:25,288][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:12:25,791][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:12:26,295][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:12:26,798][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:12:27,301][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:12:27,810][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:12:28,313][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:12:28,815][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:12:29,317][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:12:29,819][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:12:30,323][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:12:30,841][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:12:31,342][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:12:31,855][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:12:32,356][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:12:32,861][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:12:33,370][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:12:33,874][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:12:34,402][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:12:34,915][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:12:35,421][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:12:35,925][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:12:36,428][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:12:36,931][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:12:37,447][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:12:37,950][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:12:38,454][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:12:38,960][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:12:39,465][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:12:39,969][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:12:40,472][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:12:40,974][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:12:41,478][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:12:41,982][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:12:42,492][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:12:42,994][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:12:43,503][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:12:44,003][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:12:44,505][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:12:45,008][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:12:45,513][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:12:46,009][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:12:46,506][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:12:47,004][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:12:47,498][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:12:47,995][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:12:48,490][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:12:48,991][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:12:49,487][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:12:49,988][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:12:50,492][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:12:50,989][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:12:51,488][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:12:51,987][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:12:52,490][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:12:52,992][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:12:53,487][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:12:53,987][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10109 tokens. -[2025-11-10 04:12:54,862][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 55.66%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:33 -[2025-11-10 04:12:55,493][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:12:55,495][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:12:55,497][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:12:56,292][__main__][INFO] - Iteration 625 took 1m 4s (43.98% Gen, 54.79% Train). Generation: 28s, Training: 35s. Estimated remaining time: 43h 43m 26s. Estimated total time: 53h 57m 25s. Time estimates for 10 more iterations: 10m 47s, 100 more iterations: 1h 47m 54s, 500 more iterations: 8h 59m 34s. -[2025-11-10 04:12:56,294][__main__][INFO] - Starting iteration 625. -[2025-11-10 04:12:56,775][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 62 and human policies 1. -[2025-11-10 04:12:56,776][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:13:13,725][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 04:13:24,635][__main__][INFO] - Number of regex retries in iteration 625: 1 -[2025-11-10 04:13:24,635][__main__][INFO] - agents played in iteration 625 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:13:25,436][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:13:25,461][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:13:25,484][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:13:25,508][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:13:25,509][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:13:25,509][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:13:26,396][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:13:26,854][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:13:27,362][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:13:27,865][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:13:28,372][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:13:28,873][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:13:29,376][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:13:29,878][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:13:30,384][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:13:30,889][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:13:31,387][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:13:31,894][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:13:32,395][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:13:32,895][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:13:33,392][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:13:33,893][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:13:34,397][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:13:34,902][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:13:35,402][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:13:35,907][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:13:36,411][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:13:36,912][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:13:37,414][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:13:37,914][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:13:38,421][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:13:38,926][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:13:39,427][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:13:39,941][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:13:40,443][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:13:40,952][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:13:41,456][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:13:41,958][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:13:42,461][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:13:42,965][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:13:43,467][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:13:43,973][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:13:44,473][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:13:44,982][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:13:45,487][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:13:45,989][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:13:46,497][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:13:46,997][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:13:47,496][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:13:47,997][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:13:48,496][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:13:48,993][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:13:49,488][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:13:49,987][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:13:50,492][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:13:50,988][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:13:51,484][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:13:51,986][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:13:52,487][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:13:52,994][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:13:53,492][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:13:53,991][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:13:54,486][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:13:54,983][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:13:55,479][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:13:55,976][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:13:56,473][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:13:56,974][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:13:57,471][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:13:57,971][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:13:58,471][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9989 tokens. -[2025-11-10 04:13:59,346][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.02%, Current % of VRAM taken: 55.51%, Block Peak % of device VRAM: 62.12%, ΔTime: 00:00:32 -[2025-11-10 04:14:00,088][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:14:00,089][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:14:00,091][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:14:01,124][__main__][INFO] - Iteration 626 took 1m 4s (43.29% Gen, 55.10% Train). Generation: 27s, Training: 35s. Estimated remaining time: 43h 22m 22s. Estimated total time: 53h 37m 26s. Time estimates for 10 more iterations: 10m 43s, 100 more iterations: 1h 47m 14s, 500 more iterations: 8h 56m 14s. -[2025-11-10 04:14:01,126][__main__][INFO] - Starting iteration 626. -[2025-11-10 04:14:01,615][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 62 and human policies 1. -[2025-11-10 04:14:01,616][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:14:27,337][__main__][INFO] - Number of regex retries in iteration 626: 0 -[2025-11-10 04:14:27,339][__main__][INFO] - agents played in iteration 626 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:14:28,146][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:14:28,178][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:14:28,204][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:14:28,227][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:14:28,228][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:14:28,228][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:14:29,090][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:14:29,561][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:14:30,068][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:14:30,568][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:14:31,073][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:14:31,574][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:14:32,091][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:14:32,594][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:14:33,097][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:14:33,600][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:14:34,106][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:14:34,614][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:14:35,117][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:14:35,623][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:14:36,128][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:14:36,631][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:14:37,134][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:14:37,635][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:14:38,134][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:14:38,638][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:14:39,142][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:14:39,648][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:14:40,152][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:14:40,657][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:14:41,160][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:14:41,660][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:14:42,163][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:14:42,674][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:14:43,177][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:14:43,680][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:14:44,186][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:14:44,686][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:14:45,201][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:14:45,704][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:14:46,210][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:14:46,711][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:14:47,211][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:14:47,712][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:14:48,215][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:14:48,716][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:14:49,225][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:14:49,726][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:14:50,231][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:14:50,732][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:14:51,236][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:14:51,740][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:14:52,244][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:14:52,744][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:14:53,251][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:14:53,751][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:14:54,254][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:14:54,754][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:14:55,255][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:14:55,758][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:14:56,258][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:14:56,757][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:14:57,259][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:14:57,760][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:14:58,271][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:14:58,768][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:14:59,263][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:14:59,775][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:15:00,270][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:15:00,767][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:15:01,266][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10113 tokens. -[2025-11-10 04:15:02,095][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 55.71%, Block Peak % of device VRAM: 62.20%, ΔTime: 00:00:33 -[2025-11-10 04:15:02,730][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:15:02,731][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:15:02,733][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:15:03,621][__main__][INFO] - Iteration 627 took 1m 2s (41.49% Gen, 57.08% Train). Generation: 25s, Training: 35s. Estimated remaining time: 41h 24m 13s. Estimated total time: 51h 40m 19s. Time estimates for 10 more iterations: 10m 20s, 100 more iterations: 1h 43m 20s, 500 more iterations: 8h 36m 43s. -[2025-11-10 04:15:03,623][__main__][INFO] - Starting iteration 627. -[2025-11-10 04:15:04,084][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 62 and human policies 1. -[2025-11-10 04:15:04,084][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:15:27,692][__main__][INFO] - Number of regex retries in iteration 627: 0 -[2025-11-10 04:15:27,693][__main__][INFO] - agents played in iteration 627 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:15:28,481][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:15:28,510][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:15:28,536][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:15:28,560][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:15:28,560][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:15:28,561][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:15:29,414][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:15:30,029][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:15:30,535][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:15:31,038][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:15:31,537][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:15:32,040][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:15:32,539][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:15:33,047][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:15:33,548][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:15:34,050][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:15:34,550][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:15:35,055][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:15:35,557][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:15:36,056][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:15:36,562][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:15:37,063][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:15:37,563][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:15:38,064][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:15:38,566][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:15:39,068][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:15:39,568][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:15:40,069][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:15:40,570][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:15:41,070][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:15:41,579][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:15:42,084][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:15:42,583][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:15:43,081][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:15:43,581][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:15:44,087][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:15:44,587][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:15:45,086][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:15:45,604][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:15:46,105][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:15:46,607][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:15:47,108][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:15:47,608][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:15:48,121][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:15:48,622][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:15:49,122][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:15:49,622][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:15:50,123][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:15:50,635][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:15:51,139][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:15:51,644][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:15:52,148][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:15:52,649][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:15:53,155][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:15:53,657][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:15:54,159][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:15:54,666][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:15:55,169][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:15:55,673][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:15:56,187][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:15:56,690][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:15:57,231][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:15:57,735][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:15:58,236][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:15:58,743][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:15:59,242][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:15:59,742][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:16:00,239][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:16:00,734][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:16:01,234][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:16:01,730][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10024 tokens. -[2025-11-10 04:16:02,562][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 55.61%, Block Peak % of device VRAM: 62.17%, ΔTime: 00:00:33 -[2025-11-10 04:16:03,281][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:16:03,283][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:16:03,284][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:16:04,219][__main__][INFO] - Iteration 628 took 1m 0s (39.26% Gen, 59.19% Train). Generation: 23s, Training: 35s. Estimated remaining time: 39h 49m 40s. Estimated total time: 50h 6m 47s. Time estimates for 10 more iterations: 10m 1s, 100 more iterations: 1h 40m 13s, 500 more iterations: 8h 21m 7s. -[2025-11-10 04:16:04,221][__main__][INFO] - Starting iteration 628. -[2025-11-10 04:16:04,686][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 62 and human policies 1. -[2025-11-10 04:16:04,688][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:16:30,794][__main__][INFO] - Number of regex retries in iteration 628: 0 -[2025-11-10 04:16:30,797][__main__][INFO] - agents played in iteration 628 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:16:31,662][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:16:31,719][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:16:31,762][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:16:31,794][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:16:31,794][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:16:31,796][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:16:32,662][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:16:33,120][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:16:33,626][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:16:34,131][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:16:34,636][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:16:35,143][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:16:35,646][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:16:36,151][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:16:36,653][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:16:37,154][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:16:37,657][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:16:38,160][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:16:38,659][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:16:39,163][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:16:39,665][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:16:40,166][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:16:40,666][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:16:41,169][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:16:41,672][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:16:42,172][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:16:42,674][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:16:43,181][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:16:43,680][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:16:44,180][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:16:44,682][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:16:45,185][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:16:45,688][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:16:46,205][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:16:46,708][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:16:47,221][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:16:47,723][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:16:48,227][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:16:48,732][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:16:49,235][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:16:49,738][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:16:50,247][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:16:50,753][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:16:51,258][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:16:51,761][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:16:52,268][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:16:52,772][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:16:53,274][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:16:53,777][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:16:54,276][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:16:54,777][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:16:55,278][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:16:55,778][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:16:56,278][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:16:56,779][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:16:57,281][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:16:57,780][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:16:58,284][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:16:58,786][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:16:59,299][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:16:59,798][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:17:00,308][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:17:00,811][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:17:01,312][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:17:01,832][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:17:02,333][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:17:02,837][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:17:03,343][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:17:03,842][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:17:04,350][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:17:04,852][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10078 tokens. -[2025-11-10 04:17:05,804][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 55.59%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:33 -[2025-11-10 04:17:06,455][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:17:06,457][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:17:06,459][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:17:07,274][__main__][INFO] - Iteration 629 took 1m 2s (41.71% Gen, 56.98% Train). Generation: 26s, Training: 35s. Estimated remaining time: 41h 51m 15s. Estimated total time: 52h 9m 26s. Time estimates for 10 more iterations: 10m 25s, 100 more iterations: 1h 44m 18s, 500 more iterations: 8h 41m 34s. -[2025-11-10 04:17:07,276][__main__][INFO] - Starting iteration 629. -[2025-11-10 04:17:07,741][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 62 and human policies 1. -[2025-11-10 04:17:07,741][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:17:31,322][__main__][INFO] - Number of regex retries in iteration 629: 0 -[2025-11-10 04:17:31,323][__main__][INFO] - agents played in iteration 629 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:17:32,176][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:17:32,203][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:17:32,229][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:17:32,252][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:17:32,253][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:17:32,253][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:17:33,034][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:17:33,492][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:17:33,996][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:17:34,492][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:17:34,988][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:17:35,485][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:17:35,983][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:17:36,482][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:17:36,978][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:17:37,475][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:17:37,979][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:17:38,478][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:17:38,980][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:17:39,479][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:17:39,978][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:17:40,476][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:17:40,976][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:17:41,482][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:17:41,985][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:17:42,486][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:17:43,000][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:17:43,504][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:17:44,013][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:17:44,515][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:17:45,019][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:17:45,523][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:17:46,028][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:17:46,533][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:17:47,038][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:17:47,546][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:17:48,059][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:17:48,561][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:17:49,065][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:17:49,570][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:17:50,072][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:17:50,575][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:17:51,082][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:17:51,586][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:17:52,091][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:17:52,594][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:17:53,129][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:17:53,633][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:17:54,137][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:17:54,647][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:17:55,151][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:17:55,659][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:17:56,157][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:17:56,660][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:17:57,163][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:17:57,671][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:17:58,177][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:17:58,684][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:17:59,187][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:17:59,692][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:18:00,206][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:18:00,709][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:18:01,214][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:18:01,716][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:18:02,216][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:18:02,733][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:18:03,243][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:18:03,744][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:18:04,249][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:18:04,755][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:18:05,264][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10029 tokens. -[2025-11-10 04:18:06,203][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 55.61%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:33 -[2025-11-10 04:18:06,920][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:18:06,921][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:18:06,923][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:18:07,837][__main__][INFO] - Iteration 630 took 1m 0s (39.24% Gen, 59.24% Train). Generation: 23s, Training: 35s. Estimated remaining time: 39h 45m 39s. Estimated total time: 50h 4m 49s. Time estimates for 10 more iterations: 10m 0s, 100 more iterations: 1h 40m 9s, 500 more iterations: 8h 20m 48s. -[2025-11-10 04:18:07,839][__main__][INFO] - Starting iteration 630. -[2025-11-10 04:18:08,301][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 62 and human policies 1. -[2025-11-10 04:18:08,302][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:18:20,345][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 04:18:30,936][__main__][INFO] - Number of regex retries in iteration 630: 1 -[2025-11-10 04:18:30,937][__main__][INFO] - agents played in iteration 630 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:18:31,732][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:18:31,755][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:18:31,777][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:18:31,799][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:18:31,800][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:18:31,801][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:18:32,583][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:18:33,047][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:18:33,547][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:18:34,042][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:18:34,541][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:18:35,035][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:18:35,540][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:18:36,041][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:18:36,533][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:18:37,034][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:18:37,529][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:18:38,034][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:18:38,534][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:18:39,034][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:18:39,546][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:18:40,047][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:18:40,550][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:18:41,049][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:18:41,548][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:18:42,057][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:18:42,555][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:18:43,055][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:18:43,556][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:18:44,056][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:18:44,564][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:18:45,064][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:18:45,563][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:18:46,063][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:18:46,564][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:18:47,065][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:18:47,565][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:18:48,065][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:18:48,565][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:18:49,499][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:18:50,211][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:18:50,808][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:18:51,325][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:18:51,835][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:18:52,348][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:18:52,853][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:18:53,359][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:18:53,860][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:18:54,359][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:18:54,861][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:18:55,359][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:18:55,861][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:18:56,367][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:18:56,871][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:18:57,375][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:18:57,876][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:18:58,380][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:18:58,883][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:18:59,387][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:18:59,886][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:19:00,389][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:19:00,889][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:19:01,392][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:19:01,892][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:19:02,392][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:19:02,894][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:19:03,397][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:19:03,896][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:19:04,401][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:19:04,901][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:19:05,401][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10034 tokens. -[2025-11-10 04:19:06,333][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.04%, Current % of VRAM taken: 55.53%, Block Peak % of device VRAM: 62.18%, ΔTime: 00:00:33 -[2025-11-10 04:19:07,002][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:19:07,004][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:19:07,006][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:19:08,724][__main__][INFO] - Iteration 631 took 1m 0s (37.46% Gen, 59.70% Train). Generation: 22s, Training: 36s. Estimated remaining time: 40h 0m 58s. Estimated total time: 50h 21m 10s. Time estimates for 10 more iterations: 10m 4s, 100 more iterations: 1h 40m 42s, 500 more iterations: 8h 23m 31s. -[2025-11-10 04:19:08,726][__main__][INFO] - Starting iteration 631. -[2025-11-10 04:19:09,197][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 63 and human policies 1. -[2025-11-10 04:19:09,198][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:19:41,140][__main__][INFO] - Number of regex retries in iteration 631: 0 -[2025-11-10 04:19:41,141][__main__][INFO] - agents played in iteration 631 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:19:41,905][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:19:41,931][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:19:41,958][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:19:41,980][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:19:41,981][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:19:41,982][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:19:42,788][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:19:43,245][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:19:43,751][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:19:44,249][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:19:44,749][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:19:45,252][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:19:45,751][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:19:46,252][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:19:46,750][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:19:47,255][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:19:47,753][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:19:48,259][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:19:48,764][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:19:49,265][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:19:49,770][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:19:50,275][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:19:50,773][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:19:51,275][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:19:51,778][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:19:52,279][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:19:52,781][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:19:53,286][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:19:53,793][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:19:54,300][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:19:54,802][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:19:55,304][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:19:55,806][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:19:56,309][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:19:56,815][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:19:57,316][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:19:57,818][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:19:58,329][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:19:58,838][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:19:59,353][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:19:59,857][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:20:00,361][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:20:00,873][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:20:01,376][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:20:01,885][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:20:02,394][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:20:02,899][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:20:03,403][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:20:03,913][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:20:04,419][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:20:04,934][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:20:05,437][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:20:05,949][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:20:06,453][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:20:06,956][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:20:07,461][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:20:07,968][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:20:08,494][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:20:09,002][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:20:09,510][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:20:10,015][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:20:10,520][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:20:11,022][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:20:11,537][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:20:12,043][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:20:12,560][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:20:13,067][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:20:13,574][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:20:14,081][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:20:14,586][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:20:15,093][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10055 tokens. -[2025-11-10 04:20:15,966][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 55.64%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:33 -[2025-11-10 04:20:16,736][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:20:16,738][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:20:16,740][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:20:17,795][__main__][INFO] - Iteration 632 took 1m 8s (46.56% Gen, 51.89% Train). Generation: 31s, Training: 35s. Estimated remaining time: 46h 48m 35s. Estimated total time: 57h 9m 56s. Time estimates for 10 more iterations: 11m 25s, 100 more iterations: 1h 54m 19s, 500 more iterations: 9h 31m 39s. -[2025-11-10 04:20:17,798][__main__][INFO] - Starting iteration 632. -[2025-11-10 04:20:18,262][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 63 and human policies 1. -[2025-11-10 04:20:18,263][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:20:30,345][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 04:20:41,471][__main__][INFO] - Number of regex retries in iteration 632: 1 -[2025-11-10 04:20:41,471][__main__][INFO] - agents played in iteration 632 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:20:42,234][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:20:42,258][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:20:42,280][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:20:42,303][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:20:42,304][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:20:42,304][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:20:43,069][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:20:43,520][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:20:44,028][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:20:44,526][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:20:45,024][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:20:45,522][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:20:46,018][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:20:46,531][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:20:47,031][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:20:47,529][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:20:48,041][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:20:48,538][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:20:49,045][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:20:49,544][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:20:50,046][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:20:50,563][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:20:51,064][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:20:51,566][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:20:52,066][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:20:52,568][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:20:53,074][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:20:53,570][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:20:54,074][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:20:54,577][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:20:55,076][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:20:55,767][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:20:56,270][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:20:56,778][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:20:57,299][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:20:57,808][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:20:58,315][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:20:58,817][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:20:59,320][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:20:59,833][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:21:00,335][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:21:00,841][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:21:01,344][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:21:01,849][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:21:02,353][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:21:02,857][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:21:03,359][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:21:03,862][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:21:04,363][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:21:04,870][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:21:05,376][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:21:05,878][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:21:06,384][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:21:06,889][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:21:07,394][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:21:07,901][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:21:08,406][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:21:08,906][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:21:09,410][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:21:09,914][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:21:10,413][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:21:10,914][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:21:11,415][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:21:11,917][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:21:12,421][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:21:12,925][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:21:13,431][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:21:13,933][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:21:14,436][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:21:14,937][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:21:15,439][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10081 tokens. -[2025-11-10 04:21:16,363][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 55.60%, Block Peak % of device VRAM: 62.20%, ΔTime: 00:00:33 -[2025-11-10 04:21:17,066][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:21:17,068][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:21:17,070][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:21:17,908][__main__][INFO] - Iteration 633 took 59s (38.91% Gen, 59.68% Train). Generation: 23s, Training: 35s. Estimated remaining time: 39h 20m 1s. Estimated total time: 49h 42m 22s. Time estimates for 10 more iterations: 9m 56s, 100 more iterations: 1h 39m 24s, 500 more iterations: 8h 17m 3s. -[2025-11-10 04:21:17,911][__main__][INFO] - Starting iteration 633. -[2025-11-10 04:21:18,377][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 63 and human policies 1. -[2025-11-10 04:21:18,378][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:21:51,125][__main__][INFO] - Number of regex retries in iteration 633: 0 -[2025-11-10 04:21:51,126][__main__][INFO] - agents played in iteration 633 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:21:51,893][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:21:51,922][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:21:51,948][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:21:51,971][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:21:51,972][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:21:51,973][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:21:52,752][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:21:53,211][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:21:53,714][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:21:54,222][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:21:54,727][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:21:55,235][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:21:55,739][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:21:56,242][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:21:56,745][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:21:57,248][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:21:57,762][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:21:58,263][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:21:58,764][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:21:59,270][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:21:59,771][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:22:00,280][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:22:00,780][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:22:01,285][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:22:01,802][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:22:02,301][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:22:02,804][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:22:03,305][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:22:03,805][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:22:04,311][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:22:04,811][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:22:05,312][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:22:05,812][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:22:06,313][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:22:06,820][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:22:07,321][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:22:07,822][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:22:08,323][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:22:08,826][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:22:09,327][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:22:09,831][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:22:10,331][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:22:10,830][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:22:11,327][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:22:11,825][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:22:12,327][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:22:12,831][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:22:13,333][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:22:13,835][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:22:14,336][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:22:14,843][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:22:15,347][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:22:15,850][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:22:16,351][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:22:16,853][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:22:17,356][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:22:17,856][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:22:18,355][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:22:18,855][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:22:19,355][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:22:19,855][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:22:20,355][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:22:20,856][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:22:21,371][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:22:21,876][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:22:22,377][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:22:22,886][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:22:23,390][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:22:23,892][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:22:24,394][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:22:24,894][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10092 tokens. -[2025-11-10 04:22:25,781][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 55.63%, Block Peak % of device VRAM: 62.24%, ΔTime: 00:00:33 -[2025-11-10 04:22:26,538][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:22:26,540][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:22:26,542][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:22:27,496][__main__][INFO] - Iteration 634 took 1m 9s (47.38% Gen, 51.24% Train). Generation: 32s, Training: 35s. Estimated remaining time: 47h 12m 29s. Estimated total time: 57h 36m 0s. Time estimates for 10 more iterations: 11m 31s, 100 more iterations: 1h 55m 12s, 500 more iterations: 9h 36m 0s. -[2025-11-10 04:22:27,499][__main__][INFO] - Starting iteration 634. -[2025-11-10 04:22:27,967][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 63 and human policies 1. -[2025-11-10 04:22:27,968][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:22:54,255][__main__][INFO] - Number of regex retries in iteration 634: 0 -[2025-11-10 04:22:54,255][__main__][INFO] - agents played in iteration 634 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:22:55,025][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:22:55,051][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:22:55,075][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:22:55,098][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:22:55,098][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:22:55,099][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:22:55,892][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:22:56,350][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:22:56,857][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:22:57,356][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:22:57,857][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:22:58,368][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:22:58,874][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:22:59,376][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:22:59,884][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:23:00,384][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:23:00,897][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:23:01,396][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:23:01,898][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:23:02,401][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:23:03,931][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:23:04,450][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:23:04,960][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:23:05,461][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:23:05,963][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:23:06,464][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:23:06,967][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:23:07,486][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:23:07,991][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:23:08,495][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:23:08,999][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:23:09,500][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:23:10,008][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:23:10,507][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:23:11,009][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:23:11,521][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:23:12,024][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:23:12,529][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:23:13,031][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:23:13,537][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:23:14,039][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:23:14,545][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:23:15,048][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:23:15,554][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:23:16,053][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:23:16,553][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:23:17,053][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:23:17,552][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:23:18,054][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:23:18,559][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:23:19,060][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:23:19,573][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:23:20,075][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:23:20,580][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:23:21,078][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:23:21,577][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:23:22,094][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:23:22,596][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:23:23,096][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:23:23,600][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:23:24,100][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:23:24,609][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:23:25,111][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:23:25,614][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:23:26,121][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:23:26,625][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:23:27,131][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:23:27,635][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:23:28,139][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:23:28,642][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:23:29,149][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10089 tokens. -[2025-11-10 04:23:30,053][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 55.66%, Block Peak % of device VRAM: 62.22%, ΔTime: 00:00:34 -[2025-11-10 04:23:30,695][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:23:30,697][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:23:30,699][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:23:31,548][__main__][INFO] - Iteration 635 took 1m 3s (41.34% Gen, 57.32% Train). Generation: 26s, Training: 36s. Estimated remaining time: 42h 34m 30s. Estimated total time: 52h 59m 5s. Time estimates for 10 more iterations: 10m 35s, 100 more iterations: 1h 45m 58s, 500 more iterations: 8h 49m 50s. -[2025-11-10 04:23:31,551][__main__][INFO] - Starting iteration 635. -[2025-11-10 04:23:32,019][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 63 and human policies 1. -[2025-11-10 04:23:32,020][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:24:04,629][__main__][INFO] - Number of regex retries in iteration 635: 0 -[2025-11-10 04:24:04,630][__main__][INFO] - agents played in iteration 635 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:24:05,437][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:24:05,466][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:24:05,491][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:24:05,514][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:24:05,515][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:24:05,515][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:24:06,395][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:24:06,851][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:24:07,361][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:24:07,864][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:24:08,368][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:24:08,871][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:24:09,373][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:24:09,877][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:24:10,380][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:24:10,880][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:24:11,381][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:24:11,886][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:24:12,403][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:24:12,906][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:24:13,411][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:24:13,913][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:24:14,415][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:24:14,920][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:24:15,418][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:24:15,918][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:24:16,424][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:24:16,921][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:24:17,419][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:24:17,915][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:24:18,410][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:24:18,909][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:24:19,405][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:24:19,902][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:24:20,401][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:24:20,897][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:24:21,396][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:24:21,894][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:24:22,392][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:24:22,909][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:24:23,407][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:24:23,905][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:24:24,406][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:24:24,909][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:24:25,415][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:24:25,913][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:24:26,414][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:24:26,919][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:24:27,427][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:24:27,928][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:24:28,431][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:24:28,934][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:24:29,439][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:24:29,942][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:24:30,443][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:24:30,951][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:24:31,455][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:24:31,955][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:24:32,456][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:24:32,957][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:24:33,459][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:24:33,958][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:24:34,460][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:24:34,966][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:24:35,471][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:24:35,973][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:24:36,475][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:24:36,976][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:24:37,483][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:24:37,987][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:24:38,489][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9995 tokens. -[2025-11-10 04:24:39,347][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.09%, Current % of VRAM taken: 55.59%, Block Peak % of device VRAM: 62.05%, ΔTime: 00:00:32 -[2025-11-10 04:24:40,105][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:24:40,107][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:24:40,108][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:24:41,121][__main__][INFO] - Iteration 636 took 1m 9s (47.19% Gen, 51.34% Train). Generation: 32s, Training: 35s. Estimated remaining time: 47h 9m 24s. Estimated total time: 57h 35m 8s. Time estimates for 10 more iterations: 11m 31s, 100 more iterations: 1h 55m 10s, 500 more iterations: 9h 35m 51s. -[2025-11-10 04:24:41,123][__main__][INFO] - Starting iteration 636. -[2025-11-10 04:24:41,592][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 63 and human policies 1. -[2025-11-10 04:24:41,593][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:24:53,467][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 04:25:03,620][__main__][INFO] - Number of regex retries in iteration 636: 1 -[2025-11-10 04:25:03,620][__main__][INFO] - agents played in iteration 636 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:25:04,404][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:25:04,430][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:25:04,453][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:25:04,476][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:25:04,476][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:25:04,477][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:25:05,363][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:25:05,822][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:25:06,336][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:25:06,838][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:25:07,339][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:25:07,841][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:25:08,344][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:25:08,845][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:25:09,348][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:25:09,855][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:25:10,356][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:25:10,856][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:25:11,359][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:25:11,864][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:25:12,414][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:25:14,498][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:25:15,002][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:25:15,527][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:25:16,031][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:25:16,538][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:25:17,044][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:25:17,549][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:25:18,052][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:25:18,551][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:25:19,046][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:25:19,546][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:25:20,044][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:25:20,547][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:25:21,047][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:25:21,545][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:25:22,043][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:25:22,541][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:25:23,043][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:25:23,546][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:25:24,045][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:25:24,547][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:25:25,045][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:25:25,543][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:25:26,043][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:25:26,545][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:25:27,046][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:25:27,553][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:25:28,053][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:25:28,556][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:25:29,056][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:25:29,557][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:25:30,062][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:25:30,567][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:25:31,071][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:25:31,574][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:25:32,085][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:25:32,588][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:25:33,095][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:25:33,598][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:25:34,107][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:25:34,612][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:25:35,124][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:25:35,622][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:25:36,128][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:25:36,630][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:25:37,131][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:25:37,631][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:25:38,134][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:25:38,636][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:25:39,137][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10123 tokens. -[2025-11-10 04:25:40,055][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.04%, Current % of VRAM taken: 55.53%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:34 -[2025-11-10 04:25:40,707][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:25:40,708][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:25:40,710][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:25:41,519][__main__][INFO] - Iteration 637 took 59s (36.76% Gen, 61.89% Train). Generation: 22s, Training: 37s. Estimated remaining time: 39h 29m 37s. Estimated total time: 49h 56m 22s. Time estimates for 10 more iterations: 9m 59s, 100 more iterations: 1h 39m 52s, 500 more iterations: 8h 19m 23s. -[2025-11-10 04:25:41,521][__main__][INFO] - Starting iteration 637. -[2025-11-10 04:25:41,992][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 63 and human policies 1. -[2025-11-10 04:25:41,993][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:26:14,510][__main__][INFO] - Number of regex retries in iteration 637: 0 -[2025-11-10 04:26:14,510][__main__][INFO] - agents played in iteration 637 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:26:15,318][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:26:15,342][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:26:15,367][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:26:15,390][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:26:15,390][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:26:15,391][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:26:16,291][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:26:16,749][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:26:17,260][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:26:17,763][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:26:18,264][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:26:18,767][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:26:19,274][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:26:19,773][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:26:20,277][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:26:20,782][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:26:21,286][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:26:21,792][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:26:22,292][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:26:22,801][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:26:23,304][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:26:23,804][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:26:24,316][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:26:24,811][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:26:25,313][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:26:25,810][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:26:26,310][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:26:26,825][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:26:27,325][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:26:27,823][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:26:28,324][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:26:28,825][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:26:29,327][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:26:29,827][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:26:30,328][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:26:30,829][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:26:31,328][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:26:31,839][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:26:32,337][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:26:32,841][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:26:33,337][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:26:33,840][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:26:34,349][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:26:34,849][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:26:35,351][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:26:35,854][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:26:36,351][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:26:36,852][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:26:37,358][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:26:37,861][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:26:38,362][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:26:38,864][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:26:39,369][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:26:39,870][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:26:40,374][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:26:40,877][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:26:41,374][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:26:41,875][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:26:42,382][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:26:42,886][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:26:43,391][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:26:43,893][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:26:44,394][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:26:44,893][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:26:45,400][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:26:45,901][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:26:46,408][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:26:46,910][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:26:47,413][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:26:47,921][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:26:48,425][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10144 tokens. -[2025-11-10 04:26:49,303][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 55.68%, Block Peak % of device VRAM: 62.38%, ΔTime: 00:00:33 -[2025-11-10 04:26:50,044][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:26:50,045][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:26:50,047][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:26:50,968][__main__][INFO] - Iteration 638 took 1m 8s (47.14% Gen, 51.52% Train). Generation: 32s, Training: 35s. Estimated remaining time: 47h 0m 57s. Estimated total time: 57h 28m 51s. Time estimates for 10 more iterations: 11m 29s, 100 more iterations: 1h 54m 57s, 500 more iterations: 9h 34m 48s. -[2025-11-10 04:26:50,970][__main__][INFO] - Starting iteration 638. -[2025-11-10 04:26:51,437][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 63 and human policies 1. -[2025-11-10 04:26:51,438][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:27:09,710][__main__][INFO] - Number of regex retries in iteration 638: 0 -[2025-11-10 04:27:09,710][__main__][INFO] - agents played in iteration 638 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:27:10,605][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:27:10,628][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:27:10,651][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:27:10,674][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:27:10,675][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:27:10,675][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:27:11,531][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:27:11,988][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:27:12,495][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:27:12,997][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:27:13,497][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:27:13,998][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:27:14,498][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:27:15,002][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:27:15,510][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:27:16,010][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:27:16,511][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:27:17,012][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:27:17,512][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:27:18,013][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:27:18,512][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:27:19,012][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:27:19,514][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:27:20,016][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:27:20,518][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:27:21,019][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:27:21,515][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:27:22,014][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:27:22,510][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:27:23,009][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:27:23,512][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:27:24,011][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:27:24,507][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:27:25,005][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:27:25,505][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:27:26,003][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:27:26,499][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:27:26,997][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:27:27,500][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:27:27,996][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:27:28,492][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:27:28,989][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:27:29,486][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:27:29,988][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:27:30,486][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:27:30,982][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:27:31,480][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:27:31,985][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:27:32,486][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:27:32,984][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:27:33,479][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:27:33,975][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:27:34,469][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:27:34,965][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:27:35,459][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:27:35,961][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:27:36,474][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:27:36,973][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:27:37,472][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:27:37,973][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:27:38,469][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:27:38,973][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:27:39,473][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:27:39,974][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:27:40,476][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:27:40,976][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:27:41,477][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:27:41,975][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:27:42,476][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:27:42,983][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:27:43,486][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10045 tokens. -[2025-11-10 04:27:44,390][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 55.67%, Block Peak % of device VRAM: 62.14%, ΔTime: 00:00:32 -[2025-11-10 04:27:45,107][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:27:45,109][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:27:45,111][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:27:46,019][__main__][INFO] - Iteration 639 took 54s (33.48% Gen, 64.86% Train). Generation: 18s, Training: 35s. Estimated remaining time: 35h 0m 19s. Estimated total time: 45h 29m 8s. Time estimates for 10 more iterations: 9m 5s, 100 more iterations: 1h 30m 58s, 500 more iterations: 7h 34m 51s. -[2025-11-10 04:27:46,021][__main__][INFO] - Starting iteration 639. -[2025-11-10 04:27:46,484][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 63 and human policies 1. -[2025-11-10 04:27:46,485][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:28:16,918][__main__][INFO] - Number of regex retries in iteration 639: 0 -[2025-11-10 04:28:16,921][__main__][INFO] - agents played in iteration 639 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:28:17,715][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:28:17,739][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:28:17,764][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:28:17,787][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:28:17,788][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:28:17,789][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:28:18,648][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:28:19,105][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:28:19,614][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:28:20,119][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:28:20,618][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:28:21,116][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:28:21,616][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:28:22,116][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:28:22,618][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:28:23,126][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:28:23,626][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:28:24,130][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:28:24,633][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:28:25,129][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:28:25,626][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:28:26,123][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:28:26,625][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:28:27,129][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:28:27,630][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:28:28,130][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:28:28,630][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:28:29,132][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:28:29,635][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:28:30,135][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:28:30,648][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:28:31,146][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:28:31,649][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:28:32,148][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:28:32,648][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:28:33,153][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:28:33,653][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:28:34,156][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:28:34,658][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:28:35,157][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:28:35,661][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:28:36,164][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:28:36,669][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:28:37,180][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:28:37,683][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:28:38,184][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:28:38,686][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:28:39,186][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:28:39,687][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:28:40,190][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:28:40,690][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:28:41,190][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:28:41,689][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:28:42,195][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:28:42,693][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:28:43,197][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:28:43,702][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:28:44,205][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:28:44,707][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:28:45,208][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:28:45,708][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:28:46,212][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:28:46,713][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:28:47,214][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:28:47,715][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:28:48,216][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:28:48,716][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:28:49,217][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:28:49,721][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:28:50,224][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:28:50,724][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10078 tokens. -[2025-11-10 04:28:51,654][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 55.56%, Block Peak % of device VRAM: 62.47%, ΔTime: 00:00:33 -[2025-11-10 04:28:52,293][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:28:52,295][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:28:52,297][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:28:53,129][__main__][INFO] - Iteration 640 took 1m 6s (45.67% Gen, 53.08% Train). Generation: 30s, Training: 35s. Estimated remaining time: 45h 2m 19s. Estimated total time: 55h 32m 15s. Time estimates for 10 more iterations: 11m 6s, 100 more iterations: 1h 51m 4s, 500 more iterations: 9h 15m 22s. -[2025-11-10 04:28:53,131][__main__][INFO] - Starting iteration 640. -[2025-11-10 04:28:53,597][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 63 and human policies 1. -[2025-11-10 04:28:53,598][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:29:19,251][__main__][INFO] - Number of regex retries in iteration 640: 0 -[2025-11-10 04:29:19,252][__main__][INFO] - agents played in iteration 640 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:29:20,052][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:29:20,077][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:29:20,114][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:29:20,138][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:29:20,139][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:29:20,140][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:29:21,015][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:29:21,477][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:29:21,984][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:29:22,484][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:29:22,994][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:29:23,493][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:29:23,995][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:29:24,508][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:29:25,011][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:29:25,516][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:29:26,017][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:29:26,522][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:29:27,025][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:29:27,528][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:29:28,026][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:29:28,529][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:29:29,027][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:29:29,528][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:29:30,027][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:29:30,524][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:29:31,021][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:29:31,521][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:29:32,022][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:29:32,518][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:29:33,014][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:29:33,516][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:29:34,017][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:29:34,518][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:29:35,025][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:29:35,526][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:29:36,028][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:29:36,529][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:29:37,032][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:29:37,535][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:29:38,036][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:29:38,538][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:29:39,041][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:29:39,542][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:29:40,046][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:29:40,557][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:29:41,058][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:29:41,582][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:29:42,083][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:29:42,585][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:29:43,088][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:29:43,591][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:29:44,094][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:29:44,597][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:29:45,099][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:29:45,601][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:29:46,105][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:29:46,608][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:29:47,108][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:29:47,608][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:29:48,114][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:29:48,617][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:29:49,117][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:29:49,620][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:29:50,124][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:29:50,623][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:29:51,123][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:29:51,625][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:29:52,130][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:29:52,635][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:29:53,140][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10103 tokens. -[2025-11-10 04:29:54,043][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 55.72%, Block Peak % of device VRAM: 62.24%, ΔTime: 00:00:33 -[2025-11-10 04:29:54,803][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:29:54,805][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:29:54,807][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:29:56,658][__main__][INFO] - Iteration 641 took 1m 3s (40.68% Gen, 56.38% Train). Generation: 25s, Training: 35s. Estimated remaining time: 42h 2m 4s. Estimated total time: 52h 33m 4s. Time estimates for 10 more iterations: 10m 30s, 100 more iterations: 1h 45m 6s, 500 more iterations: 8h 45m 30s. -[2025-11-10 04:29:56,660][__main__][INFO] - Starting iteration 641. -[2025-11-10 04:29:57,153][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 64 and human policies 1. -[2025-11-10 04:29:57,153][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:30:11,891][mllm.models.large_language_model_local][WARNING] - Response Proposal: x hats, y books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 04:30:22,601][__main__][INFO] - Number of regex retries in iteration 641: 1 -[2025-11-10 04:30:22,602][__main__][INFO] - agents played in iteration 641 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:30:23,407][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:30:23,430][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:30:23,454][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:30:23,478][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:30:23,478][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:30:23,479][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:30:24,355][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:30:24,817][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:30:25,330][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:30:25,837][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:30:26,347][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:30:26,849][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:30:27,349][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:30:27,854][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:30:28,353][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:30:28,858][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:30:29,369][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:30:29,872][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:30:30,374][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:30:30,877][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:30:31,378][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:30:31,881][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:30:32,384][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:30:32,892][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:30:33,395][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:30:33,897][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:30:34,399][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:30:34,928][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:30:35,432][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:30:35,936][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:30:36,437][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:30:36,934][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:30:37,437][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:30:37,939][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:30:38,438][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:30:38,936][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:30:39,437][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:30:39,936][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:30:40,434][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:30:40,935][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:30:41,435][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:30:41,943][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:30:42,444][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:30:42,946][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:30:43,459][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:30:43,958][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:30:44,460][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:30:44,960][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:30:45,461][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:30:45,988][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:30:46,487][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:30:46,986][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:30:47,488][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:30:47,990][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:30:48,492][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:30:48,994][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:30:49,495][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:30:49,996][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:30:50,499][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:30:50,999][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:30:51,499][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:30:51,999][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:30:52,503][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:30:53,004][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:30:53,508][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:30:54,012][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:30:54,513][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:30:55,017][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:30:55,516][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:30:56,019][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:30:56,529][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10097 tokens. -[2025-11-10 04:30:57,448][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:33 -[2025-11-10 04:30:58,104][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:30:58,105][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:30:58,107][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:30:59,086][__main__][INFO] - Iteration 642 took 1m 1s (41.09% Gen, 57.33% Train). Generation: 25s, Training: 35s. Estimated remaining time: 41h 4m 40s. Estimated total time: 51h 36m 42s. Time estimates for 10 more iterations: 10m 19s, 100 more iterations: 1h 43m 13s, 500 more iterations: 8h 36m 7s. -[2025-11-10 04:30:59,088][__main__][INFO] - Starting iteration 642. -[2025-11-10 04:30:59,556][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 64 and human policies 1. -[2025-11-10 04:30:59,556][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:31:30,112][__main__][INFO] - Number of regex retries in iteration 642: 0 -[2025-11-10 04:31:30,112][__main__][INFO] - agents played in iteration 642 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:31:30,900][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:31:30,923][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:31:30,947][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:31:30,969][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:31:30,970][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:31:30,970][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:31:31,838][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:31:32,293][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:31:32,802][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:31:33,315][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:31:33,814][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:31:34,315][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:31:34,821][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:31:35,322][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:31:35,825][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:31:36,331][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:31:36,832][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:31:37,333][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:31:37,835][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:31:38,332][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:31:38,828][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:31:39,328][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:31:39,834][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:31:40,331][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:31:40,856][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:31:41,366][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:31:41,861][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:31:42,364][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:31:42,863][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:31:43,363][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:31:43,860][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:31:44,358][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:31:44,857][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:31:45,359][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:31:45,857][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:31:46,354][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:31:46,854][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:31:47,357][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:31:47,861][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:31:48,360][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:31:48,861][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:31:49,372][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:31:49,873][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:31:50,376][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:31:50,879][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:31:51,379][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:31:51,887][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:31:52,391][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:31:52,901][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:31:53,408][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:31:53,911][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:31:54,409][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:31:54,910][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:31:55,416][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:31:55,917][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:31:56,417][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:31:56,916][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:31:57,417][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:31:57,918][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:31:58,419][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:31:58,924][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:31:59,422][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:31:59,923][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:32:00,428][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:32:00,933][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:32:01,435][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:32:01,935][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:32:02,438][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:32:02,941][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:32:03,443][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:32:03,942][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10000 tokens. -[2025-11-10 04:32:04,747][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.03%, Current % of VRAM taken: 55.53%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:32 -[2025-11-10 04:32:05,466][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:32:05,468][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:32:05,470][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:32:06,360][__main__][INFO] - Iteration 643 took 1m 6s (45.74% Gen, 52.93% Train). Generation: 30s, Training: 35s. Estimated remaining time: 45h 7m 5s. Estimated total time: 55h 40m 14s. Time estimates for 10 more iterations: 11m 8s, 100 more iterations: 1h 51m 20s, 500 more iterations: 9h 16m 42s. -[2025-11-10 04:32:06,362][__main__][INFO] - Starting iteration 643. -[2025-11-10 04:32:06,823][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 64 and human policies 1. -[2025-11-10 04:32:06,824][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:32:31,371][__main__][INFO] - Number of regex retries in iteration 643: 0 -[2025-11-10 04:32:31,372][__main__][INFO] - agents played in iteration 643 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:32:32,247][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:32:32,276][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:32:32,303][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:32:32,340][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:32:32,341][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:32:32,342][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:32:33,240][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:32:33,697][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:32:34,205][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:32:34,705][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:32:35,207][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:32:35,708][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:32:36,209][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:32:36,708][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:32:37,210][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:32:37,717][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:32:38,221][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:32:38,723][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:32:39,229][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:32:39,732][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:32:40,235][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:32:40,738][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:32:41,239][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:32:41,746][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:32:42,249][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:32:42,751][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:32:43,252][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:32:43,754][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:32:44,282][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:32:44,785][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:32:45,288][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:32:45,789][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:32:46,294][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:32:46,797][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:32:47,304][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:32:47,804][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:32:48,303][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:32:48,803][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:32:49,308][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:32:49,807][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:32:50,304][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:32:50,801][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:32:51,297][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:32:51,794][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:32:52,312][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:32:52,811][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:32:53,313][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:32:53,819][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:32:54,321][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:32:54,820][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:32:55,320][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:32:55,819][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:32:56,318][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:32:56,817][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:32:57,321][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:32:57,823][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:32:58,324][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:32:58,830][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:32:59,338][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:32:59,840][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:33:00,342][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:33:00,846][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:33:01,348][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:33:01,858][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:33:02,361][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:33:02,864][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:33:03,370][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:33:03,873][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:33:04,378][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:33:04,881][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:33:05,383][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9944 tokens. -[2025-11-10 04:33:06,281][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 55.66%, Block Peak % of device VRAM: 62.20%, ΔTime: 00:00:33 -[2025-11-10 04:33:06,983][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:33:06,984][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:33:06,987][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:33:07,862][__main__][INFO] - Iteration 644 took 1m 1s (40.22% Gen, 58.35% Train). Generation: 24s, Training: 35s. Estimated remaining time: 40h 17m 48s. Estimated total time: 50h 51m 59s. Time estimates for 10 more iterations: 10m 10s, 100 more iterations: 1h 41m 43s, 500 more iterations: 8h 28m 39s. -[2025-11-10 04:33:07,865][__main__][INFO] - Starting iteration 644. -[2025-11-10 04:33:08,332][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 64 and human policies 1. -[2025-11-10 04:33:08,333][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:33:35,866][__main__][INFO] - Number of regex retries in iteration 644: 0 -[2025-11-10 04:33:35,866][__main__][INFO] - agents played in iteration 644 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:33:36,661][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:33:36,684][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:33:36,707][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:33:36,730][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:33:36,731][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:33:36,732][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:33:37,617][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:33:38,077][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:33:38,589][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:33:39,089][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:33:39,590][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:33:40,091][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:33:40,591][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:33:41,106][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:33:41,607][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:33:42,112][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:33:42,617][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:33:43,120][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:33:43,624][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:33:44,128][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:33:44,638][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:33:45,156][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:33:45,658][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:33:46,159][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:33:46,662][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:33:47,168][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:33:47,672][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:33:48,176][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:33:48,680][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:33:49,182][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:33:49,683][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:33:50,183][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:33:50,687][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:33:51,185][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:33:51,690][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:33:52,187][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:33:52,684][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:33:53,195][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:33:53,696][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:33:54,204][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:33:54,703][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:33:55,203][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:33:55,720][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:33:56,221][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:33:56,726][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:33:57,227][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:33:57,730][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:33:58,239][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:33:58,742][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:33:59,247][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:33:59,752][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:34:00,261][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:34:00,763][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:34:01,263][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:34:01,766][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:34:02,271][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:34:02,775][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:34:03,277][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:34:03,779][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:34:04,285][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:34:04,788][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:34:05,290][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:34:05,793][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:34:06,306][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:34:06,805][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:34:07,317][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:34:07,818][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:34:08,320][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:34:08,826][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:34:09,331][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:34:09,834][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10219 tokens. -[2025-11-10 04:34:10,653][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 55.62%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:33 -[2025-11-10 04:34:11,400][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:34:11,402][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:34:11,404][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:34:12,475][__main__][INFO] - Iteration 645 took 1m 4s (42.92% Gen, 55.40% Train). Generation: 27s, Training: 35s. Estimated remaining time: 42h 51m 55s. Estimated total time: 53h 27m 11s. Time estimates for 10 more iterations: 10m 41s, 100 more iterations: 1h 46m 54s, 500 more iterations: 8h 54m 31s. -[2025-11-10 04:34:12,477][__main__][INFO] - Starting iteration 645. -[2025-11-10 04:34:12,963][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 64 and human policies 1. -[2025-11-10 04:34:12,963][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:34:37,415][__main__][INFO] - Number of regex retries in iteration 645: 0 -[2025-11-10 04:34:37,417][__main__][INFO] - agents played in iteration 645 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:34:38,277][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:34:38,306][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:34:38,332][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:34:38,355][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:34:38,356][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:34:38,357][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:34:39,225][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:34:39,696][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:34:40,210][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:34:40,714][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:34:41,214][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:34:41,721][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:34:42,222][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:34:42,725][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:34:43,229][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:34:43,732][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:34:44,243][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:34:44,745][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:34:45,250][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:34:45,751][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:34:46,256][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:34:46,760][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:34:47,261][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:34:47,765][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:34:48,272][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:34:48,774][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:34:49,277][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:34:49,782][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:34:50,287][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:34:50,791][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:34:51,294][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:34:51,798][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:34:52,311][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:34:52,813][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:34:53,316][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:34:53,819][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:34:54,319][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:34:54,831][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:34:55,331][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:34:55,835][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:34:56,341][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:34:56,847][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:34:57,350][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:34:57,852][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:34:58,354][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:34:58,860][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:34:59,357][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:34:59,861][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:35:00,360][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:35:00,861][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:35:01,363][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:35:01,864][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:35:02,365][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:35:02,870][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:35:03,372][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:35:03,871][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:35:04,368][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:35:04,870][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:35:05,370][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:35:05,870][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:35:06,372][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:35:06,870][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:35:07,368][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:35:07,873][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:35:08,377][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:35:08,880][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:35:09,384][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:35:09,887][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:35:10,391][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:35:10,892][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:35:11,396][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10049 tokens. -[2025-11-10 04:35:12,307][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.25%, Current % of VRAM taken: 55.74%, Block Peak % of device VRAM: 62.23%, ΔTime: 00:00:33 -[2025-11-10 04:35:12,975][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:35:12,976][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:35:12,978][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:35:13,843][__main__][INFO] - Iteration 646 took 1m 0s (40.17% Gen, 58.41% Train). Generation: 24s, Training: 35s. Estimated remaining time: 40h 7m 44s. Estimated total time: 50h 44m 1s. Time estimates for 10 more iterations: 10m 8s, 100 more iterations: 1h 41m 28s, 500 more iterations: 8h 27m 20s. -[2025-11-10 04:35:13,845][__main__][INFO] - Starting iteration 646. -[2025-11-10 04:35:14,312][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 64 and human policies 1. -[2025-11-10 04:35:14,313][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:35:38,890][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 04:35:39,683][__main__][INFO] - Number of regex retries in iteration 646: 1 -[2025-11-10 04:35:39,683][__main__][INFO] - agents played in iteration 646 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:35:40,518][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:35:40,545][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:35:40,571][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:35:40,593][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:35:40,594][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:35:40,595][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:35:41,379][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:35:41,833][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:35:42,335][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:35:42,843][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:35:43,345][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:35:43,841][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:35:44,346][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:35:44,846][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:35:45,350][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:35:45,849][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:35:46,348][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:35:46,847][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:35:47,349][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:35:47,849][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:35:48,356][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:35:48,857][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:35:49,359][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:35:49,860][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:35:50,361][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:35:50,868][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:35:51,369][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:35:51,872][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:35:52,373][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:35:52,872][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:35:53,376][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:35:53,878][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:35:54,380][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:35:54,886][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:35:55,387][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:35:55,888][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:35:56,390][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:35:56,892][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:35:57,391][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:35:57,893][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:35:58,393][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:35:58,893][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:35:59,394][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:35:59,898][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:36:00,395][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:36:00,893][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:36:01,390][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:36:01,889][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:36:02,391][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:36:02,892][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:36:03,399][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:36:03,900][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:36:04,405][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:36:04,907][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:36:05,406][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:36:05,909][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:36:06,413][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:36:06,915][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:36:07,418][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:36:07,932][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:36:08,437][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:36:08,941][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:36:09,447][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:36:09,951][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:36:10,453][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:36:10,954][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:36:11,459][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:36:11,965][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:36:12,466][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:36:12,967][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:36:13,472][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10099 tokens. -[2025-11-10 04:36:14,378][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:33 -[2025-11-10 04:36:15,125][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:36:15,126][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:36:15,128][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:36:16,143][__main__][INFO] - Iteration 647 took 1m 1s (41.03% Gen, 57.33% Train). Generation: 25s, Training: 35s. Estimated remaining time: 40h 54m 15s. Estimated total time: 51h 31m 35s. Time estimates for 10 more iterations: 10m 18s, 100 more iterations: 1h 43m 3s, 500 more iterations: 8h 35m 15s. -[2025-11-10 04:36:16,146][__main__][INFO] - Starting iteration 647. -[2025-11-10 04:36:16,619][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 64 and human policies 1. -[2025-11-10 04:36:16,619][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:36:39,703][__main__][INFO] - Number of regex retries in iteration 647: 0 -[2025-11-10 04:36:39,703][__main__][INFO] - agents played in iteration 647 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:36:40,485][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:36:40,515][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:36:40,544][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:36:40,574][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:36:40,575][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:36:40,577][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:36:41,356][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:36:41,816][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:36:42,331][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:36:42,834][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:36:43,334][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:36:43,832][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:36:44,334][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:36:44,850][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:36:45,353][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:36:45,855][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:36:46,357][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:36:46,862][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:36:47,378][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:36:47,878][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:36:48,379][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:36:48,885][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:36:49,390][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:36:49,897][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:36:51,601][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:36:52,299][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:36:52,802][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:36:53,308][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:36:53,811][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:36:54,312][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:36:54,818][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:36:55,319][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:36:55,819][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:36:56,319][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:36:56,818][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:36:57,316][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:36:57,818][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:36:58,316][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:36:58,814][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:36:59,316][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:36:59,821][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:37:00,320][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:37:00,831][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:37:01,334][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:37:01,839][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:37:02,345][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:37:02,852][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:37:03,370][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:37:03,875][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:37:04,380][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:37:04,877][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:37:05,380][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:37:05,883][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:37:06,383][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:37:06,884][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:37:07,387][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:37:07,889][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:37:08,395][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:37:08,896][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:37:09,398][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:37:09,898][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:37:10,399][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:37:10,899][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:37:11,402][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:37:11,900][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:37:12,401][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:37:12,900][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:37:13,400][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:37:13,903][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:37:14,402][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:37:14,903][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10194 tokens. -[2025-11-10 04:37:15,804][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 55.63%, Block Peak % of device VRAM: 62.51%, ΔTime: 00:00:34 -[2025-11-10 04:37:16,472][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:37:16,474][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:37:16,476][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:37:17,339][__main__][INFO] - Iteration 648 took 1m 0s (38.01% Gen, 60.56% Train). Generation: 23s, Training: 36s. Estimated remaining time: 39h 57m 42s. Estimated total time: 50h 36m 3s. Time estimates for 10 more iterations: 10m 7s, 100 more iterations: 1h 41m 12s, 500 more iterations: 8h 26m 0s. -[2025-11-10 04:37:17,341][__main__][INFO] - Starting iteration 648. -[2025-11-10 04:37:17,813][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 64 and human policies 1. -[2025-11-10 04:37:17,814][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:37:38,500][mllm.models.large_language_model_local][WARNING] - Response Proposal: x hats, 10 books, 10 balls +Given Bob's higher value for hats and balls, and your higher value for balls, it's strategic to maximize your share of the balls while contesting Bob for the hats. The books, having equal values, can be divided fairly. Thus, I propose to take all 10 balls to maximize my potential points from this item. I will also propose a fair split for the books and a minimal claim for hats to stay under the total item quantity and avoid proportional distribution. did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 07:51:29,639][__main__][INFO] - Number of regex retries in iteration 688: 1 +[2026-03-26 07:51:29,640][__main__][INFO] - agents played in iteration 688 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:51:30,417][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:51:30,441][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:51:30,466][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:51:30,490][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:51:30,491][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:51:30,491][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:51:31,323][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:51:31,782][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:51:32,292][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:51:32,797][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:51:33,297][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:51:33,797][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:51:34,297][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:51:34,796][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:51:35,307][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:51:35,811][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:51:36,315][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:51:36,815][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:51:37,315][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:51:37,814][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:51:38,314][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:51:38,817][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:51:39,319][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:51:39,830][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:51:40,329][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:51:40,831][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:51:41,332][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:51:41,832][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:51:42,331][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:51:42,832][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:51:43,332][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:51:43,830][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:51:44,332][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:51:44,831][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:51:45,327][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:51:45,827][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:51:46,325][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:51:46,824][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:51:47,323][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:51:47,825][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:51:48,327][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:51:48,828][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:51:49,329][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:51:49,827][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:51:50,325][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:51:50,823][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:51:51,322][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:51:51,820][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:51:52,316][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:51:52,815][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:51:53,312][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:51:53,809][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:51:54,310][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:51:54,806][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:51:55,303][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:51:55,800][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:51:56,297][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:51:56,794][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:51:57,290][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:51:57,786][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:51:58,282][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:51:58,780][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:51:59,278][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:51:59,775][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:52:00,274][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:52:00,772][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:52:01,274][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:52:01,770][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:52:02,272][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:52:02,767][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:52:03,263][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10068 tokens. +[2026-03-26 07:52:04,140][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.04%, Current % of VRAM taken: 60.47%, Block Peak % of device VRAM: 62.12%, ΔTime: 00:00:32 +[2026-03-26 07:52:04,882][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:52:04,884][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:52:04,886][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:52:05,609][__main__][INFO] - Iteration 689 took 51s (30.39% Gen, 68.21% Train). Generation: 15s, Training: 35s. Estimated remaining time: 32h 17m 21s. Estimated total time: 43h 3m 46s. Time estimates for 10 more iterations: 8m 36s, 100 more iterations: 1h 26m 7s, 500 more iterations: 7h 10m 37s. +[2026-03-26 07:52:05,611][__main__][INFO] - Starting iteration 689. +[2026-03-26 07:52:06,026][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 68 and human policies 1. +[2026-03-26 07:52:06,027][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:52:20,001][__main__][INFO] - Number of regex retries in iteration 689: 0 +[2026-03-26 07:52:20,002][__main__][INFO] - agents played in iteration 689 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:52:20,785][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:52:20,811][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:52:20,837][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:52:20,866][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:52:20,867][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:52:20,867][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:52:21,715][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:52:22,172][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:52:22,679][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:52:23,188][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:52:23,692][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:52:24,194][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:52:24,695][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:52:25,196][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:52:25,696][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:52:26,204][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:52:26,714][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:52:27,214][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:52:27,715][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:52:28,215][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:52:28,713][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:52:29,212][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:52:29,711][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:52:30,216][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:52:30,717][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:52:31,221][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:52:31,719][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:52:32,246][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:52:32,752][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:52:33,263][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:52:33,771][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:52:34,271][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:52:34,774][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:52:35,275][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:52:35,777][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:52:36,279][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:52:36,781][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:52:37,282][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:52:37,788][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:52:38,295][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:52:38,799][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:52:39,300][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:52:39,803][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:52:40,303][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:52:40,803][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:52:41,304][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:52:41,804][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:52:42,307][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:52:42,805][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:52:43,307][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:52:43,809][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:52:44,317][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:52:44,821][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:52:45,321][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:52:45,823][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:52:46,322][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:52:46,821][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:52:47,321][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:52:47,820][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:52:48,319][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:52:48,818][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:52:49,316][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:52:49,820][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:52:50,321][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:52:50,824][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:52:51,327][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:52:51,830][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:52:52,328][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:52:52,826][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:52:53,325][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:52:53,823][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10041 tokens. +[2026-03-26 07:52:54,697][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.03%, Current % of VRAM taken: 60.46%, Block Peak % of device VRAM: 62.12%, ΔTime: 00:00:32 +[2026-03-26 07:52:55,448][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:52:55,450][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:52:55,452][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:52:56,182][__main__][INFO] - Iteration 690 took 50s (27.86% Gen, 70.68% Train). Generation: 13s, Training: 35s. Estimated remaining time: 31h 0m 40s. Estimated total time: 41h 47m 56s. Time estimates for 10 more iterations: 8m 21s, 100 more iterations: 1h 23m 35s, 500 more iterations: 6h 57m 59s. +[2026-03-26 07:52:56,184][__main__][INFO] - Starting iteration 690. +[2026-03-26 07:52:56,599][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 68 and human policies 1. +[2026-03-26 07:52:56,600][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:53:11,072][__main__][INFO] - Number of regex retries in iteration 690: 0 +[2026-03-26 07:53:11,072][__main__][INFO] - agents played in iteration 690 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:53:11,834][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:53:11,859][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:53:11,883][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:53:11,908][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:53:11,909][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:53:11,909][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:53:12,751][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:53:13,215][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:53:13,721][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:53:14,224][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:53:14,725][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:53:15,227][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:53:15,738][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:53:16,239][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:53:16,739][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:53:17,243][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:53:17,741][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:53:18,242][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:53:18,740][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:53:19,238][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:53:19,737][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:53:20,238][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:53:20,736][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:53:21,237][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:53:21,744][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:53:22,281][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:53:22,784][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:53:23,288][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:53:23,794][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:53:24,294][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:53:24,794][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:53:25,293][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:53:25,794][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:53:26,297][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:53:26,805][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:53:27,304][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:53:27,808][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:53:28,310][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:53:28,813][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:53:29,312][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:53:29,812][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:53:30,311][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:53:30,813][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:53:31,316][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:53:31,829][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:53:32,329][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:53:32,833][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:53:33,334][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:53:33,834][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:53:34,336][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:53:34,839][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:53:35,338][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:53:35,839][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:53:36,339][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:53:36,844][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:53:37,345][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:53:37,855][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:53:38,356][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:53:38,855][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:53:39,356][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:53:39,858][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:53:40,361][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:53:40,860][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:53:41,359][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:53:41,857][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:53:42,356][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:53:42,853][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:53:43,351][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:53:43,852][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:53:44,354][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:53:44,860][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10126 tokens. +[2026-03-26 07:53:45,750][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 60.49%, Block Peak % of device VRAM: 62.13%, ΔTime: 00:00:33 +[2026-03-26 07:53:46,502][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:53:46,505][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:53:46,507][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:53:47,800][__main__][INFO] - Iteration 691 took 51s (28.27% Gen, 69.21% Train). Generation: 14s, Training: 35s. Estimated remaining time: 31h 51m 57s. Estimated total time: 42h 40m 4s. Time estimates for 10 more iterations: 8m 32s, 100 more iterations: 1h 25m 20s, 500 more iterations: 7h 6m 40s. +[2026-03-26 07:53:47,803][__main__][INFO] - Starting iteration 691. +[2026-03-26 07:53:48,205][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 69 and human policies 1. +[2026-03-26 07:53:48,206][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:54:01,332][__main__][INFO] - Number of regex retries in iteration 691: 0 +[2026-03-26 07:54:01,333][__main__][INFO] - agents played in iteration 691 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:54:02,097][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:54:02,121][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:54:02,145][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:54:02,169][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:54:02,170][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:54:02,170][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:54:03,023][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:54:03,487][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:54:03,993][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:54:04,495][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:54:05,001][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:54:05,503][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:54:06,014][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:54:06,521][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:54:07,024][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:54:07,526][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:54:08,026][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:54:08,527][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:54:09,029][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:54:09,539][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:54:10,045][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:54:10,546][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:54:11,049][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:54:11,549][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:54:12,049][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:54:12,549][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:54:13,049][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:54:13,550][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:54:14,054][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:54:14,554][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:54:15,053][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:54:15,552][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:54:16,058][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:54:16,569][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:54:17,075][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:54:17,576][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:54:18,076][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:54:18,578][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:54:19,079][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:54:19,577][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:54:20,077][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:54:20,576][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:54:21,076][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:54:21,577][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:54:22,086][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:54:22,585][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:54:23,085][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:54:23,586][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:54:24,086][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:54:24,586][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:54:25,087][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:54:25,587][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:54:26,090][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:54:26,592][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:54:27,091][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:54:27,590][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:54:28,089][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:54:28,589][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:54:29,089][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:54:29,588][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:54:30,088][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:54:30,589][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:54:31,087][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:54:31,585][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:54:32,088][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:54:32,588][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:54:33,087][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:54:33,586][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:54:34,084][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:54:34,582][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:54:35,080][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9976 tokens. +[2026-03-26 07:54:35,949][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.02%, Current % of VRAM taken: 60.45%, Block Peak % of device VRAM: 62.07%, ΔTime: 00:00:32 +[2026-03-26 07:54:36,700][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:54:36,702][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:54:36,704][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:54:37,435][__main__][INFO] - Iteration 692 took 49s (26.66% Gen, 71.85% Train). Generation: 13s, Training: 35s. Estimated remaining time: 30h 12m 33s. Estimated total time: 41h 1m 30s. Time estimates for 10 more iterations: 8m 12s, 100 more iterations: 1h 22m 3s, 500 more iterations: 6h 50m 15s. +[2026-03-26 07:54:37,437][__main__][INFO] - Starting iteration 692. +[2026-03-26 07:54:37,837][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 69 and human policies 1. +[2026-03-26 07:54:37,838][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:54:52,693][__main__][INFO] - Number of regex retries in iteration 692: 0 +[2026-03-26 07:54:52,694][__main__][INFO] - agents played in iteration 692 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:54:53,461][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:54:53,490][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:54:53,517][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:54:53,543][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:54:53,544][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:54:53,544][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:54:54,416][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:54:54,876][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:54:55,386][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:54:55,889][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:54:56,398][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:54:56,907][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:54:57,410][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:54:57,912][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:54:58,415][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:54:58,916][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:54:59,416][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:54:59,916][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:55:00,420][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:55:00,925][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:55:01,425][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:55:01,926][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:55:02,427][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:55:02,930][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:55:03,429][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:55:03,939][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:55:04,442][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:55:04,946][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:55:05,447][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:55:05,948][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:55:06,449][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:55:06,949][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:55:07,448][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:55:07,949][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:55:08,456][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:55:08,963][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:55:09,465][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:55:09,965][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:55:10,467][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:55:10,968][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:55:11,468][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:55:11,967][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:55:12,466][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:55:12,964][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:55:13,464][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:55:13,963][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:55:14,463][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:55:14,961][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:55:15,459][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:55:15,960][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:55:16,461][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:55:16,960][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:55:17,460][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:55:17,961][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:55:18,466][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:55:18,970][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:55:19,472][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:55:19,976][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:55:20,477][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:55:20,979][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:55:21,479][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:55:21,979][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:55:22,481][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:55:22,981][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:55:23,482][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:55:23,983][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:55:24,483][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:55:24,986][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:55:25,493][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:55:26,001][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:55:26,512][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10043 tokens. +[2026-03-26 07:55:27,405][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 60.62%, Block Peak % of device VRAM: 62.05%, ΔTime: 00:00:32 +[2026-03-26 07:55:28,155][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:55:28,158][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:55:28,159][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:55:28,898][__main__][INFO] - Iteration 693 took 51s (29.09% Gen, 69.46% Train). Generation: 14s, Training: 35s. Estimated remaining time: 31h 43m 17s. Estimated total time: 42h 33m 5s. Time estimates for 10 more iterations: 8m 30s, 100 more iterations: 1h 25m 6s, 500 more iterations: 7h 5m 30s. +[2026-03-26 07:55:28,901][__main__][INFO] - Starting iteration 693. +[2026-03-26 07:55:29,305][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 69 and human policies 1. +[2026-03-26 07:55:29,306][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:55:44,150][__main__][INFO] - Number of regex retries in iteration 693: 0 +[2026-03-26 07:55:44,151][__main__][INFO] - agents played in iteration 693 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:55:44,917][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:55:44,943][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:55:44,970][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:55:44,997][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:55:44,998][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:55:44,998][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:55:45,848][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:55:46,310][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:55:46,818][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:55:47,319][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:55:47,821][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:55:48,334][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:55:48,836][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:55:49,339][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:55:49,841][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:55:50,345][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:55:50,844][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:55:51,343][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:55:51,844][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:55:52,346][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:55:52,845][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:55:53,344][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:55:53,842][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:55:54,340][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:55:54,842][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:55:55,344][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:55:55,841][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:55:56,338][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:55:56,835][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:55:57,332][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:55:57,828][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:55:58,325][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:55:58,824][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:55:59,321][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:55:59,819][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:56:00,318][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:56:00,816][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:56:01,315][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:56:01,818][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:56:02,323][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:56:02,821][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:56:03,319][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:56:03,817][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:56:04,316][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:56:04,815][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:56:05,313][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:56:05,810][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:56:06,312][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:56:06,810][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:56:07,339][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:56:07,837][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:56:08,338][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:56:08,843][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:56:09,342][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:56:09,840][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:56:10,339][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:56:10,838][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:56:11,337][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:56:11,835][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:56:12,336][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:56:12,836][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:56:13,337][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:56:13,834][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:56:14,338][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:56:14,842][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:56:15,356][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:56:15,857][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:56:16,359][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:56:16,860][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:56:17,360][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:56:17,862][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10144 tokens. +[2026-03-26 07:56:18,757][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.22%, ΔTime: 00:00:32 +[2026-03-26 07:56:19,507][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:56:19,509][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:56:19,511][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:56:20,245][__main__][INFO] - Iteration 694 took 50s (29.14% Gen, 69.41% Train). Generation: 14s, Training: 35s. Estimated remaining time: 31h 36m 20s. Estimated total time: 42h 27m 0s. Time estimates for 10 more iterations: 8m 29s, 100 more iterations: 1h 24m 54s, 500 more iterations: 7h 4m 30s. +[2026-03-26 07:56:20,247][__main__][INFO] - Starting iteration 694. +[2026-03-26 07:56:20,649][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 69 and human policies 1. +[2026-03-26 07:56:20,650][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:56:31,456][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 07:56:34,894][__main__][INFO] - Number of regex retries in iteration 694: 1 +[2026-03-26 07:56:34,895][__main__][INFO] - agents played in iteration 694 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:56:35,669][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:56:35,696][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:56:35,722][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:56:35,748][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:56:35,748][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:56:35,749][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:56:36,608][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:56:37,076][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:56:37,579][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:56:38,082][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:56:38,582][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:56:39,081][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:56:39,580][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:56:40,079][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:56:40,578][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:56:41,079][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:56:41,584][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:56:42,091][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:56:42,598][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:56:43,098][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:56:43,598][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:56:44,099][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:56:44,600][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:56:45,099][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:56:45,599][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:56:46,098][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:56:46,598][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:56:47,097][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:56:47,597][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:56:48,096][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:56:48,596][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:56:49,095][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:56:49,596][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:56:50,095][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:56:50,594][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:56:51,120][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:56:51,619][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:56:52,117][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:56:52,615][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:56:53,113][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:56:53,612][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:56:54,111][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:56:54,610][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:56:55,108][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:56:55,607][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:56:56,105][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:56:56,603][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:56:57,098][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:56:57,594][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:56:58,089][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:56:58,583][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:56:59,080][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:56:59,578][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:57:00,076][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:57:00,573][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:57:01,073][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:57:01,571][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:57:02,071][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:57:02,569][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:57:03,066][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:57:03,575][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:57:04,074][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:57:04,577][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:57:05,085][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:57:05,609][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:57:06,112][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:57:06,617][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:57:07,120][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:57:07,623][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:57:08,124][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:57:08,625][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9956 tokens. +[2026-03-26 07:57:09,502][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.58%, Block Peak % of device VRAM: 62.00%, ΔTime: 00:00:32 +[2026-03-26 07:57:10,255][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:57:10,257][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:57:10,259][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:57:11,143][__main__][INFO] - Iteration 695 took 50s (28.21% Gen, 70.03% Train). Generation: 14s, Training: 35s. Estimated remaining time: 31h 13m 14s. Estimated total time: 42h 4m 45s. Time estimates for 10 more iterations: 8m 24s, 100 more iterations: 1h 24m 9s, 500 more iterations: 7h 0m 47s. +[2026-03-26 07:57:11,146][__main__][INFO] - Starting iteration 695. +[2026-03-26 07:57:11,554][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 69 and human policies 1. +[2026-03-26 07:57:11,555][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:57:27,077][__main__][INFO] - Number of regex retries in iteration 695: 0 +[2026-03-26 07:57:27,078][__main__][INFO] - agents played in iteration 695 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:57:27,850][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:57:27,876][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:57:27,902][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:57:27,929][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:57:27,930][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:57:27,931][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:57:28,775][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:57:29,235][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:57:29,746][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:57:30,258][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:57:30,761][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:57:31,270][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:57:31,772][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:57:32,275][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:57:32,776][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:57:33,285][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:57:33,791][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:57:34,290][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:57:34,791][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:57:35,290][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:57:35,789][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:57:36,291][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:57:36,789][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:57:37,291][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:57:37,789][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:57:38,287][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:57:38,785][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:57:39,283][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:57:39,786][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:57:40,294][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:57:40,797][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:57:41,301][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:57:41,800][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:57:42,304][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:57:42,802][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:57:43,300][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:57:43,800][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:57:44,302][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:57:44,808][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:57:45,315][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:57:45,815][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:57:46,315][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:57:46,812][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:57:47,310][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:57:47,808][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:57:48,306][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:57:48,807][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:57:49,308][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:57:49,811][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:57:50,312][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:57:50,814][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:57:51,315][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:57:51,815][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:57:52,313][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:57:52,815][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:57:53,312][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:57:53,815][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:57:54,315][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:57:54,820][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:57:55,322][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:57:55,822][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:57:56,321][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:57:56,820][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:57:57,318][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:57:57,816][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:57:58,314][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:57:58,812][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:57:59,309][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:57:59,807][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:58:00,303][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:58:00,800][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10078 tokens. +[2026-03-26 07:58:01,688][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.02%, Current % of VRAM taken: 60.45%, Block Peak % of device VRAM: 62.07%, ΔTime: 00:00:32 +[2026-03-26 07:58:02,428][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:58:02,430][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:58:02,432][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:58:03,300][__main__][INFO] - Iteration 696 took 51s (30.00% Gen, 68.32% Train). Generation: 15s, Training: 35s. Estimated remaining time: 32h 14m 58s. Estimated total time: 43h 7m 21s. Time estimates for 10 more iterations: 8m 37s, 100 more iterations: 1h 26m 14s, 500 more iterations: 7h 11m 13s. +[2026-03-26 07:58:03,303][__main__][INFO] - Starting iteration 696. +[2026-03-26 07:58:03,718][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 69 and human policies 1. +[2026-03-26 07:58:03,719][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:58:08,281][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 07:58:17,593][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 1 z ball did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 07:58:18,450][__main__][INFO] - Number of regex retries in iteration 696: 2 +[2026-03-26 07:58:18,451][__main__][INFO] - agents played in iteration 696 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:58:19,218][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:58:19,242][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:58:19,266][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:58:19,291][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:58:19,291][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:58:19,292][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:58:20,129][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:58:20,588][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:58:21,091][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:58:21,589][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:58:22,091][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:58:22,587][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:58:23,086][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:58:23,583][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:58:24,089][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:58:24,601][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:58:25,104][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:58:25,605][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:58:26,106][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:58:26,606][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:58:27,108][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:58:27,608][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:58:28,114][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:58:28,615][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:58:29,119][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:58:29,623][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:58:30,121][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:58:30,625][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:58:31,127][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:58:31,635][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:58:32,145][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:58:32,647][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:58:33,152][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:58:33,656][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:58:34,164][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:58:34,667][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:58:35,169][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:58:35,670][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:58:36,172][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:58:36,675][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:58:37,176][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:58:37,675][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:58:38,175][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:58:38,674][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:58:39,172][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:58:39,671][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:58:40,169][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:58:40,666][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:58:41,163][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:58:41,662][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:58:42,163][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:58:42,660][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:58:43,159][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:58:43,659][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:58:44,156][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:58:44,656][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:58:45,155][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:58:45,655][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:58:46,155][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:58:46,654][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:58:47,154][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:58:47,653][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:58:48,154][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:58:48,651][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:58:49,147][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:58:49,643][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:58:50,140][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:58:50,636][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:58:51,131][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:58:51,627][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:58:52,124][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10048 tokens. +[2026-03-26 07:58:53,005][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 60.51%, Block Peak % of device VRAM: 62.34%, ΔTime: 00:00:32 +[2026-03-26 07:58:53,751][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:58:53,754][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:58:53,755][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:58:54,496][__main__][INFO] - Iteration 697 took 50s (29.01% Gen, 69.53% Train). Generation: 14s, Training: 35s. Estimated remaining time: 31h 25m 41s. Estimated total time: 42h 18m 55s. Time estimates for 10 more iterations: 8m 27s, 100 more iterations: 1h 24m 37s, 500 more iterations: 7h 3m 9s. +[2026-03-26 07:58:54,498][__main__][INFO] - Starting iteration 697. +[2026-03-26 07:58:54,905][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 69 and human policies 1. +[2026-03-26 07:58:54,905][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:59:08,656][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 07:59:13,882][__main__][INFO] - Number of regex retries in iteration 697: 1 +[2026-03-26 07:59:13,883][__main__][INFO] - agents played in iteration 697 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 07:59:14,653][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:59:14,678][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:59:14,702][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:59:14,726][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 07:59:14,727][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 07:59:14,727][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 07:59:15,584][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 07:59:16,050][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 07:59:16,563][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 07:59:17,066][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 07:59:17,569][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 07:59:18,076][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 07:59:18,588][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 07:59:19,100][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 07:59:19,606][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 07:59:20,106][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 07:59:20,607][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 07:59:21,110][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 07:59:21,616][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 07:59:22,129][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 07:59:22,631][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 07:59:23,132][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 07:59:23,634][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 07:59:24,136][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 07:59:24,636][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 07:59:25,137][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 07:59:25,642][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 07:59:26,154][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 07:59:26,656][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 07:59:27,162][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 07:59:27,663][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 07:59:28,163][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 07:59:28,666][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 07:59:29,170][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 07:59:29,673][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 07:59:30,174][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 07:59:30,684][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 07:59:31,193][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 07:59:31,694][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 07:59:32,198][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 07:59:32,698][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 07:59:33,202][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 07:59:33,701][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 07:59:34,199][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 07:59:34,697][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 07:59:35,196][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 07:59:35,698][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 07:59:36,205][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 07:59:36,713][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 07:59:37,221][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 07:59:37,721][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 07:59:38,225][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 07:59:38,729][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 07:59:39,233][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 07:59:39,734][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 07:59:40,234][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 07:59:40,736][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 07:59:41,235][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 07:59:41,734][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 07:59:42,232][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 07:59:42,733][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 07:59:43,237][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 07:59:43,742][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 07:59:44,248][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 07:59:44,751][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 07:59:45,251][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 07:59:45,753][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 07:59:46,254][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 07:59:46,756][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 07:59:47,256][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 07:59:47,757][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10122 tokens. +[2026-03-26 07:59:48,672][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 60.54%, Block Peak % of device VRAM: 62.33%, ΔTime: 00:00:33 +[2026-03-26 07:59:49,425][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 07:59:49,427][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 07:59:49,429][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 07:59:50,313][__main__][INFO] - Iteration 698 took 55s (34.25% Gen, 64.15% Train). Generation: 18s, Training: 35s. Estimated remaining time: 35h 16m 16s. Estimated total time: 46h 10m 26s. Time estimates for 10 more iterations: 9m 14s, 100 more iterations: 1h 32m 20s, 500 more iterations: 7h 41m 44s. +[2026-03-26 07:59:50,315][__main__][INFO] - Starting iteration 698. +[2026-03-26 07:59:50,725][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 69 and human policies 1. +[2026-03-26 07:59:50,726][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 07:59:56,259][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 08:00:06,741][__main__][INFO] - Number of regex retries in iteration 698: 1 +[2026-03-26 08:00:06,743][__main__][INFO] - agents played in iteration 698 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:00:07,517][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:00:07,542][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:00:07,566][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:00:07,591][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:00:07,591][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:00:07,592][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:00:08,411][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:00:08,871][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:00:09,378][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:00:09,890][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:00:10,399][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:00:10,903][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:00:11,407][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:00:11,907][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:00:12,408][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:00:12,909][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:00:13,410][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:00:13,911][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:00:14,419][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:00:14,949][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:00:15,455][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:00:15,962][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:00:16,464][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:00:16,971][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:00:17,477][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:00:17,978][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:00:18,480][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:00:18,984][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:00:19,486][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:00:19,987][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:00:20,491][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:00:20,993][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:00:21,492][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:00:21,995][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:00:22,494][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:00:22,998][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:00:23,498][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:00:24,000][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:00:24,503][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:00:25,012][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:00:25,516][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:00:26,020][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:00:26,522][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:00:27,026][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:00:27,527][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:00:28,026][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:00:28,524][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:00:29,027][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:00:29,530][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:00:30,031][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:00:30,533][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:00:31,041][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:00:31,547][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:00:32,057][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:00:32,568][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:00:33,071][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:00:33,574][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:00:34,077][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:00:34,580][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:00:35,083][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:00:35,584][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:00:36,083][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:00:36,586][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:00:37,083][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:00:37,581][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:00:38,079][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:00:38,578][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:00:39,075][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:00:39,572][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:00:40,067][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:00:40,564][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10262 tokens. +[2026-03-26 08:00:41,460][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 60.49%, Block Peak % of device VRAM: 62.40%, ΔTime: 00:00:33 +[2026-03-26 08:00:42,209][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:00:42,211][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:00:42,213][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:00:43,071][__main__][INFO] - Iteration 699 took 52s (30.60% Gen, 67.76% Train). Generation: 16s, Training: 35s. Estimated remaining time: 32h 42m 18s. Estimated total time: 43h 37m 21s. Time estimates for 10 more iterations: 8m 43s, 100 more iterations: 1h 27m 14s, 500 more iterations: 7h 16m 13s. +[2026-03-26 08:00:43,074][__main__][INFO] - Starting iteration 699. +[2026-03-26 08:00:43,481][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 69 and human policies 1. +[2026-03-26 08:00:43,482][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:01:00,751][__main__][INFO] - Number of regex retries in iteration 699: 0 +[2026-03-26 08:01:00,752][__main__][INFO] - agents played in iteration 699 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:01:01,518][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:01:01,541][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:01:01,565][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:01:01,588][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:01:01,589][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:01:01,590][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:01:02,427][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:01:02,887][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:01:03,400][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:01:03,901][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:01:04,402][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:01:04,903][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:01:05,409][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:01:05,920][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:01:06,421][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:01:06,926][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:01:07,428][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:01:07,930][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:01:08,440][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:01:08,944][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:01:09,447][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:01:09,951][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:01:10,455][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:01:10,956][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:01:11,456][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:01:11,956][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:01:12,455][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:01:12,954][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:01:13,453][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:01:13,953][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:01:14,453][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:01:14,956][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:01:15,456][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:01:15,959][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:01:16,459][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:01:16,960][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:01:17,463][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:01:17,962][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:01:18,465][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:01:18,964][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:01:19,462][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:01:19,961][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:01:20,460][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:01:20,961][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:01:21,458][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:01:21,959][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:01:22,457][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:01:22,955][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:01:23,450][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:01:23,946][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:01:24,443][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:01:24,939][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:01:25,436][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:01:25,933][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:01:26,438][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:01:26,952][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:01:27,459][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:01:27,960][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:01:28,460][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:01:28,962][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:01:29,462][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:01:29,962][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:01:30,462][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:01:30,962][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:01:31,462][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:01:31,964][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:01:32,467][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:01:32,966][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:01:33,467][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:01:33,978][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:01:34,485][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10104 tokens. +[2026-03-26 08:01:35,372][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 60.54%, Block Peak % of device VRAM: 62.27%, ΔTime: 00:00:32 +[2026-03-26 08:01:36,121][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:01:36,123][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:01:36,125][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:01:37,064][__main__][INFO] - Iteration 700 took 53s (32.23% Gen, 66.01% Train). Generation: 17s, Training: 35s. Estimated remaining time: 33h 43m 14s. Estimated total time: 44h 39m 11s. Time estimates for 10 more iterations: 8m 55s, 100 more iterations: 1h 29m 18s, 500 more iterations: 7h 26m 31s. +[2026-03-26 08:01:37,066][__main__][INFO] - Starting iteration 700. +[2026-03-26 08:01:37,475][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 69 and human policies 1. +[2026-03-26 08:01:37,476][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:01:50,749][__main__][INFO] - Number of regex retries in iteration 700: 0 +[2026-03-26 08:01:50,749][__main__][INFO] - agents played in iteration 700 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:01:51,516][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:01:51,540][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:01:51,565][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:01:51,589][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:01:51,590][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:01:51,590][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:01:52,435][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:01:52,898][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:01:53,406][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:01:53,911][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:01:54,412][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:01:54,913][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:01:55,418][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:01:55,917][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:01:56,415][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:01:56,912][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:01:57,410][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:01:57,908][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:01:58,409][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:01:58,911][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:01:59,415][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:01:59,920][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:02:00,424][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:02:00,927][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:02:01,430][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:02:01,936][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:02:02,439][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:02:02,941][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:02:03,444][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:02:03,945][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:02:04,446][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:02:04,943][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:02:05,442][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:02:05,940][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:02:06,441][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:02:06,940][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:02:07,438][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:02:07,936][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:02:08,434][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:02:08,932][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:02:09,428][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:02:09,924][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:02:10,421][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:02:10,918][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:02:11,416][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:02:11,913][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:02:12,414][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:02:12,913][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:02:13,412][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:02:13,910][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:02:14,408][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:02:14,908][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:02:15,406][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:02:15,908][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:02:16,406][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:02:16,912][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:02:17,411][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:02:17,911][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:02:18,436][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:02:18,941][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:02:19,442][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:02:19,945][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:02:20,454][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:02:20,966][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:02:21,470][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:02:21,973][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:02:22,476][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:02:22,979][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:02:23,486][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:02:23,991][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:02:24,494][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10170 tokens. +[2026-03-26 08:02:25,391][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 60.60%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 +[2026-03-26 08:02:26,145][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:02:26,147][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:02:26,149][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:02:27,497][__main__][INFO] - Iteration 701 took 50s (26.53% Gen, 70.77% Train). Generation: 13s, Training: 35s. Estimated remaining time: 30h 44m 19s. Estimated total time: 41h 41m 7s. Time estimates for 10 more iterations: 8m 20s, 100 more iterations: 1h 23m 22s, 500 more iterations: 6h 56m 51s. +[2026-03-26 08:02:27,500][__main__][INFO] - Starting iteration 701. +[2026-03-26 08:02:27,905][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 70 and human policies 1. +[2026-03-26 08:02:27,905][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:02:32,721][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 08:02:32,804][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 08:02:42,256][__main__][INFO] - Number of regex retries in iteration 701: 2 +[2026-03-26 08:02:42,256][__main__][INFO] - agents played in iteration 701 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:02:43,028][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:02:43,052][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:02:43,076][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:02:43,100][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:02:43,100][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:02:43,101][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:02:43,927][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:02:44,384][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:02:44,889][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:02:45,390][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:02:45,889][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:02:46,388][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:02:46,886][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:02:47,385][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:02:47,882][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:02:48,379][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:02:48,879][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:02:49,375][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:02:49,873][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:02:50,370][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:02:50,874][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:02:51,370][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:02:51,867][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:02:52,364][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:02:52,862][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:02:53,360][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:02:53,858][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:02:54,357][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:02:54,854][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:02:55,358][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:02:55,857][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:02:56,362][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:02:56,863][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:02:57,367][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:02:57,876][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:02:58,378][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:02:58,880][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:02:59,382][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:02:59,883][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:03:00,381][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:03:00,881][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:03:01,382][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:03:01,885][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:03:02,386][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:03:02,884][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:03:03,383][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:03:03,882][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:03:04,386][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:03:04,886][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:03:05,384][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:03:05,882][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:03:06,383][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:03:06,883][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:03:07,382][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:03:07,880][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:03:08,385][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:03:08,886][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:03:09,385][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:03:09,886][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:03:10,389][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:03:10,890][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:03:11,393][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:03:11,896][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:03:12,397][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:03:12,897][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:03:13,401][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:03:13,901][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:03:14,401][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:03:14,900][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:03:15,399][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:03:15,899][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10067 tokens. +[2026-03-26 08:03:16,818][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 60.48%, Block Peak % of device VRAM: 62.03%, ΔTime: 00:00:32 +[2026-03-26 08:03:17,681][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:03:17,684][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:03:17,685][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:03:18,412][__main__][INFO] - Iteration 702 took 50s (28.41% Gen, 70.14% Train). Generation: 14s, Training: 35s. Estimated remaining time: 31h 7m 45s. Estimated total time: 42h 5m 24s. Time estimates for 10 more iterations: 8m 25s, 100 more iterations: 1h 24m 10s, 500 more iterations: 7h 0m 54s. +[2026-03-26 08:03:18,414][__main__][INFO] - Starting iteration 702. +[2026-03-26 08:03:18,819][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 70 and human policies 1. +[2026-03-26 08:03:18,819][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:03:32,995][__main__][INFO] - Number of regex retries in iteration 702: 0 +[2026-03-26 08:03:32,996][__main__][INFO] - agents played in iteration 702 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:03:33,766][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:03:33,792][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:03:33,821][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:03:33,849][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:03:33,850][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:03:33,850][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:03:34,730][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:03:35,195][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:03:35,701][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:03:36,211][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:03:36,717][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:03:37,227][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:03:37,726][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:03:38,226][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:03:38,728][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:03:39,230][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:03:39,738][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:03:40,245][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:03:40,746][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:03:41,249][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:03:41,749][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:03:42,250][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:03:42,750][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:03:43,251][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:03:43,751][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:03:44,258][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:03:44,759][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:03:45,259][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:03:45,758][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:03:46,258][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:03:46,758][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:03:47,258][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:03:47,761][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:03:48,260][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:03:48,759][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:03:49,258][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:03:49,759][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:03:50,259][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:03:50,759][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:03:51,267][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:03:51,769][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:03:52,269][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:03:52,772][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:03:53,274][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:03:53,776][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:03:54,276][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:03:54,779][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:03:55,278][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:03:55,776][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:03:56,278][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:03:56,776][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:03:57,278][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:03:57,779][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:03:58,278][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:03:58,777][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:03:59,279][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:03:59,779][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:04:00,279][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:04:00,779][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:04:01,281][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:04:01,780][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:04:02,281][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:04:02,782][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:04:03,280][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:04:03,778][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:04:04,280][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:04:04,784][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:04:05,288][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:04:05,797][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:04:06,301][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:04:06,802][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10175 tokens. +[2026-03-26 08:04:07,705][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.04%, Current % of VRAM taken: 60.47%, Block Peak % of device VRAM: 62.34%, ΔTime: 00:00:32 +[2026-03-26 08:04:08,459][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:04:08,461][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:04:08,463][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:04:09,229][__main__][INFO] - Iteration 703 took 50s (28.12% Gen, 70.36% Train). Generation: 14s, Training: 35s. Estimated remaining time: 31h 2m 4s. Estimated total time: 42h 0m 33s. Time estimates for 10 more iterations: 8m 24s, 100 more iterations: 1h 24m 1s, 500 more iterations: 7h 0m 5s. +[2026-03-26 08:04:09,231][__main__][INFO] - Starting iteration 703. +[2026-03-26 08:04:09,644][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 70 and human policies 1. +[2026-03-26 08:04:09,644][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:04:25,191][__main__][INFO] - Number of regex retries in iteration 703: 0 +[2026-03-26 08:04:25,192][__main__][INFO] - agents played in iteration 703 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:04:25,975][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:04:25,999][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:04:26,023][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:04:26,047][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:04:26,048][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:04:26,048][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:04:26,894][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:04:27,360][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:04:27,872][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:04:28,374][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:04:28,880][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:04:29,387][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:04:29,887][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:04:30,388][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:04:30,889][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:04:31,390][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:04:31,897][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:04:32,402][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:04:32,902][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:04:33,403][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:04:33,902][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:04:34,401][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:04:34,900][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:04:35,399][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:04:35,900][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:04:36,405][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:04:36,913][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:04:37,423][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:04:37,923][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:04:38,423][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:04:38,925][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:04:39,426][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:04:39,927][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:04:40,426][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:04:40,927][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:04:41,431][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:04:41,931][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:04:42,433][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:04:42,942][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:04:43,449][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:04:43,952][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:04:44,452][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:04:44,957][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:04:45,458][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:04:45,958][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:04:46,458][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:04:46,959][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:04:47,461][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:04:47,959][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:04:48,458][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:04:48,961][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:04:49,459][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:04:49,962][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:04:50,463][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:04:50,965][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:04:51,469][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:04:51,969][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:04:52,469][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:04:52,976][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:04:53,478][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:04:53,982][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:04:54,482][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:04:54,985][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:04:55,482][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:04:55,984][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:04:56,482][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:04:56,980][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:04:57,478][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:04:57,977][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:04:58,476][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:04:58,979][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10104 tokens. +[2026-03-26 08:04:59,896][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.14%, ΔTime: 00:00:33 +[2026-03-26 08:05:00,647][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:05:00,650][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:05:00,651][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:05:01,382][__main__][INFO] - Iteration 704 took 51s (30.05% Gen, 68.54% Train). Generation: 15s, Training: 35s. Estimated remaining time: 32h 7m 35s. Estimated total time: 43h 6m 56s. Time estimates for 10 more iterations: 8m 37s, 100 more iterations: 1h 26m 13s, 500 more iterations: 7h 11m 9s. +[2026-03-26 08:05:01,384][__main__][INFO] - Starting iteration 704. +[2026-03-26 08:05:01,791][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 70 and human policies 1. +[2026-03-26 08:05:01,791][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:05:16,640][__main__][INFO] - Number of regex retries in iteration 704: 0 +[2026-03-26 08:05:16,641][__main__][INFO] - agents played in iteration 704 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:05:17,426][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:05:17,450][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:05:17,474][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:05:17,497][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:05:17,498][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:05:17,498][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:05:18,340][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:05:18,800][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:05:19,312][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:05:19,817][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:05:20,318][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:05:20,820][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:05:21,321][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:05:21,821][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:05:22,322][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:05:22,826][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:05:23,325][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:05:23,823][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:05:24,323][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:05:24,822][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:05:25,320][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:05:25,820][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:05:26,317][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:05:26,818][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:05:27,319][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:05:27,819][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:05:28,320][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:05:28,822][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:05:29,322][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:05:29,822][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:05:30,323][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:05:30,821][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:05:31,320][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:05:31,818][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:05:32,316][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:05:32,819][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:05:33,316][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:05:33,817][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:05:34,315][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:05:34,816][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:05:35,314][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:05:35,811][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:05:36,308][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:05:36,814][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:05:37,312][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:05:37,811][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:05:38,309][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:05:38,806][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:05:39,304][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:05:39,802][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:05:40,301][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:05:40,799][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:05:41,298][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:05:41,796][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:05:42,295][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:05:42,792][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:05:43,291][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:05:43,787][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:05:44,283][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:05:44,781][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:05:45,277][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:05:45,774][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:05:46,272][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:05:46,770][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:05:47,268][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:05:47,765][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:05:48,263][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:05:48,761][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:05:49,260][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:05:49,758][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:05:50,258][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10059 tokens. +[2026-03-26 08:05:51,132][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 60.50%, Block Peak % of device VRAM: 62.01%, ΔTime: 00:00:32 +[2026-03-26 08:05:51,883][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:05:51,885][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:05:51,887][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:05:52,628][__main__][INFO] - Iteration 705 took 50s (29.21% Gen, 69.33% Train). Generation: 14s, Training: 35s. Estimated remaining time: 31h 21m 41s. Estimated total time: 42h 21m 53s. Time estimates for 10 more iterations: 8m 28s, 100 more iterations: 1h 24m 43s, 500 more iterations: 7h 3m 38s. +[2026-03-26 08:05:52,630][__main__][INFO] - Starting iteration 705. +[2026-03-26 08:05:53,035][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 70 and human policies 1. +[2026-03-26 08:05:53,036][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:06:05,288][__main__][INFO] - Number of regex retries in iteration 705: 0 +[2026-03-26 08:06:05,289][__main__][INFO] - agents played in iteration 705 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:06:06,149][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:06:06,174][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:06:06,199][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:06:06,224][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:06:06,224][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:06:06,225][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:06:07,075][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:06:07,539][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:06:08,054][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:06:08,560][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:06:09,064][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:06:09,571][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:06:10,072][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:06:10,572][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:06:11,074][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:06:11,575][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:06:12,077][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:06:12,584][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:06:13,092][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:06:13,594][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:06:14,096][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:06:14,598][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:06:15,098][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:06:15,599][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:06:16,099][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:06:16,600][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:06:17,099][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:06:17,598][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:06:18,104][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:06:18,610][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:06:19,109][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:06:19,610][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:06:20,120][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:06:20,622][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:06:21,122][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:06:21,624][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:06:22,125][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:06:22,624][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:06:23,124][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:06:23,625][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:06:24,126][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:06:24,627][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:06:25,128][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:06:25,629][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:06:26,130][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:06:26,631][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:06:27,132][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:06:27,628][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:06:28,125][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:06:28,622][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:06:29,123][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:06:29,620][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:06:30,119][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:06:30,617][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:06:31,114][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:06:31,611][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:06:32,110][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:06:32,607][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:06:33,105][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:06:33,602][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:06:34,099][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:06:34,596][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:06:35,093][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:06:35,591][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:06:36,088][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:06:36,585][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:06:37,083][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:06:37,583][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:06:38,079][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:06:38,577][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:06:39,075][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10052 tokens. +[2026-03-26 08:06:39,965][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 60.50%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:32 +[2026-03-26 08:06:40,710][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:06:40,712][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:06:40,714][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:06:41,581][__main__][INFO] - Iteration 706 took 48s (25.24% Gen, 72.97% Train). Generation: 12s, Training: 35s. Estimated remaining time: 29h 26m 19s. Estimated total time: 40h 27m 21s. Time estimates for 10 more iterations: 8m 5s, 100 more iterations: 1h 20m 54s, 500 more iterations: 6h 44m 33s. +[2026-03-26 08:06:41,584][__main__][INFO] - Starting iteration 706. +[2026-03-26 08:06:42,001][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 70 and human policies 1. +[2026-03-26 08:06:42,001][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:06:58,281][__main__][INFO] - Number of regex retries in iteration 706: 0 +[2026-03-26 08:06:58,282][__main__][INFO] - agents played in iteration 706 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:06:59,065][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:06:59,089][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:06:59,112][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:06:59,136][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:06:59,136][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:06:59,137][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:06:59,989][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:07:00,454][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:07:00,964][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:07:01,468][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:07:01,969][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:07:02,469][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:07:02,974][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:07:03,490][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:07:03,990][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:07:04,494][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:07:04,996][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:07:05,501][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:07:06,001][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:07:06,501][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:07:07,002][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:07:07,506][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:07:08,005][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:07:08,513][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:07:09,015][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:07:09,516][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:07:10,017][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:07:10,519][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:07:11,023][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:07:11,536][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:07:12,048][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:07:12,548][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:07:13,050][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:07:13,552][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:07:14,053][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:07:14,554][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:07:15,055][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:07:15,559][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:07:16,070][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:07:16,577][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:07:17,082][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:07:17,583][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:07:18,083][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:07:18,585][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:07:19,083][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:07:19,582][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:07:20,081][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:07:20,584][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:07:21,090][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:07:21,599][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:07:22,104][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:07:22,623][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:07:23,129][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:07:23,629][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:07:24,136][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:07:24,635][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:07:25,134][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:07:25,633][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:07:26,132][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:07:26,636][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:07:27,146][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:07:27,654][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:07:28,155][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:07:28,659][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:07:29,160][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:07:29,661][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:07:30,161][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:07:30,661][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:07:31,163][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:07:31,666][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:07:32,167][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10143 tokens. +[2026-03-26 08:07:33,047][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 60.61%, Block Peak % of device VRAM: 62.28%, ΔTime: 00:00:33 +[2026-03-26 08:07:33,809][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:07:33,811][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:07:33,812][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:07:34,551][__main__][INFO] - Iteration 707 took 52s (30.98% Gen, 67.61% Train). Generation: 16s, Training: 35s. Estimated remaining time: 32h 45m 39s. Estimated total time: 43h 47m 33s. Time estimates for 10 more iterations: 8m 45s, 100 more iterations: 1h 27m 35s, 500 more iterations: 7h 17m 55s. +[2026-03-26 08:07:34,553][__main__][INFO] - Starting iteration 707. +[2026-03-26 08:07:34,960][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 70 and human policies 1. +[2026-03-26 08:07:34,961][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:07:50,046][__main__][INFO] - Number of regex retries in iteration 707: 0 +[2026-03-26 08:07:50,048][__main__][INFO] - agents played in iteration 707 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:07:50,813][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:07:50,837][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:07:50,862][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:07:50,886][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:07:50,887][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:07:50,887][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:07:51,730][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:07:52,193][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:07:52,711][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:07:53,225][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:07:53,727][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:07:54,232][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:07:54,739][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:07:55,242][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:07:55,745][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:07:56,248][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:07:56,753][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:07:57,265][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:07:57,781][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:07:58,286][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:07:58,790][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:07:59,295][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:07:59,798][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:08:00,300][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:08:00,800][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:08:01,301][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:08:01,802][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:08:02,306][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:08:02,810][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:08:03,316][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:08:03,819][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:08:04,322][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:08:04,824][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:08:05,325][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:08:05,829][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:08:06,329][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:08:06,830][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:08:07,331][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:08:07,835][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:08:08,335][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:08:08,839][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:08:09,341][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:08:09,842][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:08:10,344][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:08:10,844][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:08:11,349][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:08:11,859][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:08:12,364][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:08:12,872][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:08:13,374][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:08:13,879][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:08:14,383][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:08:14,885][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:08:15,386][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:08:15,887][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:08:16,392][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:08:16,891][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:08:17,391][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:08:17,893][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:08:18,397][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:08:18,898][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:08:19,401][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:08:19,902][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:08:20,402][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:08:20,901][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:08:21,401][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:08:21,903][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:08:22,403][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:08:22,902][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:08:23,401][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:08:23,900][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10144 tokens. +[2026-03-26 08:08:24,791][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 60.56%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:33 +[2026-03-26 08:08:25,554][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:08:25,556][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:08:25,558][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:08:26,300][__main__][INFO] - Iteration 708 took 51s (29.39% Gen, 69.17% Train). Generation: 15s, Training: 35s. Estimated remaining time: 31h 44m 14s. Estimated total time: 42h 47m 0s. Time estimates for 10 more iterations: 8m 33s, 100 more iterations: 1h 25m 34s, 500 more iterations: 7h 7m 50s. +[2026-03-26 08:08:26,302][__main__][INFO] - Starting iteration 708. +[2026-03-26 08:08:26,707][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 70 and human policies 1. +[2026-03-26 08:08:26,707][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:08:39,392][__main__][INFO] - Number of regex retries in iteration 708: 0 +[2026-03-26 08:08:39,393][__main__][INFO] - agents played in iteration 708 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:08:40,156][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:08:40,180][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:08:40,204][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:08:40,228][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:08:40,228][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:08:40,229][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:08:41,084][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:08:41,538][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:08:42,046][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:08:42,545][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:08:43,042][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:08:43,540][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:08:44,038][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:08:44,539][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:08:45,041][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:08:45,542][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:08:46,047][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:08:46,553][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:08:47,057][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:08:47,563][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:08:48,071][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:08:48,575][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:08:49,077][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:08:49,578][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:08:50,079][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:08:50,581][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:08:51,087][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:08:51,586][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:08:52,086][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:08:52,586][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:08:53,093][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:08:53,593][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:08:54,092][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:08:54,591][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:08:55,090][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:08:55,589][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:08:56,090][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:08:56,590][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:08:57,098][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:08:57,602][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:08:58,104][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:08:58,613][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:08:59,124][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:08:59,624][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:09:00,125][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:09:00,625][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:09:01,126][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:09:01,627][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:09:02,129][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:09:02,634][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:09:03,133][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:09:03,632][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:09:04,136][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:09:04,634][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:09:05,135][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:09:05,636][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:09:06,135][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:09:06,636][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:09:07,135][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:09:07,633][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:09:08,131][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:09:08,628][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:09:09,126][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:09:09,625][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:09:10,127][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:09:10,625][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:09:11,125][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:09:11,623][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:09:12,122][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:09:12,621][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:09:13,121][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10159 tokens. +[2026-03-26 08:09:14,009][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.18%, ΔTime: 00:00:32 +[2026-03-26 08:09:14,758][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:09:14,761][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:09:14,762][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:09:15,492][__main__][INFO] - Iteration 709 took 48s (26.00% Gen, 72.50% Train). Generation: 12s, Training: 35s. Estimated remaining time: 29h 35m 43s. Estimated total time: 40h 39m 18s. Time estimates for 10 more iterations: 8m 7s, 100 more iterations: 1h 21m 18s, 500 more iterations: 6h 46m 33s. +[2026-03-26 08:09:15,494][__main__][INFO] - Starting iteration 709. +[2026-03-26 08:09:15,906][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 70 and human policies 1. +[2026-03-26 08:09:15,907][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:09:32,183][__main__][INFO] - Number of regex retries in iteration 709: 0 +[2026-03-26 08:09:32,184][__main__][INFO] - agents played in iteration 709 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:09:32,957][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:09:32,981][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:09:33,004][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:09:33,028][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:09:33,028][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:09:33,029][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:09:33,858][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:09:34,316][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:09:34,822][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:09:35,321][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:09:35,826][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:09:36,326][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:09:36,829][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:09:37,337][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:09:37,839][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:09:38,345][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:09:38,850][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:09:39,352][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:09:39,852][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:09:40,353][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:09:40,854][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:09:41,356][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:09:41,860][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:09:42,362][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:09:42,865][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:09:43,366][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:09:43,869][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:09:44,370][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:09:44,871][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:09:45,372][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:09:45,880][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:09:46,379][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:09:46,877][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:09:47,376][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:09:47,874][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:09:48,375][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:09:48,871][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:09:49,369][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:09:49,867][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:09:50,369][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:09:50,870][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:09:51,368][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:09:51,871][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:09:52,373][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:09:52,879][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:09:53,380][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:09:53,884][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:09:54,388][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:09:54,913][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:09:55,416][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:09:55,919][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:09:56,421][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:09:56,923][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:09:57,426][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:09:57,928][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:09:58,427][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:09:58,924][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:09:59,424][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:09:59,923][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:10:00,423][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:10:00,922][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:10:01,420][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:10:01,921][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:10:02,418][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:10:02,920][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:10:03,416][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:10:03,916][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:10:04,416][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:10:04,913][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:10:05,409][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:10:05,906][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10217 tokens. +[2026-03-26 08:10:06,781][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 60.54%, Block Peak % of device VRAM: 62.44%, ΔTime: 00:00:32 +[2026-03-26 08:10:07,536][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:10:07,538][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:10:07,540][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:10:08,270][__main__][INFO] - Iteration 710 took 52s (31.08% Gen, 67.52% Train). Generation: 16s, Training: 35s. Estimated remaining time: 32h 33m 44s. Estimated total time: 43h 38m 12s. Time estimates for 10 more iterations: 8m 43s, 100 more iterations: 1h 27m 16s, 500 more iterations: 7h 16m 22s. +[2026-03-26 08:10:08,272][__main__][INFO] - Starting iteration 710. +[2026-03-26 08:10:08,669][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 70 and human policies 1. +[2026-03-26 08:10:08,670][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:10:13,882][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 08:10:23,943][__main__][INFO] - Number of regex retries in iteration 710: 1 +[2026-03-26 08:10:23,944][__main__][INFO] - agents played in iteration 710 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:10:24,700][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:10:24,724][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:10:24,747][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:10:24,771][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:10:24,771][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:10:24,772][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:10:25,605][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:10:26,064][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:10:26,568][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:10:27,067][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:10:27,566][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:10:28,067][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:10:28,565][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:10:29,065][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:10:29,565][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:10:30,065][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:10:30,566][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:10:31,065][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:10:31,563][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:10:32,064][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:10:32,564][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:10:33,063][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:10:33,563][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:10:34,062][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:10:34,563][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:10:35,066][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:10:35,566][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:10:36,077][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:10:36,580][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:10:37,082][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:10:37,587][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:10:38,086][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:10:38,597][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:10:39,097][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:10:39,597][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:10:40,095][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:10:40,594][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:10:41,093][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:10:41,590][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:10:42,090][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:10:42,594][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:10:43,096][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:10:43,595][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:10:44,095][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:10:44,599][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:10:45,102][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:10:45,601][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:10:46,103][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:10:46,602][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:10:47,101][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:10:47,601][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:10:48,102][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:10:48,601][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:10:49,100][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:10:49,602][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:10:50,099][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:10:50,596][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:10:51,094][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:10:51,591][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:10:52,088][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:10:52,585][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:10:53,083][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:10:53,582][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:10:54,083][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:10:54,583][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:10:55,083][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:10:55,587][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:10:56,086][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:10:56,585][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:10:57,085][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:10:57,588][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10110 tokens. +[2026-03-26 08:10:58,471][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 60.60%, Block Peak % of device VRAM: 62.18%, ΔTime: 00:00:32 +[2026-03-26 08:10:59,227][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:10:59,229][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:10:59,231][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:11:00,576][__main__][INFO] - Iteration 711 took 51s (29.42% Gen, 67.98% Train). Generation: 15s, Training: 35s. Estimated remaining time: 32h 10m 1s. Estimated total time: 43h 15m 21s. Time estimates for 10 more iterations: 8m 39s, 100 more iterations: 1h 26m 30s, 500 more iterations: 7h 12m 33s. +[2026-03-26 08:11:00,578][__main__][INFO] - Starting iteration 711. +[2026-03-26 08:11:00,979][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 71 and human policies 1. +[2026-03-26 08:11:00,979][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:11:17,449][__main__][INFO] - Number of regex retries in iteration 711: 0 +[2026-03-26 08:11:17,449][__main__][INFO] - agents played in iteration 711 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:11:18,206][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:11:18,230][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:11:18,254][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:11:18,278][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:11:18,278][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:11:18,279][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:11:19,095][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:11:19,554][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:11:20,058][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:11:20,566][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:11:21,063][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:11:21,562][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:11:22,060][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:11:22,558][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:11:23,065][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:11:23,575][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:11:24,082][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:11:24,586][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:11:25,095][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:11:25,595][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:11:26,095][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:11:26,595][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:11:27,102][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:11:27,606][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:11:28,117][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:11:28,621][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:11:29,123][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:11:29,624][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:11:30,151][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:11:30,650][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:11:31,150][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:11:31,656][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:11:32,167][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:11:32,681][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:11:33,192][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:11:33,696][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:11:34,203][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:11:34,706][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:11:35,210][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:11:35,715][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:11:36,218][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:11:36,720][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:11:37,224][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:11:37,725][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:11:38,226][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:11:38,728][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:11:39,229][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:11:39,728][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:11:40,226][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:11:40,724][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:11:41,225][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:11:41,725][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:11:42,229][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:11:42,734][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:11:43,235][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:11:43,737][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:11:44,237][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:11:44,737][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:11:45,239][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:11:45,739][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:11:46,238][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:11:46,736][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:11:47,234][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:11:47,737][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:11:48,240][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:11:48,740][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:11:49,241][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:11:49,744][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:11:50,248][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:11:50,748][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:11:51,252][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10115 tokens. +[2026-03-26 08:11:52,160][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 60.54%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:33 +[2026-03-26 08:11:52,917][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:11:52,919][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:11:52,921][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:11:53,756][__main__][INFO] - Iteration 712 took 52s (31.21% Gen, 67.21% Train). Generation: 16s, Training: 35s. Estimated remaining time: 32h 52m 39s. Estimated total time: 43h 58m 52s. Time estimates for 10 more iterations: 8m 47s, 100 more iterations: 1h 27m 57s, 500 more iterations: 7h 19m 48s. +[2026-03-26 08:11:53,758][__main__][INFO] - Starting iteration 712. +[2026-03-26 08:11:54,174][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 71 and human policies 1. +[2026-03-26 08:11:54,175][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:12:07,875][__main__][INFO] - Number of regex retries in iteration 712: 0 +[2026-03-26 08:12:07,876][__main__][INFO] - agents played in iteration 712 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:12:08,647][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:12:08,670][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:12:08,693][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:12:08,717][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:12:08,718][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:12:08,718][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:12:09,552][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:12:10,007][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:12:10,509][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:12:11,011][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:12:11,511][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:12:12,015][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:12:12,514][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:12:13,021][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:12:13,527][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:12:14,031][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:12:14,536][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:12:15,036][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:12:15,536][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:12:16,040][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:12:16,545][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:12:17,044][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:12:17,545][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:12:18,043][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:12:18,541][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:12:19,039][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:12:19,538][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:12:20,036][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:12:20,534][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:12:21,031][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:12:21,528][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:12:22,030][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:12:22,527][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:12:23,025][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:12:23,524][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:12:24,026][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:12:24,523][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:12:25,022][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:12:25,520][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:12:26,017][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:12:26,512][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:12:27,010][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:12:27,507][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:12:28,006][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:12:28,512][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:12:29,011][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:12:29,514][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:12:30,013][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:12:30,521][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:12:31,022][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:12:31,522][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:12:32,023][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:12:32,523][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:12:33,025][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:12:33,525][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:12:34,026][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:12:34,526][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:12:35,028][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:12:35,530][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:12:36,033][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:12:36,537][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:12:37,039][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:12:37,540][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:12:38,038][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:12:38,536][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:12:39,034][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:12:39,532][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:12:40,030][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:12:40,532][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:12:41,030][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:12:41,528][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10091 tokens. +[2026-03-26 08:12:42,410][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 60.53%, Block Peak % of device VRAM: 62.11%, ΔTime: 00:00:32 +[2026-03-26 08:12:43,181][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:12:43,184][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:12:43,185][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:12:43,919][__main__][INFO] - Iteration 713 took 49s (27.54% Gen, 70.98% Train). Generation: 13s, Training: 35s. Estimated remaining time: 30h 20m 14s. Estimated total time: 41h 27m 18s. Time estimates for 10 more iterations: 8m 17s, 100 more iterations: 1h 22m 54s, 500 more iterations: 6h 54m 33s. +[2026-03-26 08:12:43,922][__main__][INFO] - Starting iteration 713. +[2026-03-26 08:12:44,331][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 71 and human policies 1. +[2026-03-26 08:12:44,332][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:12:58,479][__main__][INFO] - Number of regex retries in iteration 713: 0 +[2026-03-26 08:12:58,480][__main__][INFO] - agents played in iteration 713 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:12:59,243][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:12:59,269][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:12:59,294][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:12:59,320][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:12:59,320][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:12:59,321][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:13:00,109][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:13:00,571][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:13:01,081][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:13:01,595][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:13:02,102][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:13:02,605][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:13:03,105][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:13:03,605][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:13:04,105][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:13:04,605][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:13:05,108][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:13:05,607][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:13:06,109][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:13:06,608][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:13:07,108][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:13:07,610][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:13:08,119][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:13:08,628][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:13:09,133][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:13:09,637][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:13:10,140][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:13:10,642][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:13:11,143][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:13:11,649][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:13:12,159][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:13:12,662][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:13:13,162][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:13:13,662][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:13:14,162][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:13:14,662][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:13:15,162][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:13:15,662][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:13:16,165][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:13:16,666][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:13:17,166][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:13:17,665][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:13:18,164][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:13:18,663][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:13:19,170][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:13:19,677][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:13:20,182][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:13:20,685][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:13:21,186][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:13:21,685][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:13:22,186][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:13:22,687][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:13:23,186][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:13:23,688][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:13:24,186][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:13:24,686][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:13:25,186][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:13:25,690][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:13:26,190][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:13:26,690][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:13:27,198][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:13:27,698][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:13:28,200][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:13:28,705][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:13:29,211][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:13:29,713][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:13:30,214][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:13:30,714][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:13:31,214][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:13:31,713][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:13:32,214][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10064 tokens. +[2026-03-26 08:13:33,105][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 60.51%, Block Peak % of device VRAM: 62.06%, ΔTime: 00:00:33 +[2026-03-26 08:13:33,842][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:13:33,845][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:13:33,847][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:13:34,731][__main__][INFO] - Iteration 714 took 50s (28.07% Gen, 70.17% Train). Generation: 14s, Training: 35s. Estimated remaining time: 30h 52m 8s. Estimated total time: 42h 0m 2s. Time estimates for 10 more iterations: 8m 24s, 100 more iterations: 1h 24m 0s, 500 more iterations: 7h 0m 0s. +[2026-03-26 08:13:34,733][__main__][INFO] - Starting iteration 714. +[2026-03-26 08:13:35,149][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 71 and human policies 1. +[2026-03-26 08:13:35,150][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:13:50,115][__main__][INFO] - Number of regex retries in iteration 714: 0 +[2026-03-26 08:13:50,115][__main__][INFO] - agents played in iteration 714 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:13:50,884][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:13:50,908][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:13:50,932][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:13:50,957][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:13:50,958][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:13:50,958][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:13:51,800][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:13:52,258][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:13:52,773][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:13:53,274][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:13:53,776][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:13:54,275][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:13:54,773][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:13:55,271][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:13:55,769][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:13:56,271][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:13:56,779][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:13:57,290][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:13:57,789][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:13:58,290][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:13:58,787][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:13:59,287][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:13:59,786][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:14:00,287][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:14:00,786][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:14:01,284][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:14:01,784][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:14:02,283][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:14:02,781][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:14:03,279][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:14:03,777][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:14:04,276][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:14:04,778][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:14:05,284][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:14:05,795][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:14:06,297][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:14:06,798][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:14:07,297][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:14:07,801][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:14:08,325][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:14:08,826][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:14:09,325][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:14:09,827][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:14:10,325][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:14:10,826][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:14:11,327][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:14:11,831][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:14:12,333][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:14:12,831][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:14:13,331][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:14:13,829][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:14:14,331][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:14:14,840][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:14:15,343][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:14:15,847][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:14:16,346][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:14:16,845][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:14:17,344][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:14:17,845][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:14:18,347][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:14:18,845][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:14:19,344][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:14:19,845][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:14:20,346][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:14:20,847][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:14:21,350][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:14:21,852][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:14:22,360][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:14:22,862][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:14:23,366][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:14:23,870][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10102 tokens. +[2026-03-26 08:14:24,762][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 60.60%, Block Peak % of device VRAM: 62.09%, ΔTime: 00:00:32 +[2026-03-26 08:14:25,516][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:14:25,519][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:14:25,520][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:14:26,254][__main__][INFO] - Iteration 715 took 51s (29.28% Gen, 69.28% Train). Generation: 14s, Training: 35s. Estimated remaining time: 31h 26m 29s. Estimated total time: 42h 35m 15s. Time estimates for 10 more iterations: 8m 31s, 100 more iterations: 1h 25m 10s, 500 more iterations: 7h 5m 52s. +[2026-03-26 08:14:26,256][__main__][INFO] - Starting iteration 715. +[2026-03-26 08:14:26,658][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 71 and human policies 1. +[2026-03-26 08:14:26,658][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:14:48,651][mllm.models.large_language_model_local][WARNING] - Response Propposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 08:14:48,731][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 08:14:48,733][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 08:14:50,819][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 08:14:58,567][__main__][INFO] - Number of regex retries in iteration 715: 4 +[2026-03-26 08:14:58,568][__main__][INFO] - agents played in iteration 715 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:14:59,339][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:14:59,362][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:14:59,386][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:14:59,410][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:14:59,410][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:14:59,411][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:15:00,258][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:15:00,717][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:15:01,220][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:15:01,719][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:15:02,219][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:15:02,718][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:15:03,216][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:15:03,717][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:15:04,217][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:15:04,717][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:15:05,225][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:15:05,725][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:15:06,232][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:15:06,733][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:15:07,233][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:15:07,740][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:15:08,240][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:15:08,740][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:15:09,241][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:15:09,742][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:15:10,248][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:15:10,748][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:15:11,247][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:15:11,748][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:15:12,247][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:15:12,747][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:15:13,247][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:15:13,747][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:15:14,247][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:15:14,748][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:15:15,248][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:15:15,747][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:15:16,247][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:15:16,748][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:15:17,249][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:15:17,745][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:15:18,244][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:15:18,742][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:15:19,241][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:15:19,739][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:15:20,236][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:15:20,736][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:15:21,237][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:15:21,737][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:15:22,237][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:15:22,738][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:15:23,237][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:15:23,736][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:15:24,235][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:15:24,734][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:15:25,233][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:15:25,732][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:15:26,234][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:15:26,732][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:15:27,229][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:15:27,730][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:15:28,232][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:15:28,728][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:15:29,228][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:15:29,725][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:15:30,223][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:15:30,720][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:15:31,218][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:15:31,715][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:15:32,213][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10059 tokens. +[2026-03-26 08:15:33,093][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.03%, Current % of VRAM taken: 60.46%, Block Peak % of device VRAM: 62.12%, ΔTime: 00:00:32 +[2026-03-26 08:15:33,836][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:15:33,838][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:15:33,840][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:15:34,671][__main__][INFO] - Iteration 716 took 1m 8s (46.92% Gen, 51.86% Train). Generation: 31s, Training: 35s. Estimated remaining time: 45h 30m 46s. Estimated total time: 56h 40m 41s. Time estimates for 10 more iterations: 11m 20s, 100 more iterations: 1h 53m 21s, 500 more iterations: 9h 26m 46s. +[2026-03-26 08:15:34,673][__main__][INFO] - Starting iteration 716. +[2026-03-26 08:15:35,082][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 71 and human policies 1. +[2026-03-26 08:15:35,082][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:15:52,811][__main__][INFO] - Number of regex retries in iteration 716: 0 +[2026-03-26 08:15:52,812][__main__][INFO] - agents played in iteration 716 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:15:53,578][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.45%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:15:53,601][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.45%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:15:53,625][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.45%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:15:53,649][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.45%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:15:53,649][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:15:53,650][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:15:54,478][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:15:54,933][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:15:55,437][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:15:55,934][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:15:56,437][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:15:56,934][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:15:57,431][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:15:57,932][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:15:58,428][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:15:58,925][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:15:59,425][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:15:59,922][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:16:00,420][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:16:00,918][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:16:01,418][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:16:01,918][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:16:02,417][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:16:02,914][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:16:03,410][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:16:03,906][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:16:04,402][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:16:04,899][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:16:05,395][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:16:05,890][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:16:06,386][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:16:06,881][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:16:07,378][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:16:07,873][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:16:08,369][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:16:08,866][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:16:09,361][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:16:09,857][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:16:10,351][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:16:10,849][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:16:11,346][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:16:11,849][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:16:12,350][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:16:12,851][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:16:13,350][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:16:13,850][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:16:14,349][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:16:14,851][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:16:15,351][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:16:15,852][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:16:16,352][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:16:16,851][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:16:17,351][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:16:17,849][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:16:18,347][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:16:18,842][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:16:19,342][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:16:19,842][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:16:20,338][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:16:20,834][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:16:21,334][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:16:21,829][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:16:22,325][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:16:22,822][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:16:23,321][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:16:23,818][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:16:24,317][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:16:24,817][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:16:25,317][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:16:25,813][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:16:26,310][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10105 tokens. +[2026-03-26 08:16:27,172][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.05%, Current % of VRAM taken: 60.47%, Block Peak % of device VRAM: 62.21%, ΔTime: 00:00:32 +[2026-03-26 08:16:27,888][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:16:27,890][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:16:27,891][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:16:28,564][__main__][INFO] - Iteration 717 took 53s (33.15% Gen, 65.59% Train). Generation: 17s, Training: 35s. Estimated remaining time: 33h 23m 20s. Estimated total time: 44h 34m 8s. Time estimates for 10 more iterations: 8m 54s, 100 more iterations: 1h 29m 8s, 500 more iterations: 7h 25m 41s. +[2026-03-26 08:16:28,566][__main__][INFO] - Starting iteration 717. +[2026-03-26 08:16:28,967][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 71 and human policies 1. +[2026-03-26 08:16:28,968][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:16:32,587][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 08:16:41,430][__main__][INFO] - Number of regex retries in iteration 717: 1 +[2026-03-26 08:16:41,431][__main__][INFO] - agents played in iteration 717 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:16:42,193][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:16:42,216][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:16:42,240][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:16:42,263][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:16:42,264][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:16:42,264][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:16:43,103][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:16:43,560][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:16:44,066][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:16:44,562][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:16:45,060][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:16:45,558][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:16:46,058][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:16:46,562][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:16:47,059][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:16:47,559][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:16:48,059][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:16:48,558][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:16:49,062][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:16:49,562][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:16:50,062][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:16:50,562][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:16:51,084][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:16:51,584][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:16:52,083][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:16:52,581][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:16:53,079][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:16:53,578][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:16:54,077][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:16:54,574][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:16:55,072][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:16:55,572][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:16:56,072][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:16:56,569][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:16:57,067][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:16:57,565][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:16:58,065][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:16:58,564][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:16:59,063][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:16:59,561][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:17:00,060][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:17:00,559][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:17:01,061][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:17:01,559][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:17:02,061][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:17:02,558][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:17:03,056][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:17:03,558][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:17:04,058][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:17:04,555][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:17:05,059][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:17:05,556][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:17:06,055][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:17:06,552][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:17:07,050][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:17:07,550][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:17:08,051][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:17:08,559][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:17:09,061][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:17:09,569][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:17:10,073][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:17:10,586][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:17:11,088][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:17:11,589][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:17:12,089][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:17:12,593][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:17:13,096][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:17:13,597][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:17:14,098][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:17:14,599][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:17:15,103][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10161 tokens. +[2026-03-26 08:17:15,995][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 60.66%, Block Peak % of device VRAM: 62.23%, ΔTime: 00:00:32 +[2026-03-26 08:17:16,757][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:17:16,760][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:17:16,761][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:17:17,481][__main__][INFO] - Iteration 718 took 48s (25.69% Gen, 72.82% Train). Generation: 12s, Training: 35s. Estimated remaining time: 29h 14m 6s. Estimated total time: 40h 25m 43s. Time estimates for 10 more iterations: 8m 5s, 100 more iterations: 1h 20m 51s, 500 more iterations: 6h 44m 17s. +[2026-03-26 08:17:17,483][__main__][INFO] - Starting iteration 718. +[2026-03-26 08:17:17,897][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 71 and human policies 1. +[2026-03-26 08:17:17,898][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:17:33,171][__main__][INFO] - Number of regex retries in iteration 718: 0 +[2026-03-26 08:17:33,172][__main__][INFO] - agents played in iteration 718 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:17:33,951][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:17:33,976][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:17:34,000][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:17:34,025][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:17:34,025][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:17:34,026][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:17:34,856][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:17:35,318][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:17:35,829][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:17:36,339][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:17:36,844][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:17:37,363][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:17:37,867][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:17:38,371][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:17:38,873][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:17:39,375][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:17:39,877][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:17:40,382][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:17:40,894][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:17:41,405][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:17:41,909][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:17:42,414][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:17:42,917][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:17:43,419][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:17:43,919][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:17:44,420][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:17:44,921][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:17:45,425][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:17:45,923][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:17:46,424][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:17:46,922][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:17:47,425][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:17:47,930][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:17:48,447][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:17:48,948][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:17:49,450][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:17:49,955][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:17:50,460][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:17:50,962][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:17:51,462][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:17:51,966][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:17:52,471][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:17:52,974][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:17:53,480][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:17:53,982][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:17:54,483][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:17:54,983][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:17:55,485][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:17:55,986][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:17:56,487][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:17:56,988][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:17:57,488][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:17:57,990][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:17:58,491][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:17:58,994][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:17:59,493][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:17:59,993][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:18:00,490][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:18:00,988][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:18:01,492][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:18:02,006][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:18:02,514][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:18:03,014][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:18:03,517][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:18:04,022][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:18:04,523][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:18:05,022][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:18:05,521][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:18:06,021][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:18:06,523][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:18:07,023][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10089 tokens. +[2026-03-26 08:18:07,909][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.02%, Current % of VRAM taken: 60.45%, Block Peak % of device VRAM: 62.03%, ΔTime: 00:00:33 +[2026-03-26 08:18:08,665][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:18:08,668][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:18:08,669][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:18:09,426][__main__][INFO] - Iteration 719 took 51s (29.64% Gen, 68.89% Train). Generation: 15s, Training: 35s. Estimated remaining time: 31h 43m 58s. Estimated total time: 42h 56m 27s. Time estimates for 10 more iterations: 8m 35s, 100 more iterations: 1h 25m 52s, 500 more iterations: 7h 9m 24s. +[2026-03-26 08:18:09,428][__main__][INFO] - Starting iteration 719. +[2026-03-26 08:18:09,838][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 71 and human policies 1. +[2026-03-26 08:18:09,838][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:18:16,852][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 08:18:26,345][__main__][INFO] - Number of regex retries in iteration 719: 1 +[2026-03-26 08:18:26,346][__main__][INFO] - agents played in iteration 719 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:18:27,118][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:18:27,142][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:18:27,165][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:18:27,189][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:18:27,189][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:18:27,190][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:18:28,036][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:18:28,489][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:18:28,997][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:18:29,497][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:18:29,999][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:18:30,504][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:18:31,006][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:18:31,512][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:18:32,014][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:18:32,515][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:18:33,020][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:18:33,521][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:18:34,023][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:18:34,530][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:18:35,029][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:18:35,529][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:18:36,028][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:18:36,528][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:18:37,027][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:18:37,525][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:18:38,026][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:18:38,525][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:18:39,024][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:18:39,523][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:18:40,024][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:18:40,528][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:18:41,031][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:18:41,533][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:18:42,046][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:18:42,548][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:18:43,048][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:18:43,550][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:18:44,049][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:18:44,572][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:18:45,077][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:18:45,581][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:18:46,084][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:18:46,584][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:18:47,082][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:18:47,580][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:18:48,080][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:18:48,579][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:18:49,079][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:18:49,579][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:18:50,082][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:18:50,581][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:18:51,081][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:18:51,580][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:18:52,082][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:18:52,582][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:18:53,081][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:18:53,583][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:18:54,083][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:18:54,584][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:18:55,085][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:18:55,585][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:18:56,087][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:18:56,584][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:18:57,084][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:18:57,584][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:18:58,086][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:18:58,585][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:18:59,083][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:18:59,582][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:19:00,081][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10112 tokens. +[2026-03-26 08:19:00,974][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.57%, Block Peak % of device VRAM: 62.18%, ΔTime: 00:00:32 +[2026-03-26 08:19:01,730][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:19:01,733][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:19:01,734][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:19:02,469][__main__][INFO] - Iteration 720 took 52s (31.36% Gen, 67.24% Train). Generation: 16s, Training: 35s. Estimated remaining time: 32h 38m 14s. Estimated total time: 43h 51m 36s. Time estimates for 10 more iterations: 8m 46s, 100 more iterations: 1h 27m 43s, 500 more iterations: 7h 18m 36s. +[2026-03-26 08:19:02,472][__main__][INFO] - Starting iteration 720. +[2026-03-26 08:19:02,870][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 71 and human policies 1. +[2026-03-26 08:19:02,871][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:19:11,752][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 08:19:20,861][__main__][INFO] - Number of regex retries in iteration 720: 1 +[2026-03-26 08:19:20,862][__main__][INFO] - agents played in iteration 720 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:19:21,622][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:19:21,646][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:19:21,670][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:19:21,693][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:19:21,694][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:19:21,694][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:19:30,913][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:19:31,381][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:19:31,893][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:19:32,395][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:19:32,896][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:19:33,398][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:19:33,900][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:19:34,401][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:19:34,902][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:19:35,403][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:19:35,902][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:19:36,408][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:19:36,917][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:19:37,422][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:19:37,923][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:19:38,424][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:19:38,925][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:19:39,429][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:19:39,927][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:19:40,429][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:19:40,937][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:19:41,441][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:19:41,943][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:19:42,443][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:19:42,942][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:19:43,443][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:19:43,944][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:19:44,444][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:19:44,946][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:19:45,446][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:19:45,950][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:19:46,449][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:19:46,951][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:19:47,452][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:19:47,963][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:19:48,464][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:19:48,966][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:19:49,467][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:19:49,968][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:19:50,467][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:19:50,967][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:19:51,466][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:19:51,965][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:19:52,465][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:19:52,964][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:19:53,464][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:19:53,963][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:19:54,462][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:19:54,962][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:19:55,467][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:19:55,974][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:19:56,482][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:19:56,987][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:19:57,489][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:19:57,991][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:19:58,492][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:19:58,993][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:19:59,492][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:19:59,991][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:20:00,491][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:20:00,991][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:20:01,492][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:20:01,994][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:20:02,493][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:20:02,993][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9969 tokens. +[2026-03-26 08:20:03,890][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 60.49%, Block Peak % of device VRAM: 62.03%, ΔTime: 00:00:33 +[2026-03-26 08:20:04,642][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:20:04,644][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:20:04,646][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:20:06,030][__main__][INFO] - Iteration 721 took 1m 3s (28.48% Gen, 69.32% Train). Generation: 17s, Training: 43s. Estimated remaining time: 41h 23m 34s. Estimated total time: 52h 38m 0s. Time estimates for 10 more iterations: 10m 31s, 100 more iterations: 1h 45m 16s, 500 more iterations: 8h 46m 20s. +[2026-03-26 08:20:06,033][__main__][INFO] - Starting iteration 721. +[2026-03-26 08:20:06,441][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 72 and human policies 1. +[2026-03-26 08:20:06,442][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:20:40,859][__main__][INFO] - Number of regex retries in iteration 721: 0 +[2026-03-26 08:20:40,860][__main__][INFO] - agents played in iteration 721 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:20:41,633][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:20:41,657][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:20:41,681][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:20:41,705][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:20:41,706][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:20:41,706][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:20:42,571][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:20:43,031][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:20:43,537][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:20:44,040][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:20:44,540][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:20:45,047][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:20:45,549][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:20:46,049][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:20:46,550][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:20:47,058][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:20:47,564][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:20:48,071][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:20:48,573][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:20:49,075][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:20:49,579][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:20:50,081][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:20:50,582][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:20:51,083][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:20:51,586][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:20:52,086][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:20:52,587][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:20:53,086][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:20:53,585][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:20:54,086][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:20:54,585][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:20:55,085][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:20:55,584][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:20:56,086][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:20:56,585][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:20:57,085][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:20:57,584][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:20:58,086][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:20:58,591][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:20:59,092][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:20:59,595][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:21:00,095][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:21:00,600][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:21:01,103][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:21:01,602][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:21:02,104][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:21:02,604][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:21:03,104][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:21:03,604][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:21:04,107][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:21:04,612][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:21:05,115][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:21:05,621][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:21:06,166][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:21:06,687][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:21:07,191][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:21:07,689][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:21:08,191][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:21:08,689][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:21:09,188][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:21:09,686][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:21:10,185][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:21:10,683][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:21:11,183][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:21:11,685][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:21:12,192][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:21:12,695][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:21:13,205][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:21:13,705][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:21:14,209][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:21:14,710][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10163 tokens. +[2026-03-26 08:21:15,642][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 60.54%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:33 +[2026-03-26 08:21:16,434][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:21:16,436][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:21:16,438][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:21:17,119][__main__][INFO] - Iteration 722 took 1m 10s (48.70% Gen, 50.34% Train). Generation: 34s, Training: 35s. Estimated remaining time: 47h 38m 19s. Estimated total time: 58h 53m 56s. Time estimates for 10 more iterations: 11m 46s, 100 more iterations: 1h 57m 47s, 500 more iterations: 9h 48m 59s. +[2026-03-26 08:21:17,123][__main__][INFO] - Starting iteration 722. +[2026-03-26 08:21:17,532][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 72 and human policies 1. +[2026-03-26 08:21:17,533][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:21:41,202][__main__][INFO] - Number of regex retries in iteration 722: 0 +[2026-03-26 08:21:41,203][__main__][INFO] - agents played in iteration 722 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:21:41,970][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:21:41,994][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:21:42,018][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:21:42,043][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:21:42,043][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:21:42,044][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:21:42,906][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:21:43,364][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:21:43,887][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:21:44,390][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:21:44,892][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:21:45,392][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:21:45,897][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:21:46,408][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:21:46,916][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:21:47,419][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:21:47,924][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:21:48,425][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:21:48,928][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:21:49,434][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:21:49,942][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:21:50,452][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:21:50,957][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:21:51,459][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:21:51,962][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:21:52,463][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:21:52,968][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:21:53,475][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:21:53,984][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:21:54,489][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:21:54,991][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:21:55,489][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:21:55,988][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:21:56,487][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:21:56,986][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:21:57,485][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:21:57,993][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:21:58,497][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:21:58,998][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:21:59,498][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:21:59,998][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:22:00,497][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:22:00,997][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:22:01,500][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:22:02,002][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:22:02,503][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:22:03,003][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:22:03,503][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:22:04,005][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:22:04,505][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:22:05,004][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:22:05,502][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:22:06,003][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:22:06,508][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:22:07,016][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:22:07,522][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:22:08,027][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:22:08,531][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:22:09,031][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:22:09,530][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:22:10,030][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:22:10,534][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:22:11,040][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:22:11,542][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:22:12,040][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:22:12,543][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:22:13,042][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:22:13,541][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:22:14,042][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:22:14,541][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:22:15,043][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10150 tokens. +[2026-03-26 08:22:15,949][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 60.61%, Block Peak % of device VRAM: 62.18%, ΔTime: 00:00:33 +[2026-03-26 08:22:16,707][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:22:16,709][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:22:16,710][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:22:17,603][__main__][INFO] - Iteration 723 took 1m 0s (39.40% Gen, 59.11% Train). Generation: 23s, Training: 35s. Estimated remaining time: 38h 46m 57s. Estimated total time: 50h 3m 34s. Time estimates for 10 more iterations: 10m 0s, 100 more iterations: 1h 40m 7s, 500 more iterations: 8h 20m 35s. +[2026-03-26 08:22:17,605][__main__][INFO] - Starting iteration 723. +[2026-03-26 08:22:18,008][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 72 and human policies 1. +[2026-03-26 08:22:18,008][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:22:49,734][__main__][INFO] - Number of regex retries in iteration 723: 0 +[2026-03-26 08:22:49,735][__main__][INFO] - agents played in iteration 723 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:22:50,504][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:22:50,528][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:22:50,552][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:22:50,575][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:22:50,576][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:22:50,576][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:22:51,437][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:22:51,902][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:22:52,412][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:22:52,914][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:22:53,415][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:22:53,921][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:22:54,426][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:22:54,938][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:22:55,441][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:22:55,943][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:22:56,443][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:22:56,945][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:22:57,444][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:22:57,943][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:22:58,441][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:22:58,941][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:22:59,439][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:22:59,936][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:23:00,436][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:23:00,934][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:23:01,430][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:23:01,931][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:23:02,436][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:23:02,949][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:23:03,449][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:23:03,952][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:23:04,457][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:23:04,960][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:23:05,463][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:23:05,966][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:23:06,469][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:23:06,978][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:23:07,481][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:23:07,980][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:23:08,480][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:23:08,987][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:23:09,492][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:23:10,002][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:23:10,503][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:23:11,004][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:23:11,506][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:23:12,005][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:23:12,504][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:23:13,003][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:23:13,503][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:23:14,003][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:23:14,509][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:23:15,016][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:23:15,516][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:23:16,017][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:23:16,518][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:23:17,021][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:23:17,522][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:23:18,026][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:23:18,530][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:23:19,030][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:23:19,532][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:23:20,029][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:23:20,526][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:23:21,024][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:23:21,525][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:23:22,023][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:23:22,524][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:23:23,031][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:23:23,533][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10069 tokens. +[2026-03-26 08:23:24,435][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.01%, Current % of VRAM taken: 60.44%, Block Peak % of device VRAM: 62.43%, ΔTime: 00:00:33 +[2026-03-26 08:23:25,196][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:23:25,198][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:23:25,200][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:23:26,069][__main__][INFO] - Iteration 724 took 1m 8s (46.61% Gen, 52.11% Train). Generation: 31s, Training: 35s. Estimated remaining time: 45h 25m 21s. Estimated total time: 56h 43m 7s. Time estimates for 10 more iterations: 11m 20s, 100 more iterations: 1h 53m 26s, 500 more iterations: 9h 27m 11s. +[2026-03-26 08:23:26,072][__main__][INFO] - Starting iteration 724. +[2026-03-26 08:23:26,477][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 72 and human policies 1. +[2026-03-26 08:23:26,477][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:23:53,164][__main__][INFO] - Number of regex retries in iteration 724: 0 +[2026-03-26 08:23:53,165][__main__][INFO] - agents played in iteration 724 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:23:53,928][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:23:53,952][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:23:53,975][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:23:53,999][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.46%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:23:53,999][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:23:54,000][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:23:54,858][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:23:55,318][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:23:55,825][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:23:56,335][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:23:56,835][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:23:57,343][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:23:57,848][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:23:58,351][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:23:58,853][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:23:59,360][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:23:59,864][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:24:00,369][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:24:00,876][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:24:01,380][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:24:01,885][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:24:02,388][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:24:02,890][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:24:03,388][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:24:03,890][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:24:04,392][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:24:04,892][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:24:05,393][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:24:05,894][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:24:06,402][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:24:06,903][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:24:07,405][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:24:07,914][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:24:08,413][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:24:08,913][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:24:09,412][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:24:09,912][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:24:10,413][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:24:10,913][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:24:11,414][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:24:11,918][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:24:12,420][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:24:12,922][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:24:13,423][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:24:13,924][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:24:14,425][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:24:14,925][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:24:15,425][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:24:15,923][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:24:16,421][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:24:16,918][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:24:17,415][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:24:17,916][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:24:18,412][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:24:18,909][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:24:19,410][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:24:19,909][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:24:20,409][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:24:20,909][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:24:21,410][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:24:21,909][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:24:22,409][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:24:22,908][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:24:23,407][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:24:23,909][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:24:24,408][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:24:24,908][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:24:25,409][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:24:25,908][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:24:26,408][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:24:26,907][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10226 tokens. +[2026-03-26 08:24:27,810][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 60.60%, Block Peak % of device VRAM: 62.42%, ΔTime: 00:00:32 +[2026-03-26 08:24:28,560][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:24:28,562][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:24:28,564][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:24:29,294][__main__][INFO] - Iteration 725 took 1m 2s (42.48% Gen, 56.35% Train). Generation: 26s, Training: 35s. Estimated remaining time: 41h 2m 5s. Estimated total time: 52h 20m 54s. Time estimates for 10 more iterations: 10m 28s, 100 more iterations: 1h 44m 41s, 500 more iterations: 8h 43m 29s. +[2026-03-26 08:24:29,296][__main__][INFO] - Starting iteration 725. +[2026-03-26 08:24:29,710][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 72 and human policies 1. +[2026-03-26 08:24:29,711][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:24:51,625][__main__][INFO] - Number of regex retries in iteration 725: 0 +[2026-03-26 08:24:51,625][__main__][INFO] - agents played in iteration 725 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:24:52,408][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:24:52,432][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:24:52,456][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:24:52,479][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:24:52,480][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:24:52,480][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:24:53,328][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:24:53,791][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:24:54,303][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:24:54,818][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:24:55,320][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:24:55,823][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:24:56,324][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:24:56,824][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:24:57,326][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:24:57,831][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:24:58,343][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:24:58,846][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:24:59,346][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:24:59,847][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:25:00,346][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:25:00,850][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:25:01,350][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:25:01,858][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:25:02,372][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:25:02,880][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:25:03,385][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:25:03,888][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:25:04,390][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:25:04,891][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:25:05,392][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:25:05,893][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:25:06,399][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:25:06,908][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:25:07,409][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:25:07,910][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:25:08,412][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:25:08,911][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:25:09,411][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:25:09,909][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:25:10,408][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:25:10,909][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:25:11,415][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:25:11,921][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:25:12,424][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:25:12,925][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:25:13,425][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:25:13,926][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:25:14,427][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:25:14,930][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:25:15,434][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:25:15,934][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:25:16,434][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:25:16,934][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:25:17,441][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:25:17,951][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:25:18,453][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:25:18,953][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:25:19,454][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:25:19,954][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:25:20,454][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:25:20,958][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:25:21,458][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:25:21,959][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:25:22,461][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:25:22,963][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:25:23,466][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:25:23,971][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:25:24,477][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:25:24,985][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:25:25,488][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10080 tokens. +[2026-03-26 08:25:26,406][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 60.60%, Block Peak % of device VRAM: 62.16%, ΔTime: 00:00:33 +[2026-03-26 08:25:27,164][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:25:27,167][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:25:27,168][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:25:27,942][__main__][INFO] - Iteration 726 took 58s (37.63% Gen, 61.03% Train). Generation: 21s, Training: 35s. Estimated remaining time: 37h 11m 52s. Estimated total time: 48h 31m 40s. Time estimates for 10 more iterations: 9m 42s, 100 more iterations: 1h 37m 3s, 500 more iterations: 8h 5m 16s. +[2026-03-26 08:25:27,944][__main__][INFO] - Starting iteration 726. +[2026-03-26 08:25:28,352][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 72 and human policies 1. +[2026-03-26 08:25:28,353][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:25:51,267][__main__][INFO] - Number of regex retries in iteration 726: 0 +[2026-03-26 08:25:51,268][__main__][INFO] - agents played in iteration 726 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:25:52,031][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:25:52,055][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:25:52,078][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:25:52,102][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:25:52,102][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:25:52,103][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:25:52,964][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:25:53,418][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:25:53,921][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:25:54,419][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:25:54,921][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:25:55,419][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:25:55,916][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:25:56,416][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:25:56,915][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:25:57,421][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:25:57,923][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:25:58,420][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:25:58,917][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:25:59,424][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:25:59,925][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:26:00,430][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:26:00,936][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:26:01,436][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:26:01,952][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:26:02,457][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:26:02,958][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:26:03,458][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:26:03,958][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:26:04,458][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:26:04,964][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:26:05,467][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:26:05,968][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:26:06,469][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:26:06,969][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:26:07,471][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:26:08,002][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:26:08,504][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:26:09,007][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:26:09,509][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:26:10,009][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:26:10,510][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:26:11,010][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:26:11,512][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:26:12,011][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:26:12,511][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:26:13,012][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:26:13,510][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:26:14,008][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:26:14,509][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:26:15,011][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:26:15,509][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:26:16,008][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:26:16,507][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:26:17,006][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:26:17,509][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:26:18,011][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:26:18,515][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:26:19,019][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:26:19,528][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:26:20,034][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:26:20,545][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:26:21,048][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:26:21,547][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:26:22,048][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:26:22,549][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:26:23,050][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:26:23,550][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:26:24,051][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:26:24,550][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:26:25,050][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10122 tokens. +[2026-03-26 08:26:25,961][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 60.50%, Block Peak % of device VRAM: 62.16%, ΔTime: 00:00:33 +[2026-03-26 08:26:26,708][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:26:26,713][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:26:26,716][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:26:27,503][__main__][INFO] - Iteration 727 took 59s (38.74% Gen, 59.93% Train). Generation: 22s, Training: 35s. Estimated remaining time: 37h 56m 48s. Estimated total time: 49h 17m 35s. Time estimates for 10 more iterations: 9m 51s, 100 more iterations: 1h 38m 35s, 500 more iterations: 8h 12m 55s. +[2026-03-26 08:26:27,505][__main__][INFO] - Starting iteration 727. +[2026-03-26 08:26:27,924][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 72 and human policies 1. +[2026-03-26 08:26:27,925][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:26:39,952][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 08:26:40,092][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 08:26:50,668][__main__][INFO] - Number of regex retries in iteration 727: 2 +[2026-03-26 08:26:50,669][__main__][INFO] - agents played in iteration 727 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:26:51,439][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:26:51,464][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:26:51,490][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:26:51,515][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:26:51,516][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:26:51,516][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:26:52,383][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:26:52,849][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:26:53,353][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:26:53,863][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:26:54,371][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:26:54,872][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:26:55,376][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:26:55,876][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:26:56,377][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:26:56,876][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:26:57,375][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:26:57,873][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:26:58,377][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:26:58,884][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:26:59,389][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:26:59,895][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:27:00,402][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:27:00,903][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:27:01,404][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:27:01,905][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:27:02,405][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:27:02,904][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:27:03,406][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:27:03,906][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:27:04,405][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:27:04,912][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:27:05,419][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:27:05,929][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:27:06,434][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:27:06,938][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:27:07,439][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:27:07,942][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:27:08,443][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:27:08,944][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:27:09,446][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:27:09,946][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:27:10,446][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:27:10,945][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:27:11,448][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:27:11,952][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:27:12,457][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:27:12,969][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:27:13,471][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:27:13,975][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:27:14,475][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:27:14,977][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:27:15,476][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:27:15,976][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:27:16,476][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:27:16,979][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:27:17,478][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:27:17,979][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:27:18,483][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:27:18,982][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:27:19,485][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:27:19,985][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:27:20,487][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:27:20,988][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:27:21,487][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:27:21,985][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:27:22,483][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:27:22,981][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:27:23,479][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:27:23,976][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:27:24,473][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10093 tokens. +[2026-03-26 08:27:25,376][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 60.48%, Block Peak % of device VRAM: 62.16%, ΔTime: 00:00:33 +[2026-03-26 08:27:26,123][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:27:26,125][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:27:26,127][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:27:26,959][__main__][INFO] - Iteration 728 took 59s (38.53% Gen, 60.06% Train). Generation: 22s, Training: 35s. Estimated remaining time: 37h 49m 59s. Estimated total time: 49h 11m 45s. Time estimates for 10 more iterations: 9m 50s, 100 more iterations: 1h 38m 23s, 500 more iterations: 8h 11m 57s. +[2026-03-26 08:27:26,961][__main__][INFO] - Starting iteration 728. +[2026-03-26 08:27:27,375][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 72 and human policies 1. +[2026-03-26 08:27:27,376][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:27:54,335][__main__][INFO] - Number of regex retries in iteration 728: 0 +[2026-03-26 08:27:54,336][__main__][INFO] - agents played in iteration 728 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:27:55,102][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:27:55,126][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:27:55,149][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:27:55,173][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:27:55,173][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:27:55,174][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:27:56,021][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:27:56,480][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:27:56,984][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:27:57,487][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:27:57,988][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:27:58,488][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:27:58,989][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:27:59,488][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:27:59,988][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:28:00,486][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:28:00,986][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:28:01,487][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:28:01,989][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:28:02,490][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:28:02,988][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:28:03,488][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:28:03,991][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:28:04,489][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:28:04,986][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:28:05,485][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:28:05,984][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:28:06,485][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:28:06,989][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:28:07,487][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:28:07,990][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:28:08,497][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:28:08,997][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:28:09,506][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:28:10,011][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:28:10,511][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:28:11,016][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:28:11,515][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:28:12,016][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:28:12,516][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:28:13,017][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:28:13,520][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:28:14,019][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:28:14,521][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:28:15,021][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:28:15,521][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:28:16,022][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:28:16,522][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:28:17,022][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:28:17,523][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:28:18,026][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:28:18,526][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:28:19,029][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:28:19,532][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:28:20,031][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:28:20,533][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:28:21,031][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:28:21,533][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:28:22,031][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:28:22,534][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:28:23,033][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:28:23,532][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:28:24,031][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:28:24,527][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:28:25,024][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:28:25,521][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:28:26,019][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:28:26,517][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:28:27,017][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:28:27,514][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:28:28,016][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10154 tokens. +[2026-03-26 08:28:28,915][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.05%, Current % of VRAM taken: 60.48%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:32 +[2026-03-26 08:28:29,668][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:28:29,670][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:28:29,672][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:28:30,478][__main__][INFO] - Iteration 729 took 1m 3s (42.72% Gen, 56.00% Train). Generation: 26s, Training: 35s. Estimated remaining time: 41h 12m 22s. Estimated total time: 52h 35m 12s. Time estimates for 10 more iterations: 10m 31s, 100 more iterations: 1h 45m 10s, 500 more iterations: 8h 45m 52s. +[2026-03-26 08:28:30,480][__main__][INFO] - Starting iteration 729. +[2026-03-26 08:28:30,885][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 72 and human policies 1. +[2026-03-26 08:28:30,885][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:28:34,952][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 08:28:45,118][__main__][INFO] - Number of regex retries in iteration 729: 1 +[2026-03-26 08:28:45,119][__main__][INFO] - agents played in iteration 729 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:28:45,894][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:28:45,922][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:28:45,949][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:28:45,975][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:28:45,976][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:28:45,977][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:28:46,867][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:28:47,326][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:28:47,835][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:28:48,337][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:28:48,841][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:28:49,345][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:28:49,858][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:28:50,361][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:28:50,866][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:28:51,369][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:28:51,872][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:28:52,377][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:28:52,878][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:28:53,379][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:28:53,881][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:28:54,382][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:28:54,884][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:28:55,387][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:28:55,890][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:28:56,392][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:28:56,898][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:28:57,405][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:28:57,911][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:28:58,418][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:28:58,922][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:28:59,423][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:28:59,929][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:29:00,432][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:29:00,932][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:29:01,431][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:29:01,934][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:29:02,434][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:29:02,933][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:29:03,435][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:29:03,939][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:29:04,446][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:29:04,959][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:29:05,460][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:29:05,959][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:29:06,463][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:29:06,967][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:29:07,471][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:29:07,971][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:29:08,471][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:29:08,970][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:29:09,469][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:29:09,968][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:29:10,468][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:29:10,970][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:29:11,473][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:29:11,978][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:29:12,486][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:29:12,989][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:29:13,489][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:29:13,989][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:29:14,488][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:29:14,991][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:29:15,490][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:29:15,990][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:29:16,490][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:29:16,990][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:29:17,491][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:29:17,990][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:29:18,490][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:29:18,990][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10217 tokens. +[2026-03-26 08:29:19,883][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 60.54%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:33 +[2026-03-26 08:29:20,632][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:29:20,635][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:29:20,636][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:29:21,363][__main__][INFO] - Iteration 730 took 50s (28.20% Gen, 70.36% Train). Generation: 14s, Training: 35s. Estimated remaining time: 30h 40m 17s. Estimated total time: 42h 3m 58s. Time estimates for 10 more iterations: 8m 24s, 100 more iterations: 1h 24m 7s, 500 more iterations: 7h 0m 39s. +[2026-03-26 08:29:21,365][__main__][INFO] - Starting iteration 730. +[2026-03-26 08:29:21,765][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 72 and human policies 1. +[2026-03-26 08:29:21,766][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:29:26,157][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 08:29:36,794][__main__][INFO] - Number of regex retries in iteration 730: 1 +[2026-03-26 08:29:36,794][__main__][INFO] - agents played in iteration 730 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:29:37,559][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:29:37,583][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:29:37,606][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:29:37,630][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:29:37,630][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:29:37,631][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:29:38,468][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:29:38,923][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:29:39,427][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:29:39,925][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:29:40,426][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:29:40,924][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:29:41,426][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:29:41,925][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:29:42,425][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:29:42,923][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:29:43,421][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:29:43,920][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:29:44,420][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:29:44,919][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:29:45,418][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:29:45,922][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:29:46,420][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:29:46,919][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:29:47,420][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:29:47,922][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:29:48,429][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:29:48,928][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:29:49,431][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:29:49,940][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:29:50,444][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:29:50,945][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:29:51,445][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:29:51,947][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:29:52,449][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:29:52,950][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:29:53,450][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:29:53,951][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:29:54,451][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:29:54,950][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:29:55,449][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:29:55,948][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:29:56,447][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:29:56,946][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:29:57,444][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:29:57,943][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:29:58,442][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:29:58,942][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:29:59,441][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:29:59,940][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:30:00,447][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:30:00,949][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:30:01,451][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:30:01,949][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:30:02,447][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:30:02,945][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:30:03,445][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:30:03,941][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:30:04,441][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:30:04,939][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:30:05,435][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:30:05,932][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:30:06,431][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:30:06,928][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:30:07,425][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:30:07,923][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:30:08,421][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:30:08,921][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:30:09,418][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:30:09,915][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:30:10,412][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10178 tokens. +[2026-03-26 08:30:11,304][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.06%, ΔTime: 00:00:32 +[2026-03-26 08:30:12,056][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:30:12,059][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:30:12,060][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:30:13,393][__main__][INFO] - Iteration 731 took 51s (29.11% Gen, 68.31% Train). Generation: 15s, Training: 35s. Estimated remaining time: 31h 36m 52s. Estimated total time: 43h 1m 25s. Time estimates for 10 more iterations: 8m 36s, 100 more iterations: 1h 26m 2s, 500 more iterations: 7h 10m 14s. +[2026-03-26 08:30:13,395][__main__][INFO] - Starting iteration 731. +[2026-03-26 08:30:13,809][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 73 and human policies 1. +[2026-03-26 08:30:13,810][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:30:28,501][__main__][INFO] - Number of regex retries in iteration 731: 0 +[2026-03-26 08:30:28,502][__main__][INFO] - agents played in iteration 731 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:30:29,265][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:30:29,289][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:30:29,312][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:30:29,336][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:30:29,336][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:30:29,337][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:30:30,185][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:30:30,643][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:30:31,156][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:30:31,674][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:30:32,184][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:30:32,687][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:30:33,192][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:30:33,692][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:30:34,198][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:30:34,712][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:30:35,214][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:30:35,716][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:30:36,218][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:30:36,719][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:30:37,220][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:30:37,724][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:30:38,236][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:30:38,736][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:30:39,237][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:30:39,739][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:30:40,238][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:30:40,737][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:30:41,235][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:30:41,739][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:30:42,244][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:30:42,755][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:30:43,257][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:30:43,758][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:30:44,259][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:30:44,763][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:30:45,263][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:30:45,763][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:30:46,267][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:30:46,771][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:30:47,281][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:30:47,783][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:30:48,284][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:30:48,785][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:30:49,285][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:30:49,787][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:30:50,288][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:30:50,790][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:30:51,292][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:30:51,796][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:30:52,304][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:30:52,810][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:30:53,318][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:30:53,820][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:30:54,325][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:30:54,822][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:30:55,325][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:30:55,824][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:30:56,322][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:30:56,821][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:30:57,319][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:30:57,816][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:30:58,316][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:30:58,825][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:30:59,330][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:30:59,835][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:31:00,344][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:31:00,846][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:31:01,350][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:31:01,855][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:31:02,361][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10125 tokens. +[2026-03-26 08:31:03,258][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.26%, Current % of VRAM taken: 60.68%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:33 +[2026-03-26 08:31:04,080][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:31:04,083][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:31:04,084][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:31:04,966][__main__][INFO] - Iteration 732 took 51s (28.72% Gen, 69.55% Train). Generation: 14s, Training: 35s. Estimated remaining time: 31h 12m 27s. Estimated total time: 42h 37m 52s. Time estimates for 10 more iterations: 8m 31s, 100 more iterations: 1h 25m 15s, 500 more iterations: 7h 6m 18s. +[2026-03-26 08:31:04,968][__main__][INFO] - Starting iteration 732. +[2026-03-26 08:31:05,374][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 73 and human policies 1. +[2026-03-26 08:31:05,375][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:31:19,462][__main__][INFO] - Number of regex retries in iteration 732: 0 +[2026-03-26 08:31:19,463][__main__][INFO] - agents played in iteration 732 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:31:20,225][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:31:20,248][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:31:20,272][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:31:20,295][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:31:20,296][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:31:20,296][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:31:21,143][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:31:21,602][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:31:22,111][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:31:22,610][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:31:23,107][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:31:23,605][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:31:24,107][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:31:24,612][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:31:25,120][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:31:25,634][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:31:26,137][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:31:26,642][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:31:27,145][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:31:27,648][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:31:28,150][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:31:28,651][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:31:29,153][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:31:29,658][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:31:30,162][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:31:30,667][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:31:31,168][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:31:31,671][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:31:32,173][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:31:32,673][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:31:33,171][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:31:33,678][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:31:34,183][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:31:34,683][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:31:35,193][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:31:35,698][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:31:36,212][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:31:36,713][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:31:37,214][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:31:37,715][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:31:38,215][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:31:38,715][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:31:39,214][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:31:39,714][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:31:40,215][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:31:40,717][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:31:41,216][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:31:41,745][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:31:42,251][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:31:42,751][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:31:43,256][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:31:43,760][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:31:44,265][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:31:44,772][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:31:45,278][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:31:45,785][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:31:46,288][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:31:46,788][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:31:47,288][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:31:47,788][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:31:48,289][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:31:48,791][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:31:49,293][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:31:49,796][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:31:50,295][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:31:50,798][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:31:51,297][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:31:51,800][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:31:52,298][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:31:52,801][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:31:53,304][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10202 tokens. +[2026-03-26 08:31:54,196][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.23%, Current % of VRAM taken: 60.66%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:33 +[2026-03-26 08:31:54,953][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:31:54,956][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:31:54,957][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:31:55,691][__main__][INFO] - Iteration 733 took 50s (28.00% Gen, 70.54% Train). Generation: 14s, Training: 35s. Estimated remaining time: 30h 29m 36s. Estimated total time: 41h 55m 51s. Time estimates for 10 more iterations: 8m 23s, 100 more iterations: 1h 23m 51s, 500 more iterations: 6h 59m 18s. +[2026-03-26 08:31:55,696][__main__][INFO] - Starting iteration 733. +[2026-03-26 08:31:56,106][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 73 and human policies 1. +[2026-03-26 08:31:56,106][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:32:13,353][__main__][INFO] - Number of regex retries in iteration 733: 0 +[2026-03-26 08:32:13,354][__main__][INFO] - agents played in iteration 733 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:32:14,120][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:32:14,144][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:32:14,167][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:32:14,191][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:32:14,192][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:32:14,192][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:32:15,052][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:32:15,508][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:32:16,016][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:32:16,513][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:32:17,011][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:32:17,509][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:32:18,007][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:32:18,506][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:32:19,004][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:32:19,501][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:32:19,998][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:32:20,500][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:32:20,999][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:32:21,497][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:32:22,001][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:32:22,499][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:32:23,000][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:32:23,502][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:32:24,004][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:32:24,509][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:32:25,018][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:32:25,522][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:32:26,031][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:32:26,534][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:32:27,037][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:32:27,538][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:32:28,039][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:32:28,544][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:32:29,043][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:32:29,544][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:32:30,047][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:32:30,547][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:32:31,047][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:32:31,546][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:32:32,048][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:32:32,548][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:32:33,047][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:32:33,547][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:32:34,049][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:32:34,550][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:32:35,050][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:32:35,552][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:32:36,052][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:32:36,555][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:32:37,066][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:32:37,568][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:32:38,069][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:32:38,571][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:32:39,072][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:32:39,572][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:32:40,074][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:32:40,574][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:32:41,076][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:32:41,576][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:32:42,078][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:32:42,586][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:32:43,088][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:32:43,592][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:32:44,090][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:32:44,593][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:32:45,091][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:32:45,594][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:32:46,094][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:32:46,593][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:32:47,093][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10110 tokens. +[2026-03-26 08:32:47,990][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 60.50%, Block Peak % of device VRAM: 62.10%, ΔTime: 00:00:32 +[2026-03-26 08:32:48,741][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:32:48,743][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:32:48,745][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:32:49,585][__main__][INFO] - Iteration 734 took 53s (32.25% Gen, 66.17% Train). Generation: 17s, Training: 35s. Estimated remaining time: 33h 6m 52s. Estimated total time: 44h 34m 1s. Time estimates for 10 more iterations: 8m 54s, 100 more iterations: 1h 29m 8s, 500 more iterations: 7h 25m 40s. +[2026-03-26 08:32:49,588][__main__][INFO] - Starting iteration 734. +[2026-03-26 08:32:49,998][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 73 and human policies 1. +[2026-03-26 08:32:49,999][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:33:07,872][__main__][INFO] - Number of regex retries in iteration 734: 0 +[2026-03-26 08:33:07,873][__main__][INFO] - agents played in iteration 734 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:33:08,637][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:33:08,660][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:33:08,684][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:33:08,707][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:33:08,708][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:33:08,709][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:33:09,558][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:33:10,013][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:33:10,529][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:33:11,043][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:33:11,547][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:33:12,051][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:33:12,554][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:33:13,059][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:33:13,560][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:33:14,061][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:33:14,565][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:33:15,077][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:33:15,583][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:33:16,085][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:33:16,588][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:33:17,090][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:33:17,591][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:33:18,092][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:33:18,592][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:33:19,095][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:33:19,596][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:33:20,095][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:33:20,593][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:33:21,097][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:33:21,595][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:33:22,101][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:33:22,605][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:33:23,109][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:33:23,609][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:33:24,110][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:33:24,611][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:33:25,111][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:33:25,611][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:33:26,114][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:33:26,613][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:33:27,118][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:33:27,624][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:33:28,131][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:33:28,639][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:33:29,145][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:33:29,648][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:33:30,149][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:33:30,649][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:33:31,149][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:33:31,650][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:33:32,153][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:33:32,657][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:33:33,156][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:33:33,656][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:33:34,157][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:33:34,659][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:33:35,159][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:33:35,660][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:33:36,159][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:33:36,659][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:33:37,179][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:33:37,679][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:33:38,178][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:33:38,683][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:33:39,180][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:33:39,678][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:33:40,177][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:33:40,674][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:33:41,173][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:33:41,671][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10129 tokens. +[2026-03-26 08:33:42,565][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 60.54%, Block Peak % of device VRAM: 62.15%, ΔTime: 00:00:33 +[2026-03-26 08:33:43,316][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:33:43,318][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:33:43,320][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:33:44,060][__main__][INFO] - Iteration 735 took 54s (33.06% Gen, 65.57% Train). Generation: 17s, Training: 35s. Estimated remaining time: 33h 35m 2s. Estimated total time: 45h 3m 6s. Time estimates for 10 more iterations: 9m 0s, 100 more iterations: 1h 30m 6s, 500 more iterations: 7h 30m 31s. +[2026-03-26 08:33:44,062][__main__][INFO] - Starting iteration 735. +[2026-03-26 08:33:44,462][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 73 and human policies 1. +[2026-03-26 08:33:44,463][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:34:03,363][__main__][INFO] - Number of regex retries in iteration 735: 0 +[2026-03-26 08:34:03,364][__main__][INFO] - agents played in iteration 735 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:34:04,134][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:34:04,158][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:34:04,182][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:34:04,206][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:34:04,207][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:34:04,207][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:34:05,056][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:34:05,515][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:34:06,017][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:34:06,520][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:34:07,019][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:34:07,521][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:34:08,023][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:34:08,530][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:34:09,030][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:34:09,534][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:34:10,035][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:34:10,544][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:34:11,047][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:34:11,554][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:34:12,057][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:34:12,558][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:34:13,062][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:34:13,564][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:34:14,068][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:34:14,570][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:34:15,073][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:34:15,574][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:34:16,075][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:34:16,577][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:34:17,079][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:34:17,577][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:34:18,076][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:34:18,578][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:34:19,076][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:34:19,580][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:34:20,085][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:34:20,593][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:34:21,095][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:34:21,599][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:34:22,103][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:34:22,606][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:34:23,110][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:34:23,612][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:34:24,139][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:34:24,642][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:34:25,145][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:34:25,645][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:34:26,148][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:34:26,648][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:34:27,148][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:34:27,647][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:34:28,146][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:34:28,647][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:34:29,146][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:34:29,647][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:34:30,150][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:34:30,651][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:34:31,150][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:34:31,649][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:34:32,150][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:34:32,649][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:34:33,149][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:34:33,649][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:34:34,160][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:34:34,665][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:34:35,171][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:34:35,671][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:34:36,172][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:34:36,677][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:34:37,179][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10149 tokens. +[2026-03-26 08:34:38,075][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.57%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:33 +[2026-03-26 08:34:38,832][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:34:38,834][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:34:38,836][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:34:39,572][__main__][INFO] - Iteration 736 took 55s (34.30% Gen, 64.37% Train). Generation: 18s, Training: 35s. Estimated remaining time: 34h 26m 32s. Estimated total time: 45h 55m 31s. Time estimates for 10 more iterations: 9m 11s, 100 more iterations: 1h 31m 51s, 500 more iterations: 7h 39m 15s. +[2026-03-26 08:34:39,583][__main__][INFO] - Starting iteration 736. +[2026-03-26 08:34:39,994][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 73 and human policies 1. +[2026-03-26 08:34:39,994][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:35:03,672][__main__][INFO] - Number of regex retries in iteration 736: 0 +[2026-03-26 08:35:03,673][__main__][INFO] - agents played in iteration 736 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:35:04,441][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:35:04,465][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:35:04,488][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:35:04,512][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:35:04,513][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:35:04,513][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:35:05,357][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:35:05,818][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:35:06,327][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:35:06,827][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:35:07,331][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:35:07,832][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:35:08,332][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:35:08,831][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:35:09,330][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:35:09,831][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:35:10,330][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:35:10,829][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:35:11,330][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:35:11,829][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:35:12,328][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:35:12,828][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:35:13,327][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:35:13,829][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:35:14,333][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:35:14,836][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:35:15,335][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:35:15,833][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:35:16,332][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:35:16,834][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:35:17,336][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:35:17,834][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:35:18,333][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:35:18,832][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:35:19,333][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:35:19,837][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:35:20,335][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:35:20,832][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:35:21,331][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:35:21,830][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:35:22,332][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:35:22,831][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:35:23,336][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:35:23,835][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:35:24,333][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:35:24,836][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:35:25,336][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:35:25,842][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:35:26,343][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:35:26,845][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:35:27,345][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:35:27,846][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:35:28,346][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:35:28,850][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:35:29,351][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:35:29,851][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:35:30,350][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:35:30,853][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:35:31,351][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:35:31,854][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:35:32,354][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:35:32,854][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:35:33,357][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:35:33,856][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:35:34,360][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:35:34,861][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:35:35,362][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:35:35,860][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:35:36,359][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:35:36,859][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:35:37,362][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10083 tokens. +[2026-03-26 08:35:38,264][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 60.63%, Block Peak % of device VRAM: 62.11%, ΔTime: 00:00:32 +[2026-03-26 08:35:39,020][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:35:39,023][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:35:39,024][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:35:39,759][__main__][INFO] - Iteration 737 took 59s (39.62% Gen, 59.15% Train). Generation: 23s, Training: 35s. Estimated remaining time: 38h 18m 17s. Estimated total time: 49h 48m 17s. Time estimates for 10 more iterations: 9m 57s, 100 more iterations: 1h 39m 36s, 500 more iterations: 8h 18m 2s. +[2026-03-26 08:35:39,761][__main__][INFO] - Starting iteration 737. +[2026-03-26 08:35:40,165][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 73 and human policies 1. +[2026-03-26 08:35:40,166][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:35:56,529][__main__][INFO] - Number of regex retries in iteration 737: 0 +[2026-03-26 08:35:56,529][__main__][INFO] - agents played in iteration 737 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:35:57,314][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:35:57,337][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:35:57,361][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:35:57,384][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:35:57,385][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:35:57,385][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:35:58,225][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:35:58,681][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:35:59,194][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:35:59,696][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:36:00,199][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:36:00,698][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:36:01,197][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:36:01,695][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:36:02,197][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:36:02,694][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:36:03,192][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:36:03,695][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:36:04,195][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:36:04,699][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:36:05,202][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:36:05,701][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:36:06,213][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:36:06,718][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:36:07,233][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:36:07,736][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:36:08,241][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:36:08,740][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:36:09,242][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:36:09,741][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:36:10,240][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:36:10,739][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:36:11,238][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:36:11,739][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:36:12,240][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:36:12,739][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:36:13,237][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:36:13,735][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:36:14,233][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:36:14,732][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:36:15,234][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:36:15,732][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:36:16,232][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:36:16,731][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:36:17,232][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:36:17,730][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:36:18,231][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:36:18,732][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:36:19,231][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:36:19,734][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:36:20,234][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:36:20,736][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:36:21,239][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:36:21,738][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:36:22,239][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:36:22,741][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:36:23,238][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:36:23,736][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:36:24,234][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:36:24,735][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:36:25,232][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:36:25,729][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:36:26,229][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:36:26,729][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:36:27,229][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:36:27,731][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:36:28,230][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:36:28,730][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:36:29,234][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:36:29,735][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:36:30,237][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10212 tokens. +[2026-03-26 08:36:31,142][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 60.53%, Block Peak % of device VRAM: 62.18%, ΔTime: 00:00:32 +[2026-03-26 08:36:31,896][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:36:31,899][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:36:31,900][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:36:32,637][__main__][INFO] - Iteration 738 took 52s (31.19% Gen, 67.41% Train). Generation: 16s, Training: 35s. Estimated remaining time: 32h 12m 44s. Estimated total time: 43h 43m 36s. Time estimates for 10 more iterations: 8m 44s, 100 more iterations: 1h 27m 27s, 500 more iterations: 7h 17m 16s. +[2026-03-26 08:36:32,639][__main__][INFO] - Starting iteration 738. +[2026-03-26 08:36:33,038][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 73 and human policies 1. +[2026-03-26 08:36:33,039][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:36:47,964][__main__][INFO] - Number of regex retries in iteration 738: 0 +[2026-03-26 08:36:47,965][__main__][INFO] - agents played in iteration 738 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:36:48,747][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:36:48,770][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:36:48,794][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:36:48,817][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:36:48,818][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:36:48,818][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:36:49,666][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:36:50,126][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:36:50,633][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:36:51,133][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:36:51,636][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:36:52,139][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:36:52,641][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:36:53,142][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:36:53,643][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:36:54,141][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:36:54,644][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:36:55,142][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:36:55,642][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:36:56,143][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:36:56,644][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:36:57,145][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:36:57,647][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:36:58,145][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:36:58,643][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:36:59,142][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:36:59,640][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:37:00,136][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:37:00,634][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:37:01,132][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:37:01,629][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:37:02,128][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:37:02,625][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:37:03,122][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:37:03,618][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:37:04,115][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:37:04,612][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:37:05,108][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:37:05,605][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:37:06,104][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:37:06,604][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:37:07,102][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:37:07,601][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:37:08,099][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:37:08,597][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:37:09,095][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:37:09,597][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:37:10,095][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:37:10,592][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:37:11,093][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:37:11,594][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:37:12,094][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:37:12,596][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:37:13,094][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:37:13,599][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:37:14,099][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:37:14,598][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:37:15,108][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:37:15,608][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:37:16,106][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:37:16,606][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:37:17,106][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:37:17,606][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:37:18,109][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:37:18,608][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:37:19,109][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:37:19,608][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:37:20,111][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:37:20,609][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:37:21,109][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:37:21,611][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10157 tokens. +[2026-03-26 08:37:22,505][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.21%, Current % of VRAM taken: 60.64%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 +[2026-03-26 08:37:23,272][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:37:23,274][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:37:23,276][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:37:24,016][__main__][INFO] - Iteration 739 took 50s (29.28% Gen, 69.26% Train). Generation: 14s, Training: 35s. Estimated remaining time: 30h 57m 11s. Estimated total time: 42h 28m 55s. Time estimates for 10 more iterations: 8m 29s, 100 more iterations: 1h 24m 57s, 500 more iterations: 7h 4m 49s. +[2026-03-26 08:37:24,018][__main__][INFO] - Starting iteration 739. +[2026-03-26 08:37:24,425][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 73 and human policies 1. +[2026-03-26 08:37:24,426][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:37:44,721][__main__][INFO] - Number of regex retries in iteration 739: 0 +[2026-03-26 08:37:44,722][__main__][INFO] - agents played in iteration 739 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:37:45,489][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:37:45,513][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:37:45,536][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:37:45,560][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.56%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:37:45,560][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:37:45,561][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:37:46,407][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:37:46,865][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:37:47,373][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:37:47,873][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:37:48,373][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:37:48,870][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:37:49,369][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:37:49,872][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:37:50,372][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:37:50,871][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:37:51,370][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:37:51,868][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:37:52,368][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:37:52,869][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:37:53,369][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:37:53,867][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:37:54,366][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:37:54,867][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:37:55,365][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:37:55,867][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:37:56,365][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:37:56,863][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:37:57,362][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:37:57,860][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:37:58,359][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:37:58,861][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:37:59,365][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:37:59,864][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:38:00,368][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:38:00,870][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:38:01,374][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:38:01,881][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:38:02,385][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:38:02,889][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:38:03,392][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:38:03,893][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:38:04,393][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:38:04,893][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:38:05,398][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:38:05,897][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:38:06,399][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:38:06,898][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:38:07,402][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:38:07,901][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:38:08,400][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:38:08,901][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:38:09,402][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:38:09,902][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:38:10,401][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:38:10,902][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:38:11,402][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:38:11,901][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:38:12,404][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:38:12,903][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:38:13,403][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:38:13,903][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:38:14,402][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:38:14,904][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:38:15,406][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:38:15,908][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:38:16,409][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:38:16,912][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:38:17,414][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:38:17,916][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:38:18,419][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10195 tokens. +[2026-03-26 08:38:19,325][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.30%, Current % of VRAM taken: 60.72%, Block Peak % of device VRAM: 62.46%, ΔTime: 00:00:32 +[2026-03-26 08:38:20,071][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:38:20,073][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:38:20,075][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:38:20,807][__main__][INFO] - Iteration 740 took 56s (36.00% Gen, 62.70% Train). Generation: 20s, Training: 35s. Estimated remaining time: 35h 26m 27s. Estimated total time: 46h 59m 7s. Time estimates for 10 more iterations: 9m 23s, 100 more iterations: 1h 33m 58s, 500 more iterations: 7h 49m 51s. +[2026-03-26 08:38:20,809][__main__][INFO] - Starting iteration 740. +[2026-03-26 08:38:21,217][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 73 and human policies 1. +[2026-03-26 08:38:21,217][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:38:34,891][__main__][INFO] - Number of regex retries in iteration 740: 0 +[2026-03-26 08:38:34,892][__main__][INFO] - agents played in iteration 740 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:38:35,655][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:38:35,678][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:38:35,702][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:38:35,725][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.58%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:38:35,726][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:38:35,726][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:38:36,566][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:38:37,024][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:38:37,522][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:38:38,023][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:38:38,530][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:38:39,036][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:38:39,541][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:38:40,047][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:38:40,546][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:38:41,048][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:38:41,556][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:38:42,063][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:38:42,571][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:38:43,074][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:38:43,575][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:38:44,076][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:38:44,577][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:38:45,077][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:38:45,576][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:38:46,076][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:38:46,575][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:38:47,074][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:38:47,573][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:38:48,076][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:38:48,574][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:38:49,079][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:38:49,583][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:38:50,085][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:38:50,588][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:38:51,089][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:38:51,590][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:38:52,092][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:38:52,593][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:38:53,091][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:38:53,589][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:38:54,088][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:38:54,586][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:38:55,083][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:38:55,581][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:38:56,078][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:38:56,576][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:38:57,073][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:38:57,570][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:38:58,068][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:38:58,565][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:38:59,063][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:38:59,560][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:39:00,057][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:39:00,558][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:39:01,065][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:39:01,573][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:39:02,077][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:39:02,580][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:39:03,082][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:39:03,586][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:39:04,085][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:39:04,584][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:39:05,089][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:39:05,591][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:39:06,094][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:39:06,598][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:39:07,101][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:39:07,603][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:39:08,110][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:39:08,617][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10058 tokens. +[2026-03-26 08:39:09,532][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.18%, ΔTime: 00:00:32 +[2026-03-26 08:39:10,293][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:39:10,295][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:39:10,297][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:39:11,629][__main__][INFO] - Iteration 741 took 50s (27.12% Gen, 70.23% Train). Generation: 13s, Training: 35s. Estimated remaining time: 30h 27m 6s. Estimated total time: 42h 0m 38s. Time estimates for 10 more iterations: 8m 24s, 100 more iterations: 1h 24m 1s, 500 more iterations: 7h 0m 6s. +[2026-03-26 08:39:11,631][__main__][INFO] - Starting iteration 741. +[2026-03-26 08:39:12,038][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 74 and human policies 1. +[2026-03-26 08:39:12,038][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:39:16,874][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 08:39:27,669][__main__][INFO] - Number of regex retries in iteration 741: 1 +[2026-03-26 08:39:27,670][__main__][INFO] - agents played in iteration 741 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:39:28,441][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:39:28,465][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:39:28,489][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:39:28,513][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:39:28,514][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:39:28,514][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:39:29,391][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:39:29,856][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:39:30,368][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:39:30,870][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:39:31,371][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:39:31,876][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:39:32,381][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:39:32,882][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:39:33,383][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:39:33,881][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:39:34,385][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:39:34,885][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:39:35,384][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:39:35,885][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:39:36,384][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:39:36,882][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:39:37,385][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:39:37,883][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:39:38,385][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:39:38,889][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:39:39,390][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:39:39,892][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:39:40,398][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:39:40,899][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:39:41,403][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:39:41,905][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:39:42,407][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:39:42,912][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:39:43,412][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:39:43,912][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:39:44,413][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:39:44,913][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:39:45,412][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:39:45,911][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:39:46,410][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:39:46,912][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:39:47,409][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:39:47,910][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:39:48,413][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:39:48,915][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:39:49,418][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:39:49,918][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:39:50,417][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:39:50,916][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:39:51,416][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:39:51,919][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:39:52,418][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:39:52,917][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:39:53,416][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:39:53,915][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:39:54,414][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:39:54,915][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:39:55,416][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:39:55,916][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:39:56,415][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:39:56,914][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:39:57,416][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:39:57,917][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:39:58,417][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:39:58,921][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:39:59,428][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:39:59,930][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:40:00,437][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:40:00,942][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:40:01,443][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10097 tokens. +[2026-03-26 08:40:02,388][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.05%, ΔTime: 00:00:33 +[2026-03-26 08:40:03,145][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:40:03,148][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:40:03,149][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:40:04,042][__main__][INFO] - Iteration 742 took 52s (30.06% Gen, 68.22% Train). Generation: 15s, Training: 35s. Estimated remaining time: 31h 45m 54s. Estimated total time: 43h 20m 17s. Time estimates for 10 more iterations: 8m 40s, 100 more iterations: 1h 26m 40s, 500 more iterations: 7h 13m 22s. +[2026-03-26 08:40:04,044][__main__][INFO] - Starting iteration 742. +[2026-03-26 08:40:04,446][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 74 and human policies 1. +[2026-03-26 08:40:04,447][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:40:23,447][__main__][INFO] - Number of regex retries in iteration 742: 0 +[2026-03-26 08:40:23,448][__main__][INFO] - agents played in iteration 742 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:40:24,216][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:40:24,239][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:40:24,263][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:40:24,286][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:40:24,287][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:40:24,288][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:40:25,138][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:40:25,602][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:40:26,108][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:40:26,608][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:40:27,109][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:40:27,611][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:40:28,111][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:40:28,612][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:40:29,112][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:40:29,612][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:40:30,109][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:40:30,608][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:40:31,106][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:40:31,604][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:40:32,105][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:40:32,607][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:40:33,104][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:40:33,608][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:40:34,107][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:40:34,609][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:40:35,110][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:40:35,615][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:40:36,120][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:40:36,630][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:40:37,134][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:40:37,635][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:40:38,136][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:40:38,641][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:40:39,142][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:40:39,644][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:40:40,144][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:40:40,644][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:40:41,145][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:40:41,646][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:40:42,147][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:40:42,647][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:40:43,148][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:40:43,647][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:40:44,147][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:40:44,648][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:40:45,151][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:40:45,650][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:40:46,147][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:40:46,644][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:40:47,142][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:40:47,639][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:40:48,136][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:40:48,633][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:40:49,130][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:40:49,628][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:40:50,129][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:40:50,628][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:40:51,126][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:40:51,624][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:40:52,123][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:40:52,620][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:40:53,119][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:40:53,619][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:40:54,121][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:40:54,623][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:40:55,121][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:40:55,618][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:40:56,122][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:40:56,621][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:40:57,133][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10104 tokens. +[2026-03-26 08:40:58,027][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 60.52%, Block Peak % of device VRAM: 62.24%, ΔTime: 00:00:32 +[2026-03-26 08:40:58,782][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:40:58,784][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:40:58,786][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:40:59,629][__main__][INFO] - Iteration 743 took 55s (34.43% Gen, 64.04% Train). Generation: 19s, Training: 35s. Estimated remaining time: 34h 23m 50s. Estimated total time: 45h 59m 9s. Time estimates for 10 more iterations: 9m 11s, 100 more iterations: 1h 31m 58s, 500 more iterations: 7h 39m 51s. +[2026-03-26 08:40:59,631][__main__][INFO] - Starting iteration 743. +[2026-03-26 08:41:00,041][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 74 and human policies 1. +[2026-03-26 08:41:00,042][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:41:11,670][__main__][INFO] - Number of regex retries in iteration 743: 0 +[2026-03-26 08:41:11,671][__main__][INFO] - agents played in iteration 743 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:41:12,531][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:41:12,555][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:41:12,578][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:41:12,602][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:41:12,602][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:41:12,603][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:41:13,460][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:41:13,917][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:41:14,423][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:41:14,921][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:41:15,418][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:41:15,916][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:41:16,416][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:41:16,913][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:41:17,410][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:41:17,906][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:41:18,403][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:41:18,901][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:41:19,397][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:41:19,893][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:41:20,389][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:41:20,887][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:41:21,383][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:41:21,880][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:41:22,377][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:41:22,882][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:41:23,380][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:41:23,878][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:41:24,378][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:41:24,876][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:41:25,373][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:41:25,870][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:41:26,369][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:41:26,870][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:41:27,369][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:41:27,865][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:41:28,362][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:41:28,859][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:41:29,355][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:41:29,853][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:41:30,351][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:41:30,849][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:41:31,355][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:41:31,863][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:41:32,398][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:41:32,901][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:41:33,405][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:41:33,906][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:41:34,406][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:41:34,911][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:41:35,411][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:41:35,911][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:41:36,411][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:41:36,918][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:41:37,426][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:41:37,932][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:41:38,433][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:41:38,933][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:41:39,432][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:41:39,932][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:41:40,432][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:41:40,931][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:41:41,430][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:41:41,930][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:41:42,432][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:41:42,933][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:41:43,438][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:41:43,943][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:41:44,454][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:41:44,956][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:41:45,457][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10098 tokens. +[2026-03-26 08:41:46,364][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.17%, ΔTime: 00:00:32 +[2026-03-26 08:41:47,122][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:41:47,124][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:41:47,125][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:41:47,863][__main__][INFO] - Iteration 744 took 47s (24.32% Gen, 74.14% Train). Generation: 11s, Training: 35s. Estimated remaining time: 28h 15m 0s. Estimated total time: 39h 51m 8s. Time estimates for 10 more iterations: 7m 58s, 100 more iterations: 1h 19m 42s, 500 more iterations: 6h 38m 31s. +[2026-03-26 08:41:47,866][__main__][INFO] - Starting iteration 744. +[2026-03-26 08:41:48,267][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 74 and human policies 1. +[2026-03-26 08:41:48,268][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:41:57,043][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 08:42:01,753][__main__][INFO] - Number of regex retries in iteration 744: 1 +[2026-03-26 08:42:01,754][__main__][INFO] - agents played in iteration 744 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:42:02,521][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:42:02,547][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:42:02,573][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:42:02,599][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:42:02,599][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:42:02,600][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:42:03,468][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:42:03,926][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:42:04,433][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:42:04,935][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:42:05,435][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:42:05,937][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:42:06,439][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:42:06,941][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:42:07,441][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:42:07,940][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:42:08,444][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:42:08,946][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:42:09,448][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:42:09,950][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:42:10,449][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:42:10,950][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:42:11,449][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:42:11,950][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:42:12,451][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:42:12,951][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:42:13,451][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:42:13,954][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:42:14,467][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:42:14,969][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:42:15,469][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:42:15,970][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:42:16,470][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:42:16,970][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:42:17,471][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:42:17,970][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:42:18,470][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:42:18,973][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:42:19,471][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:42:19,969][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:42:20,468][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:42:20,967][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:42:21,469][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:42:21,966][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:42:22,465][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:42:22,964][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:42:23,462][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:42:23,960][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:42:24,457][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:42:24,954][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:42:25,451][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:42:25,949][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:42:26,446][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:42:26,947][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:42:27,447][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:42:27,946][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:42:28,445][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:42:28,946][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:42:29,446][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:42:29,947][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:42:30,446][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:42:30,946][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:42:31,445][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:42:31,950][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:42:32,452][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:42:32,953][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:42:33,453][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:42:33,953][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:42:34,453][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:42:34,954][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:42:35,453][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10070 tokens. +[2026-03-26 08:42:36,361][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 60.56%, Block Peak % of device VRAM: 62.04%, ΔTime: 00:00:32 +[2026-03-26 08:42:37,124][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:42:37,126][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:42:37,128][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:42:37,884][__main__][INFO] - Iteration 745 took 49s (27.18% Gen, 71.29% Train). Generation: 13s, Training: 35s. Estimated remaining time: 29h 43m 55s. Estimated total time: 41h 20m 52s. Time estimates for 10 more iterations: 8m 16s, 100 more iterations: 1h 22m 41s, 500 more iterations: 6h 53m 28s. +[2026-03-26 08:42:37,887][__main__][INFO] - Starting iteration 745. +[2026-03-26 08:42:38,292][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 74 and human policies 1. +[2026-03-26 08:42:38,293][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:42:46,477][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 08:42:53,428][__main__][INFO] - Number of regex retries in iteration 745: 1 +[2026-03-26 08:42:53,429][__main__][INFO] - agents played in iteration 745 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:42:54,204][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:42:54,228][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:42:54,251][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:42:54,275][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:42:54,275][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:42:54,276][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:42:55,134][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:42:55,595][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:42:56,116][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:42:56,615][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:42:57,120][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:42:57,623][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:42:58,129][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:42:58,631][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:42:59,137][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:42:59,641][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:43:00,165][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:43:00,667][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:43:01,168][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:43:01,674][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:43:02,175][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:43:02,680][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:43:03,192][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:43:03,695][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:43:04,197][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:43:04,697][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:43:05,197][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:43:05,699][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:43:06,206][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:43:06,715][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:43:07,224][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:43:07,727][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:43:08,228][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:43:08,729][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:43:09,229][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:43:09,730][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:43:10,230][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:43:10,730][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:43:11,229][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:43:11,731][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:43:12,232][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:43:12,733][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:43:13,233][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:43:13,740][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:43:14,245][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:43:14,763][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:43:15,265][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:43:15,767][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:43:16,269][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:43:16,769][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:43:17,269][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:43:17,769][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:43:18,268][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:43:18,767][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:43:19,267][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:43:19,770][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:43:20,271][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:43:20,772][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:43:21,273][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:43:21,774][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:43:22,274][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:43:22,780][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:43:23,288][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:43:23,796][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:43:24,296][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:43:24,797][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:43:25,300][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:43:25,798][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:43:26,298][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:43:26,796][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:43:27,297][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10060 tokens. +[2026-03-26 08:43:28,201][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 60.56%, Block Peak % of device VRAM: 62.21%, ΔTime: 00:00:33 +[2026-03-26 08:43:28,951][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:43:28,954][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:43:28,955][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:43:29,726][__main__][INFO] - Iteration 746 took 51s (29.43% Gen, 69.07% Train). Generation: 15s, Training: 35s. Estimated remaining time: 31h 13m 53s. Estimated total time: 42h 51m 42s. Time estimates for 10 more iterations: 8m 34s, 100 more iterations: 1h 25m 43s, 500 more iterations: 7h 8m 37s. +[2026-03-26 08:43:29,728][__main__][INFO] - Starting iteration 746. +[2026-03-26 08:43:30,137][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 74 and human policies 1. +[2026-03-26 08:43:30,137][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:43:45,128][__main__][INFO] - Number of regex retries in iteration 746: 0 +[2026-03-26 08:43:45,129][__main__][INFO] - agents played in iteration 746 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:43:45,903][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:43:45,927][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:43:45,951][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:43:45,974][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:43:45,975][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:43:45,975][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:43:46,835][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:43:47,290][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:43:47,793][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:43:48,291][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:43:48,789][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:43:49,290][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:43:49,787][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:43:50,285][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:43:50,783][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:43:51,284][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:43:51,786][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:43:52,292][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:43:52,792][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:43:53,296][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:43:53,795][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:43:54,295][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:43:54,794][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:43:55,294][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:43:55,802][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:43:56,300][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:43:56,798][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:43:57,320][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:43:57,818][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:43:58,320][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:43:58,818][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:43:59,317][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:43:59,816][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:44:00,316][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:44:00,819][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:44:01,318][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:44:01,818][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:44:02,320][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:44:02,826][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:44:03,325][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:44:03,825][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:44:04,324][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:44:04,826][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:44:05,327][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:44:05,830][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:44:06,336][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:44:06,841][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:44:07,348][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:44:07,850][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:44:08,351][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:44:08,855][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:44:09,354][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:44:09,856][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:44:10,356][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:44:10,855][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:44:11,355][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:44:11,854][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:44:12,358][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:44:12,856][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:44:13,356][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:44:13,854][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:44:14,354][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:44:14,857][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:44:15,360][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:44:15,862][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:44:16,364][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:44:16,865][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:44:17,365][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:44:17,867][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:44:18,368][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:44:18,869][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10081 tokens. +[2026-03-26 08:44:19,769][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 60.62%, Block Peak % of device VRAM: 62.12%, ΔTime: 00:00:32 +[2026-03-26 08:44:20,517][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:44:20,520][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:44:20,521][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:44:21,260][__main__][INFO] - Iteration 747 took 51s (29.32% Gen, 69.23% Train). Generation: 14s, Training: 35s. Estimated remaining time: 30h 57m 31s. Estimated total time: 42h 36m 12s. Time estimates for 10 more iterations: 8m 31s, 100 more iterations: 1h 25m 12s, 500 more iterations: 7h 6m 2s. +[2026-03-26 08:44:21,262][__main__][INFO] - Starting iteration 747. +[2026-03-26 08:44:21,675][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 74 and human policies 1. +[2026-03-26 08:44:21,676][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:44:37,124][__main__][INFO] - Number of regex retries in iteration 747: 0 +[2026-03-26 08:44:37,125][__main__][INFO] - agents played in iteration 747 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:44:37,893][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:44:37,919][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:44:37,944][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:44:37,969][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:44:37,970][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:44:37,970][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:44:38,830][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:44:39,294][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:44:39,803][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:44:40,304][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:44:40,806][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:44:41,307][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:44:41,816][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:44:42,324][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:44:42,828][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:44:43,330][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:44:43,832][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:44:44,333][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:44:44,836][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:44:45,338][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:44:45,840][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:44:46,341][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:44:46,847][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:44:47,350][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:44:47,851][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:44:48,365][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:44:48,867][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:44:49,371][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:44:49,873][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:44:50,377][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:44:50,878][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:44:51,379][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:44:51,879][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:44:52,381][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:44:52,885][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:44:53,387][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:44:53,886][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:44:54,386][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:44:54,885][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:44:55,384][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:44:55,886][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:44:56,386][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:44:56,886][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:44:57,386][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:44:57,887][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:44:58,386][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:44:58,889][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:44:59,386][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:44:59,884][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:45:00,382][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:45:00,879][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:45:01,377][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:45:01,876][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:45:02,373][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:45:02,871][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:45:03,368][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:45:03,867][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:45:04,365][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:45:04,863][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:45:05,360][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:45:05,858][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:45:06,357][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:45:06,859][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:45:07,355][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:45:07,856][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:45:08,354][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:45:08,852][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:45:09,349][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:45:09,848][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:45:10,345][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:45:10,842][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10077 tokens. +[2026-03-26 08:45:11,736][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.09%, Current % of VRAM taken: 60.52%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:32 +[2026-03-26 08:45:12,503][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:45:12,505][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:45:12,507][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:45:13,357][__main__][INFO] - Iteration 748 took 51s (29.89% Gen, 68.46% Train). Generation: 15s, Training: 35s. Estimated remaining time: 31h 24m 32s. Estimated total time: 43h 4m 5s. Time estimates for 10 more iterations: 8m 36s, 100 more iterations: 1h 26m 8s, 500 more iterations: 7h 10m 40s. +[2026-03-26 08:45:13,359][__main__][INFO] - Starting iteration 748. +[2026-03-26 08:45:13,761][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 74 and human policies 1. +[2026-03-26 08:45:13,761][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:45:18,507][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 08:45:22,949][mllm.models.large_language_model_local][WARNING] - Response Proposal: x hats, y books, 10 balls -Given the values, I see that the balls have the highest value for me (10) and the lowest value for Alice (1). For hats and books, the values are the same (1) for both of us. Therefore, I will maximize my proposals for balls, and since Alice values books the most and hats equally, I will propose keeping 10 books and all 10 hats for myself. did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 04:37:45,243][__main__][INFO] - Number of regex retries in iteration 648: 1 -[2025-11-10 04:37:45,244][__main__][INFO] - agents played in iteration 648 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:37:46,006][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:37:46,032][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:37:46,056][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:37:46,079][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:37:46,080][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:37:46,081][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:37:46,860][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:37:47,312][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:37:47,813][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:37:48,313][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:37:48,809][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:37:49,307][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:37:49,802][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:37:50,298][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:37:50,803][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:37:51,302][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:37:51,800][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:37:52,302][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:37:52,800][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:37:53,300][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:37:53,801][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:37:54,299][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:37:54,808][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:37:55,309][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:37:55,813][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:37:56,318][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:37:56,817][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:37:57,321][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:37:57,834][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:37:58,340][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:37:58,844][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:37:59,347][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:37:59,852][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:38:00,363][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:38:00,867][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:38:01,383][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:38:01,884][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:38:02,386][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:38:02,896][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:38:03,396][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:38:03,904][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:38:04,406][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:38:04,912][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:38:05,416][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:38:05,917][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:38:06,420][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:38:06,924][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:38:07,424][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:38:07,931][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:38:08,434][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:38:08,933][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:38:09,435][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:38:09,934][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:38:10,434][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:38:10,933][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:38:11,434][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:38:11,936][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:38:12,437][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:38:12,949][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:38:13,448][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:38:13,949][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:38:14,451][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:38:14,952][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:38:15,456][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:38:15,971][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:38:16,476][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:38:17,006][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:38:17,509][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:38:18,013][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:38:18,519][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:38:19,025][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10055 tokens. -[2025-11-10 04:38:19,958][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 55.57%, Block Peak % of device VRAM: 62.21%, ΔTime: 00:00:33 -[2025-11-10 04:38:20,723][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:38:20,724][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:38:20,726][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:38:21,687][__main__][INFO] - Iteration 649 took 1m 3s (42.94% Gen, 55.55% Train). Generation: 27s, Training: 35s. Estimated remaining time: 42h 34m 18s. Estimated total time: 53h 13m 43s. Time estimates for 10 more iterations: 10m 38s, 100 more iterations: 1h 46m 27s, 500 more iterations: 8h 52m 17s. -[2025-11-10 04:38:21,689][__main__][INFO] - Starting iteration 649. -[2025-11-10 04:38:22,156][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 64 and human policies 1. -[2025-11-10 04:38:22,157][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:38:49,985][__main__][INFO] - Number of regex retries in iteration 649: 0 -[2025-11-10 04:38:49,987][__main__][INFO] - agents played in iteration 649 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:38:50,771][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:38:50,797][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:38:50,822][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:38:50,845][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:38:50,846][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:38:50,847][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:38:51,629][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:38:52,101][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:38:52,602][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:38:53,098][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:38:53,605][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:38:54,112][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:38:54,616][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:38:55,120][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:38:55,624][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:38:56,125][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:38:56,631][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:38:57,134][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:38:57,637][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:38:58,146][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:38:58,649][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:38:59,152][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:38:59,653][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:39:00,156][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:39:00,657][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:39:01,158][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:39:01,663][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:39:02,172][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:39:02,674][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:39:03,176][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:39:03,681][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:39:04,183][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:39:04,683][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:39:05,208][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:39:05,710][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:39:06,210][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:39:06,714][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:39:07,214][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:39:07,721][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:39:08,223][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:39:08,722][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:39:09,229][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:39:09,730][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:39:10,234][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:39:10,735][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:39:11,238][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:39:11,741][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:39:12,240][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:39:12,739][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:39:13,239][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:39:13,740][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:39:14,240][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:39:14,740][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:39:15,238][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:39:15,737][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:39:16,234][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:39:16,736][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:39:17,237][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:39:17,737][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:39:18,238][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:39:18,738][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:39:19,240][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:39:19,743][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:39:20,244][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:39:20,748][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:39:21,251][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:39:21,754][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:39:22,257][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:39:22,757][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:39:23,256][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:39:23,771][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10092 tokens. -[2025-11-10 04:39:24,754][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 55.67%, Block Peak % of device VRAM: 62.13%, ΔTime: 00:00:33 -[2025-11-10 04:39:25,411][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:39:25,413][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:39:25,415][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:39:26,275][__main__][INFO] - Iteration 650 took 1m 4s (43.40% Gen, 55.25% Train). Generation: 27s, Training: 35s. Estimated remaining time: 42h 45m 27s. Estimated total time: 53h 25m 56s. Time estimates for 10 more iterations: 10m 41s, 100 more iterations: 1h 46m 51s, 500 more iterations: 8h 54m 19s. -[2025-11-10 04:39:26,278][__main__][INFO] - Starting iteration 650. -[2025-11-10 04:39:26,755][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 64 and human policies 1. -[2025-11-10 04:39:26,756][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:39:55,148][__main__][INFO] - Number of regex retries in iteration 650: 0 -[2025-11-10 04:39:55,148][__main__][INFO] - agents played in iteration 650 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:39:55,923][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:39:55,948][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:39:55,974][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:39:55,996][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:39:55,997][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:39:55,999][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:39:56,793][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:39:57,251][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:39:57,755][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:39:58,256][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:39:58,764][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:39:59,268][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:39:59,771][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:40:00,277][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:40:00,774][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:40:01,276][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:40:01,779][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:40:02,277][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:40:02,780][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:40:03,291][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:40:03,797][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:40:04,298][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:40:04,798][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:40:05,300][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:40:05,827][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:40:06,331][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:40:06,837][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:40:07,339][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:40:07,842][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:40:08,345][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:40:08,848][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:40:09,350][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:40:09,852][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:40:10,357][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:40:10,857][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:40:11,355][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:40:11,855][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:40:12,358][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:40:12,856][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:40:13,352][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:40:13,853][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:40:14,354][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:40:14,855][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:40:15,352][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:40:15,848][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:40:16,349][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:40:16,847][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:40:17,346][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:40:17,842][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:40:18,341][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:40:18,839][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:40:19,337][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:40:19,836][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:40:20,336][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:40:20,837][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:40:21,340][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:40:21,843][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:40:22,346][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:40:22,848][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:40:23,351][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:40:23,850][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:40:24,353][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:40:24,853][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:40:25,356][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:40:25,860][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:40:26,360][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:40:26,864][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:40:27,363][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:40:27,866][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:40:28,368][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:40:28,870][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10019 tokens. -[2025-11-10 04:40:29,817][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 55.60%, Block Peak % of device VRAM: 62.33%, ΔTime: 00:00:33 -[2025-11-10 04:40:30,592][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:40:30,594][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:40:30,596][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:40:32,411][__main__][INFO] - Iteration 651 took 1m 5s (43.24% Gen, 53.99% Train). Generation: 28s, Training: 35s. Estimated remaining time: 44h 1m 12s. Estimated total time: 54h 42m 48s. Time estimates for 10 more iterations: 10m 56s, 100 more iterations: 1h 49m 25s, 500 more iterations: 9h 7m 8s. -[2025-11-10 04:40:32,413][__main__][INFO] - Starting iteration 651. -[2025-11-10 04:40:32,908][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 65 and human policies 1. -[2025-11-10 04:40:32,909][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:40:59,553][__main__][INFO] - Number of regex retries in iteration 651: 0 -[2025-11-10 04:40:59,554][__main__][INFO] - agents played in iteration 651 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:41:00,349][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:41:00,376][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:41:00,401][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:41:00,425][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:41:00,426][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:41:00,426][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:41:01,304][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:41:01,764][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:41:02,277][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:41:02,782][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:41:03,288][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:41:03,883][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:41:05,196][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:41:05,698][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:41:06,202][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:41:06,705][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:41:07,210][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:41:07,716][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:41:08,224][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:41:08,726][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:41:09,233][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:41:09,742][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:41:10,251][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:41:10,753][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:41:11,258][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:41:11,761][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:41:12,267][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:41:12,775][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:41:13,278][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:41:13,784][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:41:14,296][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:41:14,810][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:41:15,317][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:41:15,818][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:41:16,318][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:41:16,833][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:41:17,332][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:41:17,835][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:41:18,336][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:41:18,834][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:41:19,338][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:41:19,839][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:41:20,339][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:41:20,846][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:41:21,344][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:41:21,847][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:41:22,351][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:41:22,848][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:41:23,354][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:41:23,855][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:41:24,356][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:41:24,857][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:41:25,357][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:41:25,858][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:41:26,360][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:41:26,861][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:41:27,362][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:41:27,864][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:41:28,366][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:41:28,862][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:41:29,362][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:41:29,862][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:41:30,362][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:41:30,862][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:41:31,370][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:41:31,870][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:41:32,372][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:41:32,876][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:41:33,377][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:41:33,878][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:41:34,377][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10124 tokens. -[2025-11-10 04:41:35,290][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 12.99%, Current % of VRAM taken: 54.48%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:33 -[2025-11-10 04:41:35,948][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:41:35,949][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:41:35,951][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:41:36,824][__main__][INFO] - Iteration 652 took 1m 3s (41.69% Gen, 56.95% Train). Generation: 26s, Training: 36s. Estimated remaining time: 42h 33m 10s. Estimated total time: 53h 15m 50s. Time estimates for 10 more iterations: 10m 39s, 100 more iterations: 1h 46m 31s, 500 more iterations: 8h 52m 38s. -[2025-11-10 04:41:36,827][__main__][INFO] - Starting iteration 652. -[2025-11-10 04:41:37,292][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 65 and human policies 1. -[2025-11-10 04:41:37,293][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:41:53,220][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 04:42:03,040][__main__][INFO] - Number of regex retries in iteration 652: 1 -[2025-11-10 04:42:03,040][__main__][INFO] - agents played in iteration 652 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:42:03,917][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 47.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:42:03,940][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 47.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:42:03,963][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 47.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:42:03,986][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 47.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:42:03,986][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:42:03,987][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:42:04,858][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:42:05,322][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:42:05,826][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:42:06,332][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:42:06,836][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:42:07,342][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:42:07,851][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:42:08,352][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:42:08,852][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:42:09,358][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:42:09,865][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:42:10,392][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:42:10,897][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:42:11,398][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:42:11,908][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:42:12,410][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:42:12,914][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:42:13,419][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:42:13,922][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:42:14,434][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:42:14,936][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:42:15,440][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:42:15,942][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:42:16,443][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:42:16,944][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:42:17,450][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:42:17,953][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:42:18,461][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:42:18,964][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:42:19,471][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:42:19,975][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:42:20,477][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:42:20,980][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:42:21,477][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:42:21,973][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:42:22,469][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:42:22,966][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:42:23,463][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:42:23,959][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:42:24,455][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:42:24,958][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:42:25,455][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:42:25,953][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:42:26,459][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:42:26,958][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:42:27,456][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:42:27,959][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:42:28,456][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:42:28,956][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:42:29,453][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:42:29,954][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:42:30,453][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:42:30,954][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:42:31,452][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:42:31,952][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:42:32,453][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:42:32,956][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:42:33,459][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:42:33,961][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:42:34,461][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:42:34,962][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:42:35,465][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:42:35,968][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:42:36,473][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:42:36,980][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10064 tokens. -[2025-11-10 04:42:37,900][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 55.66%, Block Peak % of device VRAM: 62.16%, ΔTime: 00:00:33 -[2025-11-10 04:42:38,662][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:42:38,664][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:42:38,666][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:42:39,623][__main__][INFO] - Iteration 653 took 1m 2s (41.31% Gen, 57.16% Train). Generation: 25s, Training: 35s. Estimated remaining time: 41h 12m 51s. Estimated total time: 51h 56m 34s. Time estimates for 10 more iterations: 10m 23s, 100 more iterations: 1h 43m 53s, 500 more iterations: 8h 39m 25s. -[2025-11-10 04:42:39,626][__main__][INFO] - Starting iteration 653. -[2025-11-10 04:42:40,095][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 65 and human policies 1. -[2025-11-10 04:42:40,096][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:42:55,349][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 04:43:05,213][__main__][INFO] - Number of regex retries in iteration 653: 1 -[2025-11-10 04:43:05,214][__main__][INFO] - agents played in iteration 653 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:43:05,987][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:43:06,013][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:43:06,038][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:43:06,061][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:43:06,062][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:43:06,062][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:43:06,896][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:43:07,363][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:43:07,867][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:43:08,370][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:43:08,874][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:43:09,374][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:43:09,873][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:43:10,371][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:43:10,879][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:43:11,382][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:43:11,885][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:43:12,389][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:43:12,995][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:43:13,511][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:43:14,017][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:43:14,519][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:43:15,022][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:43:15,535][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:43:16,039][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:43:16,640][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:43:17,138][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:43:17,638][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:43:18,141][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:43:18,641][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:43:19,157][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:43:19,657][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:43:20,157][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:43:20,659][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:43:21,160][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:43:21,668][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:43:22,172][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:43:22,671][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:43:23,173][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:43:23,672][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:43:24,173][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:43:24,674][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:43:25,172][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:43:25,682][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:43:26,181][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:43:26,678][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:43:27,173][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:43:27,672][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:43:28,171][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:43:28,671][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:43:29,174][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:43:29,674][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:43:30,177][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:43:30,681][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:43:31,183][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:43:31,682][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:43:32,179][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:43:32,677][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:43:33,176][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:43:33,681][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:43:34,177][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:43:34,680][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:43:35,185][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:43:35,685][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:43:36,191][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:43:36,691][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:43:37,196][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:43:37,696][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:43:38,200][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:43:38,704][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:43:39,207][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10085 tokens. -[2025-11-10 04:43:40,198][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 55.62%, Block Peak % of device VRAM: 62.33%, ΔTime: 00:00:33 -[2025-11-10 04:43:40,843][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:43:40,845][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:43:40,847][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:43:41,674][__main__][INFO] - Iteration 654 took 1m 1s (40.79% Gen, 57.87% Train). Generation: 25s, Training: 35s. Estimated remaining time: 40h 34m 13s. Estimated total time: 51h 18m 58s. Time estimates for 10 more iterations: 10m 15s, 100 more iterations: 1h 42m 37s, 500 more iterations: 8h 33m 9s. -[2025-11-10 04:43:41,676][__main__][INFO] - Starting iteration 654. -[2025-11-10 04:43:42,163][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 65 and human policies 1. -[2025-11-10 04:43:42,164][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:44:13,634][__main__][INFO] - Number of regex retries in iteration 654: 0 -[2025-11-10 04:44:13,634][__main__][INFO] - agents played in iteration 654 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:44:14,437][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:44:14,462][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:44:14,487][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:44:14,510][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:44:14,510][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:44:14,511][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:44:15,360][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:44:15,821][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:44:16,325][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:44:16,838][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:44:17,341][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:44:17,844][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:44:18,355][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:44:18,860][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:44:19,359][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:44:19,870][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:44:20,378][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:44:20,885][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:44:21,397][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:44:21,902][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:44:22,409][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:44:22,941][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:44:23,445][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:44:23,946][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:44:24,452][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:44:24,963][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:44:25,466][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:44:25,970][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:44:26,476][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:44:26,978][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:44:27,488][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:44:27,996][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:44:28,498][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:44:29,000][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:44:29,502][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:44:30,005][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:44:30,510][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:44:31,014][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:44:31,515][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:44:32,013][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:44:32,517][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:44:33,019][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:44:33,515][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:44:34,018][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:44:34,517][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:44:35,017][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:44:35,518][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:44:36,018][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:44:36,522][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:44:37,024][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:44:37,529][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:44:38,033][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:44:38,535][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:44:39,042][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:44:39,546][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:44:40,046][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:44:40,548][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:44:41,050][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:44:41,555][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:44:42,064][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:44:42,566][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:44:43,086][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:44:43,594][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:44:44,100][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:44:44,611][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:44:45,113][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:44:45,620][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:44:46,122][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:44:46,623][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:44:47,126][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:44:47,629][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10099 tokens. -[2025-11-10 04:44:48,528][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 55.62%, Block Peak % of device VRAM: 62.24%, ΔTime: 00:00:33 -[2025-11-10 04:44:49,281][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:44:49,283][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:44:49,285][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:44:50,240][__main__][INFO] - Iteration 655 took 1m 8s (46.23% Gen, 52.37% Train). Generation: 31s, Training: 35s. Estimated remaining time: 45h 57m 59s. Estimated total time: 56h 43m 53s. Time estimates for 10 more iterations: 11m 20s, 100 more iterations: 1h 53m 27s, 500 more iterations: 9h 27m 18s. -[2025-11-10 04:44:50,242][__main__][INFO] - Starting iteration 655. -[2025-11-10 04:44:50,717][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 65 and human policies 1. -[2025-11-10 04:44:50,718][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:45:03,588][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 04:45:13,064][__main__][INFO] - Number of regex retries in iteration 655: 1 -[2025-11-10 04:45:13,064][__main__][INFO] - agents played in iteration 655 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:45:13,919][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:45:13,942][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:45:13,965][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:45:13,987][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:45:13,988][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:45:13,989][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:45:14,776][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:45:15,230][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:45:15,735][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:45:16,236][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:45:16,740][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:45:17,242][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:45:17,746][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:45:18,243][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:45:18,743][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:45:19,243][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:45:19,744][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:45:20,246][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:45:20,754][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:45:21,262][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:45:21,762][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:45:22,263][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:45:22,762][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:45:23,263][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:45:23,782][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:45:24,285][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:45:24,787][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:45:25,287][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:45:25,788][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:45:26,305][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:45:26,805][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:45:27,305][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:45:27,808][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:45:28,314][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:45:28,821][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:45:29,327][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:45:29,825][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:45:30,327][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:45:30,830][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:45:31,333][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:45:31,834][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:45:32,337][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:45:32,838][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:45:33,342][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:45:33,840][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:45:34,337][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:45:34,836][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:45:35,334][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:45:35,831][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:45:36,329][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:45:36,827][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:45:37,331][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:45:37,833][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:45:38,335][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:45:38,836][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:45:39,577][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:45:41,053][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:45:41,625][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:45:42,131][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:45:42,633][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:45:43,138][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:45:43,645][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:45:44,148][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:45:44,652][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:45:45,156][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:45:45,658][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:45:46,165][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:45:46,669][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:45:47,198][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:45:47,701][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:45:48,204][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9981 tokens. -[2025-11-10 04:45:49,180][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.05%, Current % of VRAM taken: 55.55%, Block Peak % of device VRAM: 62.14%, ΔTime: 00:00:34 -[2025-11-10 04:45:49,858][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:45:49,859][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:45:49,861][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:45:50,695][__main__][INFO] - Iteration 656 took 59s (37.26% Gen, 61.35% Train). Generation: 22s, Training: 36s. Estimated remaining time: 39h 12m 2s. Estimated total time: 49h 58m 55s. Time estimates for 10 more iterations: 9m 59s, 100 more iterations: 1h 39m 57s, 500 more iterations: 8h 19m 49s. -[2025-11-10 04:45:50,698][__main__][INFO] - Starting iteration 656. -[2025-11-10 04:45:51,172][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 65 and human policies 1. -[2025-11-10 04:45:51,173][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:46:12,965][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 04:46:23,470][__main__][INFO] - Number of regex retries in iteration 656: 1 -[2025-11-10 04:46:23,471][__main__][INFO] - agents played in iteration 656 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:46:24,266][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:46:24,293][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:46:24,319][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:46:24,342][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:46:24,342][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:46:24,343][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:46:25,220][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:46:25,677][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:46:26,181][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:46:26,681][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:46:27,181][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:46:27,682][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:46:28,194][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:46:28,694][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:46:29,195][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:46:29,697][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:46:30,198][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:46:30,701][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:46:31,206][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:46:31,711][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:46:32,217][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:46:32,718][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:46:33,220][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:46:33,722][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:46:34,229][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:46:34,737][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:46:35,237][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:46:35,739][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:46:36,240][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:46:36,742][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:46:37,244][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:46:37,749][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:46:38,252][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:46:38,755][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:46:39,260][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:46:39,762][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:46:40,266][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:46:40,771][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:46:41,305][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:46:41,813][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:46:42,318][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:46:42,821][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:46:43,322][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:46:43,826][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:46:44,328][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:46:44,829][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:46:45,327][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:46:45,825][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:46:46,329][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:46:46,830][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:46:47,331][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:46:47,837][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:46:48,337][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:46:48,841][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:46:49,339][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:46:49,840][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:46:50,344][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:46:50,844][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:46:51,344][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:46:51,845][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:46:52,346][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:46:52,851][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:46:53,352][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:46:53,851][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:46:54,355][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:46:54,859][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:46:55,360][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:46:55,863][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:46:56,368][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:46:56,868][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:46:57,371][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10130 tokens. -[2025-11-10 04:46:58,283][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 55.65%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:33 -[2025-11-10 04:46:59,019][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:46:59,020][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:46:59,022][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:46:59,975][__main__][INFO] - Iteration 657 took 1m 8s (46.94% Gen, 51.67% Train). Generation: 32s, Training: 35s. Estimated remaining time: 46h 32m 6s. Estimated total time: 57h 20m 9s. Time estimates for 10 more iterations: 11m 28s, 100 more iterations: 1h 54m 40s, 500 more iterations: 9h 33m 21s. -[2025-11-10 04:46:59,977][__main__][INFO] - Starting iteration 657. -[2025-11-10 04:47:00,460][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 65 and human policies 1. -[2025-11-10 04:47:00,460][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:47:14,871][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 04:47:23,862][__main__][INFO] - Number of regex retries in iteration 657: 1 -[2025-11-10 04:47:23,863][__main__][INFO] - agents played in iteration 657 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:47:24,736][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:47:24,763][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:47:24,789][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:47:24,812][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:47:24,812][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:47:24,813][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:47:25,579][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:47:26,031][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:47:26,531][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:47:27,033][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:47:27,529][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:47:28,026][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:47:28,534][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:47:29,039][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:47:29,540][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:47:30,040][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:47:30,539][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:47:31,042][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:47:31,543][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:47:32,045][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:47:32,553][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:47:33,056][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:47:33,561][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:47:34,064][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:47:34,566][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:47:35,070][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:47:35,571][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:47:36,072][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:47:36,575][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:47:37,077][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:47:37,579][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:47:38,091][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:47:38,592][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:47:39,109][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:47:39,611][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:47:40,123][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:47:41,912][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:47:42,436][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:47:42,940][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:47:43,442][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:47:43,944][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:47:44,446][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:47:44,950][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:47:45,453][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:47:45,954][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:47:46,454][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:47:46,955][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:47:47,456][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:47:47,956][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:47:48,460][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:47:48,962][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:47:49,463][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:47:49,964][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:47:50,466][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:47:50,966][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:47:51,467][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:47:51,965][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:47:52,466][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:47:52,967][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:47:53,469][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:47:53,967][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:47:54,469][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:47:54,966][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:47:55,469][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:47:55,974][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:47:56,476][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:47:56,978][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:47:57,491][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:47:57,994][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:47:58,495][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:47:58,996][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9982 tokens. -[2025-11-10 04:47:59,921][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 55.63%, Block Peak % of device VRAM: 62.10%, ΔTime: 00:00:34 -[2025-11-10 04:48:00,561][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:48:00,563][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:48:00,565][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:48:01,387][__main__][INFO] - Iteration 658 took 1m 0s (38.41% Gen, 60.24% Train). Generation: 23s, Training: 36s. Estimated remaining time: 39h 57m 19s. Estimated total time: 50h 46m 23s. Time estimates for 10 more iterations: 10m 9s, 100 more iterations: 1h 41m 32s, 500 more iterations: 8h 27m 43s. -[2025-11-10 04:48:01,390][__main__][INFO] - Starting iteration 658. -[2025-11-10 04:48:01,853][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 65 and human policies 1. -[2025-11-10 04:48:01,854][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:48:30,048][__main__][INFO] - Number of regex retries in iteration 658: 0 -[2025-11-10 04:48:30,049][__main__][INFO] - agents played in iteration 658 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:48:30,810][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:48:30,833][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:48:30,858][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:48:30,880][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:48:30,881][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:48:30,882][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:48:31,658][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:48:32,112][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:48:32,627][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:48:33,123][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:48:33,620][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:48:34,128][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:48:34,623][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:48:35,126][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:48:35,627][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:48:36,129][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:48:36,626][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:48:37,130][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:48:37,633][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:48:38,145][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:48:38,646][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:48:39,149][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:48:39,650][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:48:40,150][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:48:40,651][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:48:41,152][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:48:41,653][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:48:42,156][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:48:42,658][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:48:43,160][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:48:43,662][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:48:44,164][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:48:44,666][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:48:45,176][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:48:45,677][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:48:46,179][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:48:46,679][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:48:47,179][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:48:47,685][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:48:48,190][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:48:48,696][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:48:49,199][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:48:49,701][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:48:50,212][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:48:50,718][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:48:51,233][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:48:51,735][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:48:52,236][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:48:52,738][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:48:53,238][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:48:53,739][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:48:54,243][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:48:54,746][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:48:55,250][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:48:55,746][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:48:56,243][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:48:56,742][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:48:57,241][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:48:57,747][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:48:58,245][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:48:58,748][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:48:59,257][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:48:59,755][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:49:00,257][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:49:00,759][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:49:01,259][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:49:01,765][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:49:02,266][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:49:02,767][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:49:03,269][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:49:03,771][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10097 tokens. -[2025-11-10 04:49:04,691][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 55.63%, Block Peak % of device VRAM: 62.15%, ΔTime: 00:00:33 -[2025-11-10 04:49:05,432][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:49:05,434][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:49:05,436][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:49:06,557][__main__][INFO] - Iteration 659 took 1m 4s (43.57% Gen, 54.69% Train). Generation: 28s, Training: 35s. Estimated remaining time: 43h 5m 3s. Estimated total time: 53h 55m 12s. Time estimates for 10 more iterations: 10m 47s, 100 more iterations: 1h 47m 50s, 500 more iterations: 8h 59m 12s. -[2025-11-10 04:49:06,559][__main__][INFO] - Starting iteration 659. -[2025-11-10 04:49:07,029][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 65 and human policies 1. -[2025-11-10 04:49:07,032][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:49:32,071][__main__][INFO] - Number of regex retries in iteration 659: 0 -[2025-11-10 04:49:32,072][__main__][INFO] - agents played in iteration 659 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:49:32,834][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:49:32,861][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:49:32,888][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:49:32,911][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:49:32,912][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:49:32,913][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:49:33,708][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:49:34,162][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:49:34,673][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:49:35,168][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:49:35,663][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:49:36,157][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:49:36,652][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:49:37,149][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:49:37,645][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:49:38,148][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:49:38,646][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:49:39,144][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:49:39,641][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:49:40,138][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:49:40,636][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:49:41,154][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:49:41,662][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:49:42,168][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:49:42,672][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:49:43,171][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:49:43,672][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:49:44,173][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:49:44,674][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:49:45,181][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:49:45,685][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:49:46,188][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:49:46,689][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:49:47,192][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:49:47,695][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:49:48,194][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:49:48,694][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:49:49,197][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:49:49,697][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:49:50,198][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:49:50,698][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:49:51,199][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:49:51,700][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:49:52,200][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:49:52,703][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:49:53,208][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:49:53,709][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:49:54,212][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:49:54,714][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:49:55,217][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:49:55,728][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:49:56,435][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:49:57,785][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:49:58,290][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:49:58,794][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:49:59,299][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:49:59,804][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:50:00,310][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:50:00,817][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:50:01,318][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:50:01,821][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:50:02,336][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:50:02,838][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:50:03,359][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:50:03,865][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:50:04,373][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:50:04,879][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:50:05,384][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:50:05,894][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:50:06,402][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:50:06,907][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10069 tokens. -[2025-11-10 04:50:07,851][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 55.72%, Block Peak % of device VRAM: 62.53%, ΔTime: 00:00:34 -[2025-11-10 04:50:08,504][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:50:08,506][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:50:08,508][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:50:09,367][__main__][INFO] - Iteration 660 took 1m 2s (40.17% Gen, 58.45% Train). Generation: 25s, Training: 36s. Estimated remaining time: 41h 5m 41s. Estimated total time: 51h 56m 53s. Time estimates for 10 more iterations: 10m 23s, 100 more iterations: 1h 43m 53s, 500 more iterations: 8h 39m 28s. -[2025-11-10 04:50:09,369][__main__][INFO] - Starting iteration 660. -[2025-11-10 04:50:09,839][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 65 and human policies 1. -[2025-11-10 04:50:09,840][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:50:40,751][__main__][INFO] - Number of regex retries in iteration 660: 0 -[2025-11-10 04:50:40,751][__main__][INFO] - agents played in iteration 660 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:50:41,562][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:50:41,588][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:50:41,614][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:50:41,637][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:50:41,637][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:50:41,638][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:50:42,418][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:50:42,869][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:50:43,387][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:50:43,884][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:50:44,381][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:50:44,882][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:50:45,379][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:50:45,881][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:50:46,382][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:50:46,887][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:50:47,392][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:50:47,891][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:50:48,391][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:50:48,890][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:50:49,390][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:50:49,897][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:50:50,396][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:50:50,895][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:50:51,398][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:50:51,898][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:50:52,399][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:50:52,903][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:50:53,404][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:50:53,913][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:50:54,415][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:50:54,916][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:50:55,417][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:50:55,918][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:50:56,420][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:50:56,920][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:50:57,424][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:50:57,932][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:50:58,433][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:50:58,937][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:50:59,436][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:50:59,939][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:51:00,453][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:51:00,957][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:51:01,460][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:51:01,963][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:51:02,463][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:51:02,971][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:51:03,473][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:51:03,979][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:51:04,494][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:51:04,999][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:51:05,503][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:51:06,009][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:51:06,516][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:51:07,015][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:51:07,518][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:51:08,017][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:51:08,522][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:51:09,025][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:51:09,531][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:51:10,032][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:51:10,537][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:51:11,038][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:51:11,539][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:51:12,039][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:51:12,559][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:51:13,061][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:51:13,564][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:51:14,066][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:51:14,567][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10112 tokens. -[2025-11-10 04:51:15,519][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 55.55%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:33 -[2025-11-10 04:51:16,280][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:51:16,282][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:51:16,284][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:51:18,192][__main__][INFO] - Iteration 661 took 1m 8s (45.22% Gen, 51.98% Train). Generation: 30s, Training: 35s. Estimated remaining time: 46h 5m 21s. Estimated total time: 56h 57m 42s. Time estimates for 10 more iterations: 11m 23s, 100 more iterations: 1h 53m 55s, 500 more iterations: 9h 29m 37s. -[2025-11-10 04:51:18,195][__main__][INFO] - Starting iteration 661. -[2025-11-10 04:51:18,666][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 66 and human policies 1. -[2025-11-10 04:51:18,667][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:51:40,421][__main__][INFO] - Number of regex retries in iteration 661: 0 -[2025-11-10 04:51:40,421][__main__][INFO] - agents played in iteration 661 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:51:41,193][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:51:41,216][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:51:41,238][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:51:41,260][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:51:41,261][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:51:41,261][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:51:42,053][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:51:42,519][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:51:43,018][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:51:43,519][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:51:44,020][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:51:44,517][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:51:45,018][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:51:45,515][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:51:46,013][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:51:46,513][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:51:47,016][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:51:47,515][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:51:48,013][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:51:48,510][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:51:49,010][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:51:49,509][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:51:50,009][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:51:50,506][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:51:51,005][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:51:51,503][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:51:51,998][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:51:52,494][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:51:52,997][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:51:53,497][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:51:53,995][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:51:54,497][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:51:54,998][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:51:55,502][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:51:56,009][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:51:56,509][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:51:57,010][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:51:57,514][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:51:58,018][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:51:58,519][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:51:59,023][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:51:59,526][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:52:00,028][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:52:00,530][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:52:01,033][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:52:01,533][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:52:02,036][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:52:02,541][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:52:03,041][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:52:03,544][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:52:04,050][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:52:04,554][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:52:05,060][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:52:05,565][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:52:06,073][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:52:06,574][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:52:07,077][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:52:07,583][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:52:08,084][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:52:08,595][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:52:09,096][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:52:09,597][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:52:10,105][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:52:12,283][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:52:12,787][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:52:13,291][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:52:13,798][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:52:14,297][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:52:14,797][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:52:15,300][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:52:15,803][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10149 tokens. -[2025-11-10 04:52:16,772][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 55.69%, Block Peak % of device VRAM: 62.45%, ΔTime: 00:00:34 -[2025-11-10 04:52:17,418][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:52:17,419][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:52:17,421][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:52:18,240][__main__][INFO] - Iteration 662 took 59s (36.52% Gen, 62.11% Train). Generation: 21s, Training: 37s. Estimated remaining time: 38h 45m 25s. Estimated total time: 49h 38m 46s. Time estimates for 10 more iterations: 9m 55s, 100 more iterations: 1h 39m 17s, 500 more iterations: 8h 16m 27s. -[2025-11-10 04:52:18,244][__main__][INFO] - Starting iteration 662. -[2025-11-10 04:52:18,716][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 66 and human policies 1. -[2025-11-10 04:52:18,717][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:52:36,526][mllm.models.large_language_model_local][WARNING] - Response Proposal: x hats, y books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 04:52:46,662][__main__][INFO] - Number of regex retries in iteration 662: 1 -[2025-11-10 04:52:46,662][__main__][INFO] - agents played in iteration 662 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:52:47,430][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:52:47,453][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:52:47,476][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:52:47,498][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:52:47,498][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:52:47,499][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:52:48,285][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:52:48,741][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:52:49,246][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:52:49,742][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:52:50,239][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:52:50,736][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:52:51,237][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:52:51,733][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:52:52,231][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:52:52,727][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:52:53,219][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:52:53,716][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:52:54,212][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:52:54,716][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:52:55,211][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:52:55,706][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:52:56,201][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:52:56,695][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:52:57,189][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:52:57,688][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:52:58,180][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:52:58,690][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:52:59,185][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:52:59,678][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:53:00,181][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:53:00,676][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:53:01,176][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:53:01,680][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:53:02,181][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:53:02,696][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:53:03,195][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:53:03,696][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:53:04,193][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:53:04,694][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:53:05,204][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:53:05,703][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:53:06,205][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:53:06,707][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:53:07,210][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:53:07,713][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:53:08,218][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:53:08,717][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:53:09,221][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:53:09,724][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:53:10,224][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:53:10,723][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:53:11,223][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:53:11,727][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:53:12,227][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:53:12,730][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:53:13,235][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:53:13,742][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:53:14,244][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:53:14,749][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:53:15,256][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:53:15,763][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:53:16,264][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:53:16,767][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:53:17,284][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:53:17,787][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:53:18,301][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:53:18,806][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:53:19,310][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:53:19,822][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:53:20,325][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10037 tokens. -[2025-11-10 04:53:21,247][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 55.67%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:32 -[2025-11-10 04:53:22,001][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:53:22,003][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:53:22,005][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:53:22,936][__main__][INFO] - Iteration 663 took 1m 4s (43.51% Gen, 55.03% Train). Generation: 27s, Training: 35s. Estimated remaining time: 42h 36m 36s. Estimated total time: 53h 31m 2s. Time estimates for 10 more iterations: 10m 42s, 100 more iterations: 1h 47m 2s, 500 more iterations: 8h 55m 10s. -[2025-11-10 04:53:22,939][__main__][INFO] - Starting iteration 663. -[2025-11-10 04:53:23,423][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 66 and human policies 1. -[2025-11-10 04:53:23,424][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:53:48,416][__main__][INFO] - Number of regex retries in iteration 663: 0 -[2025-11-10 04:53:48,417][__main__][INFO] - agents played in iteration 663 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:53:49,191][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:53:49,214][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:53:49,237][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:53:49,259][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:53:49,260][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:53:49,261][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:53:50,051][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:53:50,507][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:53:51,004][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:53:51,516][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:53:52,010][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:53:52,503][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:53:53,001][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:53:53,494][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:53:53,993][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:53:54,488][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:53:54,983][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:53:55,498][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:53:55,991][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:53:56,487][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:53:56,980][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:53:57,474][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:53:57,969][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:53:58,469][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:53:58,967][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:53:59,463][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:53:59,961][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:54:00,461][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:54:00,955][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:54:01,449][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:54:01,949][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:54:02,446][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:54:02,950][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:54:03,447][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:54:03,946][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:54:04,445][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:54:04,942][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:54:05,442][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:54:05,948][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:54:06,449][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:54:06,953][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:54:07,454][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:54:07,957][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:54:08,463][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:54:08,963][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:54:09,466][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:54:09,969][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:54:10,470][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:54:10,974][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:54:11,476][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:54:11,979][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:54:12,485][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:54:12,990][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:54:13,497][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:54:14,005][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:54:14,511][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:54:15,030][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:54:15,535][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:54:16,037][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:54:16,539][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:54:17,042][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:54:17,545][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:54:18,046][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:54:18,547][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:54:19,051][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:54:19,552][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:54:20,054][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:54:20,555][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:54:21,055][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:54:21,749][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:54:22,270][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10092 tokens. -[2025-11-10 04:54:23,248][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 55.55%, Block Peak % of device VRAM: 62.33%, ΔTime: 00:00:33 -[2025-11-10 04:54:24,996][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:54:24,998][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:54:25,001][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:54:25,831][__main__][INFO] - Iteration 664 took 1m 2s (40.05% Gen, 58.62% Train). Generation: 24s, Training: 36s. Estimated remaining time: 41h 4m 56s. Estimated total time: 52h 0m 25s. Time estimates for 10 more iterations: 10m 24s, 100 more iterations: 1h 44m 0s, 500 more iterations: 8h 40m 4s. -[2025-11-10 04:54:25,835][__main__][INFO] - Starting iteration 664. -[2025-11-10 04:54:26,312][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 66 and human policies 1. -[2025-11-10 04:54:26,313][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:54:56,467][__main__][INFO] - Number of regex retries in iteration 664: 0 -[2025-11-10 04:54:56,468][__main__][INFO] - agents played in iteration 664 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:54:57,232][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:54:57,261][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:54:57,287][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:54:57,310][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:54:57,310][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:54:57,311][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:54:58,095][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:54:58,551][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:54:59,053][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:54:59,557][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:55:00,055][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:55:00,556][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:55:01,057][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:55:01,553][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:55:02,052][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:55:02,550][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:55:03,049][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:55:03,553][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:55:04,051][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:55:04,552][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:55:05,058][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:55:05,557][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:55:06,058][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:55:06,554][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:55:07,051][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:55:07,557][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:55:08,054][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:55:08,553][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:55:09,050][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:55:09,550][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:55:10,056][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:55:10,557][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:55:11,057][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:55:11,559][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:55:12,059][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:55:12,566][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:55:13,063][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:55:13,563][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:55:14,065][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:55:14,571][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:55:15,077][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:55:15,578][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:55:16,079][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:55:16,582][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:55:17,082][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:55:17,584][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:55:18,090][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:55:18,594][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:55:19,102][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:55:19,610][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:55:20,114][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:55:20,618][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:55:21,120][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:55:21,631][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:55:22,136][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:55:22,639][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:55:23,142][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:55:23,644][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:55:24,158][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:55:24,663][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:55:25,166][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:55:25,674][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:55:26,176][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:55:26,680][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:55:27,190][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:55:27,693][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:55:28,198][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:55:28,701][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:55:29,217][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:55:29,873][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:55:30,379][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10095 tokens. -[2025-11-10 04:55:31,341][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 55.69%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:33 -[2025-11-10 04:55:32,137][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:55:32,139][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:55:32,141][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:55:33,144][__main__][INFO] - Iteration 665 took 1m 6s (45.12% Gen, 53.38% Train). Generation: 30s, Training: 35s. Estimated remaining time: 44h 45m 2s. Estimated total time: 55h 41m 38s. Time estimates for 10 more iterations: 11m 8s, 100 more iterations: 1h 51m 23s, 500 more iterations: 9h 16m 56s. -[2025-11-10 04:55:33,147][__main__][INFO] - Starting iteration 665. -[2025-11-10 04:55:33,613][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 66 and human policies 1. -[2025-11-10 04:55:33,614][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:55:59,682][__main__][INFO] - Number of regex retries in iteration 665: 0 -[2025-11-10 04:55:59,682][__main__][INFO] - agents played in iteration 665 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:56:00,453][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:56:00,481][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:56:00,507][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:56:00,530][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:56:00,531][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:56:00,532][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:56:01,322][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:56:01,777][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:56:02,284][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:56:02,780][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:56:03,276][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:56:03,780][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:56:04,276][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:56:04,775][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:56:05,270][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:56:05,768][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:56:06,275][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:56:06,776][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:56:07,272][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:56:07,775][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:56:08,275][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:56:08,772][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:56:09,273][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:56:09,770][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:56:10,273][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:56:10,770][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:56:11,269][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:56:11,766][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:56:12,262][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:56:12,759][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:56:13,255][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:56:13,752][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:56:14,253][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:56:14,750][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:56:15,257][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:56:15,756][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:56:16,253][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:56:16,767][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:56:17,266][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:56:17,761][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:56:18,259][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:56:18,760][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:56:19,262][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:56:19,762][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:56:20,265][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:56:20,767][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:56:21,267][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:56:21,771][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:56:22,273][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:56:22,780][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:56:23,290][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:56:23,794][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:56:24,298][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:56:24,804][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:56:25,312][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:56:25,814][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:56:26,317][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:56:26,819][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:56:27,332][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:56:27,835][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:56:28,340][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:56:28,841][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:56:29,343][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:56:29,845][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:56:30,348][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:56:30,850][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:56:31,356][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:56:31,858][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:56:32,375][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:56:32,878][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:56:33,382][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10122 tokens. -[2025-11-10 04:56:34,568][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.00%, Current % of VRAM taken: 55.50%, Block Peak % of device VRAM: 62.13%, ΔTime: 00:00:33 -[2025-11-10 04:56:35,796][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:56:35,798][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:56:35,799][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:56:36,834][__main__][INFO] - Iteration 666 took 1m 3s (41.23% Gen, 57.13% Train). Generation: 26s, Training: 36s. Estimated remaining time: 41h 43m 24s. Estimated total time: 52h 41m 4s. Time estimates for 10 more iterations: 10m 32s, 100 more iterations: 1h 45m 22s, 500 more iterations: 8h 46m 50s. -[2025-11-10 04:56:36,839][__main__][INFO] - Starting iteration 666. -[2025-11-10 04:56:37,308][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 66 and human policies 1. -[2025-11-10 04:56:37,309][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:57:06,942][__main__][INFO] - Number of regex retries in iteration 666: 0 -[2025-11-10 04:57:06,943][__main__][INFO] - agents played in iteration 666 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:57:07,718][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:57:07,743][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:57:07,768][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:57:07,790][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:57:07,791][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:57:07,792][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:57:08,576][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:57:09,036][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:57:09,536][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:57:10,034][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:57:10,531][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:57:11,033][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:57:11,540][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:57:12,037][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:57:12,534][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:57:13,040][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:57:13,535][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:57:14,031][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:57:14,529][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:57:15,028][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:57:15,524][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:57:16,018][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:57:16,514][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:57:17,010][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:57:17,505][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:57:18,002][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:57:18,497][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:57:18,992][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:57:19,489][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:57:19,983][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:57:20,477][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:57:20,972][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:57:21,470][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:57:21,980][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:57:22,479][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:57:22,982][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:57:23,480][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:57:23,977][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:57:24,479][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:57:24,979][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:57:25,477][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:57:25,984][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:57:26,484][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:57:26,984][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:57:27,488][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:57:27,990][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:57:28,493][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:57:28,994][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:57:29,498][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:57:30,000][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:57:30,502][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:57:31,004][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:57:31,505][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:57:32,005][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:57:32,513][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:57:33,018][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:57:33,524][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:57:34,030][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:57:34,541][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:57:35,055][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:57:35,560][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:57:36,064][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:57:36,574][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:57:37,075][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:57:37,579][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:57:38,083][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:57:38,588][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:57:39,094][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:57:39,598][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:57:40,102][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:57:40,604][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10006 tokens. -[2025-11-10 04:57:41,555][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.05%, Current % of VRAM taken: 55.54%, Block Peak % of device VRAM: 62.44%, ΔTime: 00:00:32 -[2025-11-10 04:57:42,333][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:57:42,335][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:57:42,337][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:57:43,271][__main__][INFO] - Iteration 667 took 1m 5s (44.92% Gen, 53.66% Train). Generation: 29s, Training: 35s. Estimated remaining time: 43h 59m 26s. Estimated total time: 54h 58m 13s. Time estimates for 10 more iterations: 10m 59s, 100 more iterations: 1h 49m 56s, 500 more iterations: 9h 9m 42s. -[2025-11-10 04:57:43,274][__main__][INFO] - Starting iteration 667. -[2025-11-10 04:57:43,762][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 66 and human policies 1. -[2025-11-10 04:57:43,763][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:58:08,768][__main__][INFO] - Number of regex retries in iteration 667: 0 -[2025-11-10 04:58:08,768][__main__][INFO] - agents played in iteration 667 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:58:09,528][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:58:09,555][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:58:09,581][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:58:09,604][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:58:09,605][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:58:09,606][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:58:10,391][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:58:10,853][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:58:11,357][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:58:11,867][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:58:12,368][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:58:12,868][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:58:13,369][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:58:13,868][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:58:14,366][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:58:14,878][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:58:15,376][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:58:15,879][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:58:16,377][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:58:16,874][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:58:17,371][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:58:17,868][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:58:18,375][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:58:18,876][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:58:19,373][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:58:19,877][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:58:20,375][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:58:20,874][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:58:21,373][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:58:21,871][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:58:22,371][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:58:22,869][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:58:23,369][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:58:23,867][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:58:24,366][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:58:24,883][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:58:25,380][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:58:25,883][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:58:26,392][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:58:26,896][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:58:27,398][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:58:27,899][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:58:28,400][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:58:28,904][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:58:29,407][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:58:29,919][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:58:30,426][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:58:30,931][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:58:31,436][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:58:31,938][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:58:32,440][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:58:32,944][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:58:33,446][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:58:33,952][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:58:34,462][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:58:34,976][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:58:35,480][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:58:35,988][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:58:36,494][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:58:37,005][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:58:37,511][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:58:38,014][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:58:38,520][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:58:39,022][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:58:39,526][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:58:40,033][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:58:40,536][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:58:41,040][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:58:41,544][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:58:42,047][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:58:42,550][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10102 tokens. -[2025-11-10 04:58:43,468][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 55.59%, Block Peak % of device VRAM: 62.23%, ΔTime: 00:00:33 -[2025-11-10 04:58:44,219][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:58:44,221][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:58:44,222][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:58:45,145][__main__][INFO] - Iteration 668 took 1m 1s (40.74% Gen, 57.76% Train). Generation: 25s, Training: 35s. Estimated remaining time: 40h 9m 22s. Estimated total time: 51h 9m 10s. Time estimates for 10 more iterations: 10m 13s, 100 more iterations: 1h 42m 18s, 500 more iterations: 8h 31m 31s. -[2025-11-10 04:58:45,147][__main__][INFO] - Starting iteration 668. -[2025-11-10 04:58:45,610][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 66 and human policies 1. -[2025-11-10 04:58:45,611][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 04:59:10,053][__main__][INFO] - Number of regex retries in iteration 668: 0 -[2025-11-10 04:59:10,055][__main__][INFO] - agents played in iteration 668 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 04:59:10,893][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:59:10,923][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:59:10,949][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:59:10,972][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 04:59:10,973][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 04:59:10,974][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 04:59:11,763][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 04:59:12,222][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 04:59:12,728][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 04:59:13,227][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 04:59:13,726][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 04:59:14,229][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 04:59:14,728][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 04:59:15,227][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 04:59:15,726][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 04:59:16,222][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 04:59:16,721][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 04:59:17,218][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 04:59:17,713][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 04:59:18,211][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 04:59:18,709][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 04:59:19,208][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 04:59:19,710][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 04:59:20,207][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 04:59:20,707][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 04:59:21,203][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 04:59:21,701][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 04:59:22,200][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 04:59:22,695][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 04:59:23,193][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 04:59:23,691][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 04:59:24,187][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 04:59:24,681][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 04:59:25,177][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 04:59:25,676][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 04:59:26,179][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 04:59:26,674][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 04:59:27,169][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 04:59:27,667][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 04:59:28,170][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 04:59:28,671][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 04:59:29,171][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 04:59:29,673][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 04:59:30,187][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 04:59:30,683][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 04:59:31,182][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 04:59:31,685][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 04:59:32,188][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 04:59:32,690][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 04:59:33,193][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 04:59:33,695][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 04:59:34,218][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 04:59:34,721][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 04:59:35,225][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 04:59:35,728][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 04:59:36,229][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 04:59:36,738][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 04:59:37,242][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 04:59:37,744][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 04:59:38,248][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 04:59:38,751][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 04:59:39,258][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 04:59:39,759][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 04:59:40,260][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 04:59:40,764][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 04:59:41,267][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 04:59:41,772][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 04:59:42,279][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 04:59:42,781][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 04:59:43,306][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 04:59:43,812][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10040 tokens. -[2025-11-10 04:59:44,809][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.19%, Current % of VRAM taken: 55.68%, Block Peak % of device VRAM: 62.08%, ΔTime: 00:00:33 -[2025-11-10 04:59:45,439][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 04:59:45,440][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 04:59:45,442][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 04:59:46,325][__main__][INFO] - Iteration 669 took 1m 0s (40.26% Gen, 58.28% Train). Generation: 24s, Training: 35s. Estimated remaining time: 39h 34m 58s. Estimated total time: 50h 35m 47s. Time estimates for 10 more iterations: 10m 7s, 100 more iterations: 1h 41m 11s, 500 more iterations: 8h 25m 57s. -[2025-11-10 04:59:46,328][__main__][INFO] - Starting iteration 669. -[2025-11-10 04:59:46,790][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 66 and human policies 1. -[2025-11-10 04:59:46,791][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:00:10,127][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 05:00:15,490][__main__][INFO] - Number of regex retries in iteration 669: 1 -[2025-11-10 05:00:15,491][__main__][INFO] - agents played in iteration 669 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:00:16,253][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:00:16,276][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:00:16,298][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:00:16,321][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:00:16,321][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:00:16,322][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:00:17,100][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:00:17,561][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:00:18,064][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:00:18,567][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:00:19,066][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:00:19,564][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:00:20,069][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:00:20,567][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:00:21,063][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:00:21,557][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:00:22,050][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:00:22,544][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:00:23,038][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:00:23,536][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:00:24,038][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:00:24,535][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:00:25,030][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:00:25,528][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:00:26,025][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:00:26,527][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:00:27,023][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:00:27,519][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:00:28,016][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:00:28,513][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:00:29,010][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:00:29,512][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:00:30,007][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:00:30,511][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:00:31,007][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:00:31,499][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:00:31,994][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:00:32,491][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:00:32,989][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:00:33,488][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:00:33,990][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:00:34,490][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:00:34,991][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:00:35,490][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:00:35,988][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:00:36,489][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:00:36,996][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:00:37,499][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:00:37,997][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:00:38,504][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:00:39,006][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:00:39,511][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:00:40,015][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:00:40,516][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:00:41,016][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:00:41,519][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:00:42,019][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:00:42,518][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:00:43,019][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:00:43,523][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:00:44,028][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:00:44,528][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:00:45,036][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:00:45,539][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:00:46,043][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:00:46,551][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:00:47,055][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:00:47,557][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:00:48,057][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:00:48,559][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:00:49,064][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10116 tokens. -[2025-11-10 05:00:49,999][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.09%, Current % of VRAM taken: 55.58%, Block Peak % of device VRAM: 62.28%, ΔTime: 00:00:32 -[2025-11-10 05:00:50,752][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:00:50,754][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:00:50,755][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:00:51,739][__main__][INFO] - Iteration 670 took 1m 4s (44.19% Gen, 54.30% Train). Generation: 28s, Training: 35s. Estimated remaining time: 43h 5m 34s. Estimated total time: 54h 7m 28s. Time estimates for 10 more iterations: 10m 49s, 100 more iterations: 1h 48m 14s, 500 more iterations: 9h 1m 14s. -[2025-11-10 05:00:51,741][__main__][INFO] - Starting iteration 670. -[2025-11-10 05:00:52,224][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 66 and human policies 1. -[2025-11-10 05:00:52,224][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:01:11,323][__main__][INFO] - Number of regex retries in iteration 670: 0 -[2025-11-10 05:01:11,324][__main__][INFO] - agents played in iteration 670 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:01:12,086][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:01:12,110][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:01:12,134][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:01:12,156][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:01:12,156][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:01:12,157][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:01:12,929][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:01:13,381][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:01:13,882][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:01:14,377][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:01:14,871][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:01:15,369][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:01:15,862][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:01:16,356][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:01:16,848][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:01:17,343][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:01:17,852][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:01:18,353][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:01:18,851][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:01:19,346][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:01:19,842][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:01:20,344][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:01:20,839][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:01:21,336][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:01:21,833][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:01:22,332][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:01:22,830][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:01:23,328][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:01:23,830][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:01:24,346][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:01:24,844][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:01:25,339][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:01:25,842][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:01:26,341][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:01:26,844][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:01:27,339][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:01:27,834][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:01:28,341][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:01:28,838][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:01:29,337][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:01:29,833][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:01:30,331][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:01:30,839][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:01:31,336][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:01:31,834][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:01:32,350][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:01:32,849][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:01:33,343][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:01:33,842][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:01:34,341][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:01:34,855][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:01:35,352][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:01:35,852][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:01:36,351][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:01:36,849][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:01:37,351][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:01:37,851][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:01:38,353][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:01:38,861][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:01:39,361][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:01:39,861][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:01:40,365][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:01:40,872][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:01:41,376][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:01:41,878][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:01:42,381][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:01:42,882][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:01:43,384][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:01:43,886][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:01:44,388][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:01:44,889][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10059 tokens. -[2025-11-10 05:01:45,836][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 55.58%, Block Peak % of device VRAM: 62.41%, ΔTime: 00:00:32 -[2025-11-10 05:01:46,530][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:01:46,531][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:01:46,535][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:01:48,463][__main__][INFO] - Iteration 671 took 56s (33.96% Gen, 62.61% Train). Generation: 19s, Training: 35s. Estimated remaining time: 35h 49m 8s. Estimated total time: 46h 52m 0s. Time estimates for 10 more iterations: 9m 22s, 100 more iterations: 1h 33m 44s, 500 more iterations: 7h 48m 40s. -[2025-11-10 05:01:48,465][__main__][INFO] - Starting iteration 671. -[2025-11-10 05:01:48,930][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 67 and human policies 1. -[2025-11-10 05:01:48,931][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:02:16,209][__main__][INFO] - Number of regex retries in iteration 671: 0 -[2025-11-10 05:02:16,210][__main__][INFO] - agents played in iteration 671 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:02:17,079][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:02:17,106][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:02:17,132][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:02:17,157][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:02:17,161][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:02:17,161][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:02:18,010][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:02:18,463][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:02:18,967][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:02:19,465][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:02:19,970][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:02:20,473][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:02:20,972][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:02:21,474][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:02:21,970][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:02:22,470][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:02:22,969][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:02:23,469][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:02:23,971][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:02:24,469][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:02:24,967][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:02:25,470][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:02:25,965][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:02:26,465][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:02:26,959][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:02:27,458][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:02:27,965][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:02:28,463][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:02:28,959][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:02:29,458][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:02:29,954][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:02:30,458][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:02:30,961][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:02:31,456][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:02:31,963][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:02:32,460][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:02:32,956][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:02:33,453][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:02:33,951][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:02:34,455][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:02:34,952][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:02:35,451][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:02:35,966][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:02:36,468][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:02:36,971][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:02:37,470][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:02:37,968][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:02:38,474][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:02:38,974][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:02:39,475][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:02:39,976][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:02:40,478][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:02:40,983][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:02:41,483][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:02:41,983][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:02:42,486][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:02:42,988][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:02:43,488][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:02:43,988][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:02:44,485][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:02:44,986][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:02:45,482][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:02:45,980][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:02:46,482][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:02:46,979][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:02:47,482][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:02:47,985][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:02:48,485][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:02:48,993][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:02:49,492][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:02:49,993][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10035 tokens. -[2025-11-10 05:02:50,894][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 55.60%, Block Peak % of device VRAM: 61.99%, ΔTime: 00:00:32 -[2025-11-10 05:02:51,636][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:02:51,638][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:02:51,639][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:02:52,557][__main__][INFO] - Iteration 672 took 1m 3s (42.87% Gen, 55.68% Train). Generation: 27s, Training: 35s. Estimated remaining time: 41h 57m 25s. Estimated total time: 53h 1m 21s. Time estimates for 10 more iterations: 10m 36s, 100 more iterations: 1h 46m 2s, 500 more iterations: 8h 50m 13s. -[2025-11-10 05:02:52,559][__main__][INFO] - Starting iteration 672. -[2025-11-10 05:02:53,057][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 67 and human policies 1. -[2025-11-10 05:02:53,058][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:03:12,853][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 05:03:23,747][__main__][INFO] - Number of regex retries in iteration 672: 1 -[2025-11-10 05:03:23,748][__main__][INFO] - agents played in iteration 672 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:03:24,519][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:03:24,542][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:03:24,565][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:03:24,587][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:03:24,587][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:03:24,588][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:03:25,383][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:03:25,840][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:03:26,347][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:03:26,843][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:03:27,339][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:03:27,836][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:03:28,333][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:03:28,835][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:03:29,352][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:03:29,854][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:03:30,356][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:03:30,853][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:03:31,350][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:03:31,856][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:03:32,355][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:03:32,852][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:03:33,350][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:03:33,847][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:03:34,341][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:03:34,838][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:03:35,333][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:03:35,845][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:03:36,342][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:03:36,836][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:03:37,334][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:03:37,832][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:03:38,331][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:03:38,830][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:03:39,331][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:03:39,859][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:03:40,360][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:03:40,861][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:03:41,361][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:03:41,859][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:03:42,367][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:03:42,875][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:03:43,378][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:03:43,882][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:03:44,391][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:03:44,895][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:03:45,395][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:03:45,896][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:03:46,397][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:03:46,903][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:03:47,402][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:03:47,905][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:03:48,409][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:03:48,911][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:03:49,415][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:03:49,916][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:03:50,424][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:03:50,924][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:03:51,425][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:03:51,926][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:03:52,426][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:03:52,932][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:03:53,434][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:03:53,937][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:03:54,450][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:03:54,957][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:03:55,477][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:03:55,976][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:03:56,477][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:03:56,980][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:03:57,484][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10114 tokens. -[2025-11-10 05:03:58,473][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.20%, Current % of VRAM taken: 55.70%, Block Peak % of device VRAM: 62.15%, ΔTime: 00:00:33 -[2025-11-10 05:03:59,126][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:03:59,127][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:03:59,129][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:03:59,931][__main__][INFO] - Iteration 673 took 1m 6s (45.89% Gen, 52.91% Train). Generation: 30s, Training: 35s. Estimated remaining time: 44h 38m 42s. Estimated total time: 55h 43m 45s. Time estimates for 10 more iterations: 11m 8s, 100 more iterations: 1h 51m 27s, 500 more iterations: 9h 17m 17s. -[2025-11-10 05:03:59,934][__main__][INFO] - Starting iteration 673. -[2025-11-10 05:04:00,411][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 67 and human policies 1. -[2025-11-10 05:04:00,412][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:04:25,034][__main__][INFO] - Number of regex retries in iteration 673: 0 -[2025-11-10 05:04:25,034][__main__][INFO] - agents played in iteration 673 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:04:25,898][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:04:25,921][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:04:25,944][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:04:25,966][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:04:25,967][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:04:25,968][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:04:26,841][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:04:27,303][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:04:27,809][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:04:28,311][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:04:28,810][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:04:29,314][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:04:29,812][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:04:30,311][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:04:30,810][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:04:31,316][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:04:31,815][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:04:32,312][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:04:32,811][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:04:33,312][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:04:33,809][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:04:34,310][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:04:34,811][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:04:35,310][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:04:35,816][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:04:36,314][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:04:36,810][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:04:37,307][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:04:37,802][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:04:38,305][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:04:38,799][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:04:39,296][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:04:39,795][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:04:40,292][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:04:40,791][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:04:41,292][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:04:41,791][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:04:42,298][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:04:42,796][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:04:43,293][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:04:43,792][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:04:44,292][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:04:44,792][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:04:45,290][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:04:45,786][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:04:46,294][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:04:46,791][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:04:47,291][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:04:47,795][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:04:48,295][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:04:48,800][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:04:49,301][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:04:49,801][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:04:50,304][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:04:50,804][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:04:51,311][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:04:51,818][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:04:52,322][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:04:52,825][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:04:53,331][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:04:53,836][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:04:54,352][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:04:54,857][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:04:55,373][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:04:55,877][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:04:56,382][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:04:56,885][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:04:57,387][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:04:57,888][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:04:58,394][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:04:58,897][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9976 tokens. -[2025-11-10 05:04:59,825][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 55.62%, Block Peak % of device VRAM: 61.99%, ΔTime: 00:00:32 -[2025-11-10 05:05:00,588][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:05:00,590][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:05:00,592][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:05:01,506][__main__][INFO] - Iteration 674 took 1m 1s (40.30% Gen, 58.20% Train). Generation: 24s, Training: 35s. Estimated remaining time: 39h 48m 42s. Estimated total time: 50h 54m 46s. Time estimates for 10 more iterations: 10m 10s, 100 more iterations: 1h 41m 49s, 500 more iterations: 8h 29m 7s. -[2025-11-10 05:05:01,508][__main__][INFO] - Starting iteration 674. -[2025-11-10 05:05:01,997][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 67 and human policies 1. -[2025-11-10 05:05:01,998][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:05:27,565][__main__][INFO] - Number of regex retries in iteration 674: 0 -[2025-11-10 05:05:27,565][__main__][INFO] - agents played in iteration 674 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:05:28,349][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:05:28,377][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:05:28,404][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:05:28,427][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:05:28,428][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:05:28,429][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:05:29,224][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:05:29,916][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:05:30,430][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:05:30,928][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:05:31,429][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:05:31,924][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:05:32,421][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:05:32,916][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:05:33,416][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:05:33,911][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:05:34,408][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:05:34,905][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:05:35,403][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:05:35,899][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:05:36,393][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:05:36,893][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:05:37,397][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:05:37,894][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:05:38,392][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:05:38,884][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:05:39,378][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:05:39,881][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:05:40,379][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:05:40,877][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:05:41,393][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:05:41,887][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:05:42,382][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:05:42,876][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:05:43,372][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:05:43,879][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:05:44,373][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:05:44,870][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:05:45,387][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:05:45,884][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:05:46,388][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:05:46,886][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:05:47,382][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:05:47,891][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:05:48,392][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:05:48,894][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:05:49,400][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:05:49,904][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:05:50,410][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:05:50,909][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:05:51,410][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:05:51,915][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:05:52,417][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:05:52,921][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:05:53,423][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:05:53,930][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:05:54,433][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:05:54,935][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:05:55,440][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:05:55,949][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:05:57,111][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:05:57,631][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:05:58,133][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:05:58,637][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:05:59,140][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:05:59,645][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:06:00,147][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:06:00,653][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:06:01,166][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:06:01,668][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:06:02,171][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10072 tokens. -[2025-11-10 05:06:03,138][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.00%, Current % of VRAM taken: 55.50%, Block Peak % of device VRAM: 62.21%, ΔTime: 00:00:33 -[2025-11-10 05:06:03,790][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:06:03,794][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:06:03,795][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:06:04,660][__main__][INFO] - Iteration 675 took 1m 2s (40.80% Gen, 57.82% Train). Generation: 25s, Training: 36s. Estimated remaining time: 41h 6m 4s. Estimated total time: 52h 13m 12s. Time estimates for 10 more iterations: 10m 26s, 100 more iterations: 1h 44m 26s, 500 more iterations: 8h 42m 12s. -[2025-11-10 05:06:04,662][__main__][INFO] - Starting iteration 675. -[2025-11-10 05:06:05,138][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 67 and human policies 1. -[2025-11-10 05:06:05,138][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:06:33,552][__main__][INFO] - Number of regex retries in iteration 675: 0 -[2025-11-10 05:06:33,553][__main__][INFO] - agents played in iteration 675 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:06:34,398][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:06:34,422][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:06:34,446][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:06:34,468][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:06:34,469][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:06:34,470][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:06:35,263][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:06:35,714][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:06:36,226][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:06:36,723][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:06:37,219][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:06:37,714][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:06:38,210][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:06:38,712][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:06:39,209][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:06:39,703][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:06:40,199][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:06:40,694][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:06:41,189][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:06:41,689][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:06:42,188][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:06:42,700][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:06:43,196][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:06:43,689][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:06:44,190][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:06:44,686][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:06:45,197][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:06:45,693][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:06:46,188][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:06:46,707][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:06:47,201][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:06:47,699][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:06:48,198][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:06:48,694][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:06:49,191][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:06:49,688][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:06:50,185][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:06:50,683][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:06:51,180][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:06:51,678][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:06:52,180][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:06:52,681][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:06:53,197][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:06:53,697][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:06:54,198][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:06:54,703][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:06:55,202][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:06:55,715][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:06:56,226][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:06:56,734][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:06:57,238][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:06:57,739][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:06:58,243][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:06:58,746][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:06:59,248][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:06:59,758][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:07:00,260][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:07:00,764][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:07:01,277][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:07:01,783][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:07:02,291][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:07:02,794][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:07:03,298][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:07:03,800][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:07:04,300][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:07:04,812][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:07:05,315][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:07:05,817][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:07:06,319][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:07:06,820][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:07:07,327][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10020 tokens. -[2025-11-10 05:07:08,288][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 55.59%, Block Peak % of device VRAM: 62.17%, ΔTime: 00:00:33 -[2025-11-10 05:07:09,054][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:07:09,055][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:07:09,057][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:07:09,990][__main__][INFO] - Iteration 676 took 1m 4s (43.81% Gen, 54.75% Train). Generation: 28s, Training: 35s. Estimated remaining time: 42h 54m 26s. Estimated total time: 54h 2m 39s. Time estimates for 10 more iterations: 10m 48s, 100 more iterations: 1h 48m 5s, 500 more iterations: 9h 0m 26s. -[2025-11-10 05:07:09,992][__main__][INFO] - Starting iteration 676. -[2025-11-10 05:07:10,460][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 67 and human policies 1. -[2025-11-10 05:07:10,460][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:07:35,443][__main__][INFO] - Number of regex retries in iteration 676: 0 -[2025-11-10 05:07:35,445][__main__][INFO] - agents played in iteration 676 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:07:36,228][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:07:36,251][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:07:36,276][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:07:36,298][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:07:36,299][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:07:36,300][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:07:37,104][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:07:37,562][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:07:38,067][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:07:38,566][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:07:39,064][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:07:39,564][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:07:40,066][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:07:40,567][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:07:41,066][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:07:41,561][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:07:42,060][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:07:42,560][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:07:43,054][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:07:43,554][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:07:44,052][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:07:44,555][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:07:45,052][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:07:45,552][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:07:46,050][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:07:46,550][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:07:47,051][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:07:47,550][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:07:48,050][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:07:48,555][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:07:49,053][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:07:49,554][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:07:50,050][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:07:50,548][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:07:51,046][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:07:51,543][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:07:52,039][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:07:52,538][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:07:53,034][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:07:53,533][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:07:54,033][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:07:54,531][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:07:55,041][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:07:55,541][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:07:56,040][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:07:56,538][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:07:57,035][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:07:57,533][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:07:58,035][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:07:58,542][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:07:59,044][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:07:59,545][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:08:00,044][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:08:00,545][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:08:01,045][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:08:01,550][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:08:02,052][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:08:02,553][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:08:03,059][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:08:03,561][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:08:04,061][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:08:04,563][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:08:05,064][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:08:05,573][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:08:06,073][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:08:06,574][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:08:07,079][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:08:07,580][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:08:08,081][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:08:08,595][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:08:09,098][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10100 tokens. -[2025-11-10 05:08:10,038][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 55.57%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 -[2025-11-10 05:08:10,683][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:08:10,685][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:08:10,688][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:08:11,545][__main__][INFO] - Iteration 677 took 1m 1s (40.90% Gen, 57.69% Train). Generation: 24s, Training: 35s. Estimated remaining time: 39h 45m 4s. Estimated total time: 50h 54m 19s. Time estimates for 10 more iterations: 10m 10s, 100 more iterations: 1h 41m 48s, 500 more iterations: 8h 29m 3s. -[2025-11-10 05:08:11,548][__main__][INFO] - Starting iteration 677. -[2025-11-10 05:08:12,021][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 67 and human policies 1. -[2025-11-10 05:08:12,022][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:08:40,720][__main__][INFO] - Number of regex retries in iteration 677: 0 -[2025-11-10 05:08:40,721][__main__][INFO] - agents played in iteration 677 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:08:41,500][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:08:41,524][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:08:41,546][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:08:41,568][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:08:41,569][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:08:41,570][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:08:42,361][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:08:42,822][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:08:43,323][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:08:43,829][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:08:44,328][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:08:44,826][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:08:45,327][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:08:45,822][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:08:46,318][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:08:46,814][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:08:47,310][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:08:47,826][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:08:48,323][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:08:48,823][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:08:49,332][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:08:49,830][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:08:50,325][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:08:50,828][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:08:51,323][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:08:51,827][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:08:52,322][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:08:52,817][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:08:53,331][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:08:53,827][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:08:54,323][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:08:54,817][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:08:55,315][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:08:55,832][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:08:56,331][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:08:56,832][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:08:57,328][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:08:57,823][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:08:58,326][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:08:58,822][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:08:59,318][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:08:59,829][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:09:00,328][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:09:00,829][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:09:01,325][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:09:01,822][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:09:02,336][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:09:02,832][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:09:03,334][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:09:03,836][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:09:04,339][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:09:04,843][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:09:05,355][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:09:05,856][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:09:06,358][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:09:06,860][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:09:07,365][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:09:07,869][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:09:08,373][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:09:08,877][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:09:09,381][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:09:09,883][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:09:10,391][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:09:10,896][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:09:11,398][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:09:11,902][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:09:12,412][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:09:12,918][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:09:13,421][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:09:13,938][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:09:14,441][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10039 tokens. -[2025-11-10 05:09:15,416][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.10%, Current % of VRAM taken: 55.59%, Block Peak % of device VRAM: 62.11%, ΔTime: 00:00:33 -[2025-11-10 05:09:16,160][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:09:16,161][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:09:16,163][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:09:17,108][__main__][INFO] - Iteration 678 took 1m 5s (44.09% Gen, 54.45% Train). Generation: 28s, Training: 35s. Estimated remaining time: 43h 4m 1s. Estimated total time: 54h 14m 21s. Time estimates for 10 more iterations: 10m 50s, 100 more iterations: 1h 48m 28s, 500 more iterations: 9h 2m 23s. -[2025-11-10 05:09:17,110][__main__][INFO] - Starting iteration 678. -[2025-11-10 05:09:17,592][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 67 and human policies 1. -[2025-11-10 05:09:17,592][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:09:41,645][__main__][INFO] - Number of regex retries in iteration 678: 0 -[2025-11-10 05:09:41,647][__main__][INFO] - agents played in iteration 678 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:09:42,430][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:09:42,457][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:09:42,484][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:09:42,507][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:09:42,507][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:09:42,508][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:09:43,303][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:09:43,762][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:09:44,265][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:09:44,769][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:09:45,266][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:09:45,765][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:09:46,263][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:09:46,761][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:09:47,262][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:09:47,762][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:09:48,260][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:09:48,760][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:09:49,257][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:09:49,753][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:09:50,249][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:09:50,751][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:09:51,246][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:09:51,746][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:09:52,246][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:09:52,748][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:09:53,248][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:09:53,747][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:09:54,270][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:09:54,773][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:09:55,271][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:09:55,768][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:09:56,268][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:09:56,765][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:09:57,263][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:09:57,761][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:09:58,258][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:09:58,762][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:09:59,258][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:09:59,754][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:10:00,256][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:10:00,754][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:10:01,252][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:10:01,753][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:10:02,247][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:10:02,742][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:10:03,237][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:10:03,732][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:10:04,228][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:10:04,722][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:10:05,216][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:10:05,711][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:10:06,205][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:10:06,707][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:10:07,204][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:10:07,698][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:10:08,197][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:10:08,693][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:10:09,194][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:10:09,695][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:10:10,195][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:10:10,695][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:10:11,193][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:10:11,697][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:10:12,199][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:10:12,701][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:10:13,204][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:10:13,714][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:10:14,218][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:10:14,721][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:10:15,224][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10096 tokens. -[2025-11-10 05:10:16,172][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.22%, Current % of VRAM taken: 55.71%, Block Peak % of device VRAM: 62.42%, ΔTime: 00:00:32 -[2025-11-10 05:10:16,798][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:10:16,801][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:10:16,803][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:10:17,676][__main__][INFO] - Iteration 679 took 1m 0s (40.03% Gen, 58.51% Train). Generation: 24s, Training: 35s. Estimated remaining time: 38h 52m 56s. Estimated total time: 50h 4m 17s. Time estimates for 10 more iterations: 10m 0s, 100 more iterations: 1h 40m 8s, 500 more iterations: 8h 20m 42s. -[2025-11-10 05:10:17,679][__main__][INFO] - Starting iteration 679. -[2025-11-10 05:10:18,140][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 67 and human policies 1. -[2025-11-10 05:10:18,140][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:10:47,034][__main__][INFO] - Number of regex retries in iteration 679: 0 -[2025-11-10 05:10:47,035][__main__][INFO] - agents played in iteration 679 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:10:47,823][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:10:47,850][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:10:47,877][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:10:47,901][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:10:47,902][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:10:47,903][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:10:48,700][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:10:49,153][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:10:49,654][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:10:50,158][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:10:50,657][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:10:51,159][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:10:51,666][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:10:52,162][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:10:52,661][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:10:53,156][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:10:53,651][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:10:54,145][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:10:54,639][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:10:55,138][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:10:55,637][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:10:56,137][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:10:56,632][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:10:57,128][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:10:57,624][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:10:58,119][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:10:58,616][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:10:59,112][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:10:59,624][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:11:00,119][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:11:00,617][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:11:01,121][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:11:01,617][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:11:02,116][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:11:02,613][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:11:03,111][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:11:03,620][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:11:04,113][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:11:04,614][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:11:05,111][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:11:05,609][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:11:06,111][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:11:06,609][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:11:07,105][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:11:07,604][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:11:08,099][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:11:08,599][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:11:09,097][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:11:09,595][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:11:10,092][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:11:10,588][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:11:11,089][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:11:11,591][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:11:12,092][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:11:12,593][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:11:13,096][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:11:13,592][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:11:14,095][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:11:14,595][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:11:15,098][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:11:15,598][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:11:16,104][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:11:16,602][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:11:17,102][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:11:17,605][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:11:18,106][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:11:18,607][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:11:19,106][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:11:19,610][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:11:20,113][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:11:20,623][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10101 tokens. -[2025-11-10 05:11:21,578][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.03%, Current % of VRAM taken: 55.53%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:32 -[2025-11-10 05:11:22,329][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:11:22,330][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:11:22,332][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:11:23,271][__main__][INFO] - Iteration 680 took 1m 5s (44.36% Gen, 54.19% Train). Generation: 28s, Training: 35s. Estimated remaining time: 43h 4m 9s. Estimated total time: 54h 16m 35s. Time estimates for 10 more iterations: 10m 51s, 100 more iterations: 1h 48m 33s, 500 more iterations: 9h 2m 45s. -[2025-11-10 05:11:23,273][__main__][INFO] - Starting iteration 680. -[2025-11-10 05:11:23,737][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 67 and human policies 1. -[2025-11-10 05:11:23,737][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:11:43,990][__main__][INFO] - Number of regex retries in iteration 680: 0 -[2025-11-10 05:11:43,990][__main__][INFO] - agents played in iteration 680 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:11:44,759][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:11:44,787][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:11:44,813][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:11:44,837][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.52%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:11:44,837][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:11:44,838][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:11:45,665][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:11:46,123][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:11:46,625][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:11:47,122][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:11:47,619][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:11:48,116][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:11:48,616][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:11:49,122][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:11:49,622][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:11:50,124][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:11:50,622][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:11:51,124][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:11:51,624][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:11:52,120][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:11:52,617][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:11:53,114][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:11:53,611][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:11:54,108][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:11:54,607][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:11:55,102][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:11:55,602][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:11:56,122][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:11:56,622][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:11:57,118][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:11:57,614][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:11:58,112][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:11:58,612][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:11:59,112][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:11:59,610][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:12:00,111][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:12:00,608][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:12:01,103][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:12:01,598][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:12:02,095][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:12:02,595][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:12:03,091][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:12:03,592][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:12:04,090][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:12:04,590][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:12:05,088][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:12:05,585][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:12:06,086][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:12:06,584][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:12:07,080][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:12:07,575][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:12:08,072][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:12:08,571][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:12:09,067][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:12:09,563][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:12:10,068][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:12:10,564][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:12:11,064][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:12:11,563][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:12:12,062][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:12:12,566][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:12:13,069][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:12:13,568][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:12:14,076][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:12:14,577][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:12:15,080][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:12:15,585][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:12:16,086][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:12:16,592][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:12:17,094][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:12:17,596][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10091 tokens. -[2025-11-10 05:12:18,515][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 55.65%, Block Peak % of device VRAM: 62.13%, ΔTime: 00:00:32 -[2025-11-10 05:12:19,255][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:12:19,257][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:12:19,259][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:12:21,018][__main__][INFO] - Iteration 681 took 57s (35.36% Gen, 61.57% Train). Generation: 20s, Training: 35s. Estimated remaining time: 36h 30m 42s. Estimated total time: 47h 44m 6s. Time estimates for 10 more iterations: 9m 32s, 100 more iterations: 1h 35m 28s, 500 more iterations: 7h 57m 21s. -[2025-11-10 05:12:21,021][__main__][INFO] - Starting iteration 681. -[2025-11-10 05:12:21,498][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 68 and human policies 1. -[2025-11-10 05:12:21,499][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:12:50,121][__main__][INFO] - Number of regex retries in iteration 681: 0 -[2025-11-10 05:12:50,122][__main__][INFO] - agents played in iteration 681 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:12:50,889][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:12:50,916][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:12:50,942][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:12:50,965][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:12:50,966][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:12:50,966][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:12:51,772][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:12:52,225][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:12:52,730][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:12:53,231][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:12:53,730][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:12:54,228][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:12:54,730][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:12:55,230][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:12:55,734][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:12:56,235][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:12:56,735][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:12:57,236][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:12:57,740][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:12:58,242][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:12:58,744][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:12:59,251][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:12:59,763][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:13:00,261][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:13:00,756][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:13:01,254][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:13:01,752][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:13:02,247][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:13:02,749][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:13:03,245][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:13:03,747][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:13:04,243][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:13:04,739][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:13:05,237][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:13:05,733][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:13:06,228][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:13:06,727][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:13:07,222][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:13:07,719][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:13:08,218][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:13:08,712][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:13:09,211][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:13:09,707][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:13:10,202][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:13:10,696][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:13:11,191][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:13:11,690][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:13:12,185][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:13:12,681][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:13:13,178][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:13:13,676][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:13:14,173][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:13:14,672][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:13:15,173][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:13:15,680][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:13:16,180][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:13:16,682][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:13:17,185][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:13:17,685][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:13:18,186][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:13:18,687][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:13:19,192][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:13:19,694][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:13:20,197][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:13:20,699][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:13:21,204][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:13:21,708][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:13:22,214][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:13:22,720][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:13:23,224][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:13:23,738][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10042 tokens. -[2025-11-10 05:13:24,689][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 55.65%, Block Peak % of device VRAM: 62.12%, ΔTime: 00:00:32 -[2025-11-10 05:13:25,436][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:13:25,437][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:13:25,439][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:13:26,488][__main__][INFO] - Iteration 682 took 1m 4s (44.04% Gen, 54.34% Train). Generation: 28s, Training: 35s. Estimated remaining time: 42h 55m 3s. Estimated total time: 54h 9m 33s. Time estimates for 10 more iterations: 10m 49s, 100 more iterations: 1h 48m 19s, 500 more iterations: 9h 1m 35s. -[2025-11-10 05:13:26,490][__main__][INFO] - Starting iteration 682. -[2025-11-10 05:13:26,953][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 68 and human policies 1. -[2025-11-10 05:13:26,954][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:13:51,114][__main__][INFO] - Number of regex retries in iteration 682: 0 -[2025-11-10 05:13:51,115][__main__][INFO] - agents played in iteration 682 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:13:51,894][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:13:51,920][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:13:51,945][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:13:51,967][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:13:51,967][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:13:51,968][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:13:52,793][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:13:53,249][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:13:53,770][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:13:54,267][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:13:54,766][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:13:55,265][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:13:55,762][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:13:56,261][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:13:56,767][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:13:57,271][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:13:57,784][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:13:58,285][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:13:58,787][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:13:59,288][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:13:59,792][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:14:00,295][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:14:00,792][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:14:01,288][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:14:01,789][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:14:02,285][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:14:02,784][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:14:03,282][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:14:04,098][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:14:04,623][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:14:05,119][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:14:05,618][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:14:06,186][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:14:06,686][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:14:07,180][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:14:07,675][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:14:08,176][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:14:08,671][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:14:09,168][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:14:09,668][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:14:10,162][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:14:10,656][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:14:11,153][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:14:11,649][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:14:12,159][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:14:12,655][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:14:13,152][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:14:13,658][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:14:14,161][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:14:14,663][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:14:15,166][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:14:15,668][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:14:16,175][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:14:16,701][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:14:17,197][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:14:17,703][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:14:18,206][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:14:18,712][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:14:19,221][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:14:19,728][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:14:20,236][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:14:20,743][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:14:21,266][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:14:21,767][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:14:22,271][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:14:22,777][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:14:23,279][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:14:23,785][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:14:24,289][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:14:24,793][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:14:25,296][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10107 tokens. -[2025-11-10 05:14:26,243][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 55.66%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:33 -[2025-11-10 05:14:26,892][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:14:26,894][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:14:26,896][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:14:27,786][__main__][INFO] - Iteration 683 took 1m 0s (39.72% Gen, 58.82% Train). Generation: 24s, Training: 35s. Estimated remaining time: 39h 26m 10s. Estimated total time: 50h 41m 40s. Time estimates for 10 more iterations: 10m 8s, 100 more iterations: 1h 41m 23s, 500 more iterations: 8h 26m 56s. -[2025-11-10 05:14:27,788][__main__][INFO] - Starting iteration 683. -[2025-11-10 05:14:28,257][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 68 and human policies 1. -[2025-11-10 05:14:28,258][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:14:57,810][__main__][INFO] - Number of regex retries in iteration 683: 0 -[2025-11-10 05:14:57,811][__main__][INFO] - agents played in iteration 683 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:14:58,663][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:14:58,690][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:14:58,715][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:14:58,738][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:14:58,739][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:14:58,740][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:14:59,569][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:15:00,027][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:15:00,530][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:15:01,030][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:15:01,531][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:15:02,031][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:15:02,534][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:15:03,035][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:15:03,538][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:15:04,037][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:15:04,539][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:15:05,040][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:15:05,537][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:15:06,038][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:15:06,540][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:15:07,045][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:15:07,545][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:15:08,046][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:15:08,554][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:15:09,050][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:15:09,552][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:15:10,050][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:15:10,545][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:15:11,045][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:15:11,540][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:15:12,035][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:15:12,531][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:15:13,028][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:15:13,529][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:15:14,029][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:15:14,525][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:15:15,025][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:15:15,521][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:15:16,021][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:15:16,520][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:15:17,016][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:15:17,515][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:15:18,019][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:15:18,514][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:15:19,013][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:15:19,519][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:15:20,017][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:15:20,516][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:15:21,017][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:15:21,515][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:15:22,013][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:15:22,514][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:15:23,012][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:15:23,511][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:15:24,012][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:15:24,512][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:15:25,012][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:15:25,518][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:15:26,019][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:15:26,522][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:15:27,024][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:15:27,526][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:15:28,033][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:15:28,534][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:15:29,036][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:15:29,540][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:15:30,041][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:15:30,546][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:15:31,052][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:15:31,553][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10063 tokens. -[2025-11-10 05:15:32,523][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.04%, Current % of VRAM taken: 55.54%, Block Peak % of device VRAM: 62.09%, ΔTime: 00:00:32 -[2025-11-10 05:15:33,287][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:15:33,288][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:15:33,290][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:15:34,150][__main__][INFO] - Iteration 684 took 1m 5s (44.85% Gen, 53.84% Train). Generation: 29s, Training: 35s. Estimated remaining time: 43h 38m 4s. Estimated total time: 54h 54m 42s. Time estimates for 10 more iterations: 10m 58s, 100 more iterations: 1h 49m 49s, 500 more iterations: 9h 9m 7s. -[2025-11-10 05:15:34,153][__main__][INFO] - Starting iteration 684. -[2025-11-10 05:15:34,639][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 68 and human policies 1. -[2025-11-10 05:15:34,640][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:15:55,655][__main__][INFO] - Number of regex retries in iteration 684: 0 -[2025-11-10 05:15:55,656][__main__][INFO] - agents played in iteration 684 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:15:56,494][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:15:56,521][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:15:56,548][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:15:56,573][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:15:56,573][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:15:56,574][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:15:57,414][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:15:57,867][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:15:58,386][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:15:58,891][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:15:59,390][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:15:59,891][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:16:00,394][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:16:00,900][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:16:01,399][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:16:01,902][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:16:02,405][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:16:02,904][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:16:03,409][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:16:03,916][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:16:04,422][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:16:04,929][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:16:05,428][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:16:05,924][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:16:06,421][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:16:06,917][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:16:07,419][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:16:07,914][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:16:08,412][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:16:08,908][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:16:09,404][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:16:09,900][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:16:10,396][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:16:10,892][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:16:11,385][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:16:11,882][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:16:12,377][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:16:12,877][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:16:13,373][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:16:13,874][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:16:14,373][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:16:14,869][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:16:15,367][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:16:15,862][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:16:16,358][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:16:16,871][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:16:17,367][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:16:17,865][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:16:18,361][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:16:18,857][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:16:19,357][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:16:19,853][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:16:20,353][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:16:20,865][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:16:21,365][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:16:21,883][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:16:22,384][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:16:22,889][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:16:23,392][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:16:23,895][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:16:24,405][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:16:24,905][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:16:25,408][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:16:25,912][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:16:26,415][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:16:26,916][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:16:27,419][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:16:27,919][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:16:28,424][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:16:28,926][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:16:29,431][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10026 tokens. -[2025-11-10 05:16:30,385][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 55.57%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:32 -[2025-11-10 05:16:32,898][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:16:32,902][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:16:32,904][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:16:33,813][__main__][INFO] - Iteration 685 took 59s (35.51% Gen, 62.95% Train). Generation: 21s, Training: 37s. Estimated remaining time: 38h 1m 9s. Estimated total time: 49h 18m 46s. Time estimates for 10 more iterations: 9m 51s, 100 more iterations: 1h 38m 37s, 500 more iterations: 8h 13m 7s. -[2025-11-10 05:16:33,817][__main__][INFO] - Starting iteration 685. -[2025-11-10 05:16:34,282][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 68 and human policies 1. -[2025-11-10 05:16:34,282][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:17:05,144][__main__][INFO] - Number of regex retries in iteration 685: 0 -[2025-11-10 05:17:05,145][__main__][INFO] - agents played in iteration 685 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:17:05,913][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:17:05,941][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:17:05,967][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:17:05,990][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:17:05,990][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:17:05,991][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:17:06,812][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:17:07,269][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:17:07,777][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:17:08,275][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:17:08,776][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:17:09,279][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:17:09,777][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:17:10,275][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:17:10,778][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:17:11,275][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:17:11,772][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:17:12,269][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:17:12,768][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:17:13,269][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:17:13,770][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:17:14,267][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:17:14,771][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:17:15,270][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:17:15,773][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:17:16,274][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:17:16,772][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:17:17,272][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:17:17,769][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:17:18,269][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:17:18,765][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:17:19,263][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:17:19,766][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:17:20,262][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:17:20,758][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:17:21,256][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:17:21,755][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:17:22,251][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:17:22,752][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:17:23,254][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:17:23,758][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:17:24,257][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:17:24,760][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:17:25,259][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:17:25,762][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:17:26,271][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:17:26,774][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:17:27,277][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:17:27,780][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:17:28,284][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:17:28,786][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:17:29,287][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:17:29,785][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:17:30,288][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:17:30,791][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:17:31,292][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:17:31,803][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:17:32,307][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:17:32,809][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:17:33,312][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:17:33,813][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:17:34,321][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:17:34,825][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:17:35,323][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:17:35,822][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:17:36,320][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:17:36,838][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:17:37,340][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:17:37,843][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:17:38,342][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:17:38,846][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10170 tokens. -[2025-11-10 05:17:39,809][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 55.63%, Block Peak % of device VRAM: 62.27%, ΔTime: 00:00:33 -[2025-11-10 05:17:40,591][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:17:40,594][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:17:40,596][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:17:41,526][__main__][INFO] - Iteration 686 took 1m 7s (45.90% Gen, 52.72% Train). Generation: 30s, Training: 35s. Estimated remaining time: 44h 43m 32s. Estimated total time: 56h 2m 16s. Time estimates for 10 more iterations: 11m 12s, 100 more iterations: 1h 52m 4s, 500 more iterations: 9h 20m 22s. -[2025-11-10 05:17:41,529][__main__][INFO] - Starting iteration 686. -[2025-11-10 05:17:41,994][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 68 and human policies 1. -[2025-11-10 05:17:41,995][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:18:05,318][__main__][INFO] - Number of regex retries in iteration 686: 0 -[2025-11-10 05:18:05,318][__main__][INFO] - agents played in iteration 686 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:18:06,160][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:18:06,187][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:18:06,212][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:18:06,234][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:18:06,235][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:18:06,236][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:18:07,044][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:18:07,499][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:18:07,999][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:18:08,497][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:18:09,002][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:18:09,500][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:18:10,012][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:18:10,509][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:18:11,005][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:18:11,504][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:18:12,001][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:18:12,499][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:18:12,996][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:18:13,494][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:18:13,995][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:18:14,496][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:18:15,003][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:18:15,504][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:18:16,033][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:18:16,534][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:18:17,039][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:18:17,542][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:18:18,048][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:18:18,546][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:18:19,047][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:18:19,544][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:18:20,042][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:18:20,539][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:18:21,035][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:18:21,543][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:18:22,040][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:18:22,537][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:18:23,037][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:18:23,534][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:18:24,038][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:18:24,535][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:18:25,033][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:18:25,532][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:18:26,030][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:18:26,529][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:18:27,029][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:18:27,526][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:18:28,042][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:18:28,540][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:18:29,038][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:18:29,538][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:18:30,034][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:18:30,538][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:18:31,036][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:18:31,537][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:18:32,037][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:18:32,537][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:18:33,038][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:18:33,538][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:18:34,035][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:18:34,551][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:18:35,053][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:18:35,554][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:18:36,056][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:18:36,564][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:18:37,072][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:18:37,577][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:18:38,082][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:18:38,586][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:18:39,089][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9975 tokens. -[2025-11-10 05:18:41,212][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 55.58%, Block Peak % of device VRAM: 62.06%, ΔTime: 00:00:34 -[2025-11-10 05:18:42,132][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:18:42,135][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:18:42,137][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:18:43,012][__main__][INFO] - Iteration 687 took 1m 1s (38.22% Gen, 60.34% Train). Generation: 23s, Training: 36s. Estimated remaining time: 39h 31m 9s. Estimated total time: 50h 50m 55s. Time estimates for 10 more iterations: 10m 10s, 100 more iterations: 1h 41m 41s, 500 more iterations: 8h 28m 29s. -[2025-11-10 05:18:43,016][__main__][INFO] - Starting iteration 687. -[2025-11-10 05:18:43,497][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 68 and human policies 1. -[2025-11-10 05:18:43,498][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:19:17,337][__main__][INFO] - Number of regex retries in iteration 687: 0 -[2025-11-10 05:19:17,337][__main__][INFO] - agents played in iteration 687 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:19:18,181][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:19:18,208][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:19:18,235][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:19:18,257][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:19:18,258][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:19:18,259][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:19:19,073][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:19:19,527][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:19:20,031][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:19:20,531][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:19:21,031][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:19:21,529][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:19:22,047][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:19:22,551][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:19:23,052][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:19:23,554][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:19:24,056][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:19:24,565][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:19:25,059][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:19:25,553][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:19:26,048][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:19:26,546][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:19:27,051][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:19:27,545][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:19:28,040][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:19:28,545][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:19:29,041][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:19:29,537][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:19:30,037][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:19:30,532][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:19:31,031][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:19:31,530][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:19:32,059][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:19:32,570][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:19:33,074][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:19:33,574][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:19:34,074][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:19:34,572][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:19:35,074][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:19:35,574][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:19:36,073][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:19:36,573][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:19:37,072][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:19:37,571][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:19:38,072][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:19:38,577][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:19:39,080][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:19:39,579][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:19:40,078][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:19:40,575][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:19:41,076][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:19:41,579][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:19:42,079][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:19:42,580][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:19:43,081][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:19:43,591][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:19:44,094][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:19:44,596][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:19:45,099][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:19:45,608][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:19:46,114][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:19:46,616][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:19:47,148][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:19:47,654][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:19:48,154][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:19:48,655][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:19:49,158][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:19:49,668][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:19:50,173][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:19:50,674][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:19:51,175][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10097 tokens. -[2025-11-10 05:19:52,095][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 55.66%, Block Peak % of device VRAM: 62.43%, ΔTime: 00:00:33 -[2025-11-10 05:19:52,866][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:19:52,867][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:19:52,869][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:19:53,957][__main__][INFO] - Iteration 688 took 1m 10s (48.03% Gen, 50.43% Train). Generation: 33s, Training: 35s. Estimated remaining time: 47h 22m 4s. Estimated total time: 58h 43m 1s. Time estimates for 10 more iterations: 11m 44s, 100 more iterations: 1h 57m 26s, 500 more iterations: 9h 47m 10s. -[2025-11-10 05:19:53,959][__main__][INFO] - Starting iteration 688. -[2025-11-10 05:19:54,422][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 68 and human policies 1. -[2025-11-10 05:19:54,423][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:20:19,717][__main__][INFO] - Number of regex retries in iteration 688: 0 -[2025-11-10 05:20:19,718][__main__][INFO] - agents played in iteration 688 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:20:20,491][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:20:20,519][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:20:20,546][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:20:20,574][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:20:20,575][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:20:20,577][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:20:21,421][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:20:21,879][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:20:22,382][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:20:22,883][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:20:23,390][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:20:23,889][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:20:24,390][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:20:24,894][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:20:25,397][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:20:25,907][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:20:26,409][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:20:26,908][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:20:27,406][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:20:27,904][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:20:28,401][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:20:28,898][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:20:29,396][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:20:29,909][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:20:30,406][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:20:30,904][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:20:31,402][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:20:31,899][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:20:32,396][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:20:32,894][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:20:33,391][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:20:33,887][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:20:34,384][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:20:34,888][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:20:35,386][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:20:35,884][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:20:36,383][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:20:36,880][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:20:37,378][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:20:37,902][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:20:38,402][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:20:38,904][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:20:39,413][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:20:39,923][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:20:40,425][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:20:40,929][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:20:41,429][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:20:41,940][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:20:42,445][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:20:42,952][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:20:43,459][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:20:43,961][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:20:44,475][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:20:44,982][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:20:45,486][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:20:45,996][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:20:46,500][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:20:47,002][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:20:47,504][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:20:48,087][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:20:48,706][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:20:49,214][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:20:49,719][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:20:50,226][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:20:50,729][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:20:51,327][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:20:51,843][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:20:52,354][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:20:52,854][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:20:53,356][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:20:53,876][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9993 tokens. -[2025-11-10 05:20:54,823][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 55.56%, Block Peak % of device VRAM: 62.22%, ΔTime: 00:00:33 -[2025-11-10 05:20:55,478][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:20:55,481][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:20:55,483][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:20:56,311][__main__][INFO] - Iteration 689 took 1m 1s (40.87% Gen, 57.79% Train). Generation: 25s, Training: 35s. Estimated remaining time: 40h 12m 30s. Estimated total time: 51h 34m 29s. Time estimates for 10 more iterations: 10m 18s, 100 more iterations: 1h 43m 8s, 500 more iterations: 8h 35m 44s. -[2025-11-10 05:20:56,315][__main__][INFO] - Starting iteration 689. -[2025-11-10 05:20:56,787][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 68 and human policies 1. -[2025-11-10 05:20:56,788][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:21:26,229][__main__][INFO] - Number of regex retries in iteration 689: 0 -[2025-11-10 05:21:26,229][__main__][INFO] - agents played in iteration 689 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:21:27,100][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:21:27,127][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:21:27,154][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:21:27,177][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:21:27,177][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:21:27,179][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:21:27,975][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:21:28,431][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:21:28,934][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:21:29,440][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:21:29,938][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:21:30,443][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:21:30,943][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:21:31,442][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:21:31,939][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:21:32,445][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:21:32,945][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:21:33,444][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:21:33,939][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:21:34,435][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:21:34,932][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:21:35,429][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:21:35,928][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:21:36,432][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:21:36,930][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:21:37,435][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:21:37,934][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:21:38,434][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:21:38,932][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:21:39,431][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:21:39,932][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:21:40,430][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:21:40,928][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:21:41,427][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:21:41,923][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:21:42,419][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:21:42,915][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:21:43,412][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:21:43,910][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:21:44,409][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:21:44,907][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:21:45,405][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:21:45,902][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:21:46,401][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:21:46,903][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:21:47,406][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:21:47,908][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:21:48,409][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:21:48,910][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:21:49,412][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:21:49,916][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:21:50,419][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:21:50,919][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:21:51,422][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:21:51,925][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:21:52,428][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:21:52,931][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:21:53,435][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:21:53,937][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:21:54,444][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:21:54,945][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:21:55,445][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:21:55,952][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:21:56,457][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:21:56,966][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:21:57,472][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:21:57,977][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:21:58,503][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:21:59,004][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:21:59,505][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:22:00,013][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10045 tokens. -[2025-11-10 05:22:00,971][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 55.61%, Block Peak % of device VRAM: 62.19%, ΔTime: 00:00:33 -[2025-11-10 05:22:01,745][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:22:01,747][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:22:01,750][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:22:02,743][__main__][INFO] - Iteration 690 took 1m 5s (44.64% Gen, 53.85% Train). Generation: 29s, Training: 35s. Estimated remaining time: 43h 34m 44s. Estimated total time: 54h 57m 50s. Time estimates for 10 more iterations: 10m 59s, 100 more iterations: 1h 49m 55s, 500 more iterations: 9h 9m 38s. -[2025-11-10 05:22:02,745][__main__][INFO] - Starting iteration 690. -[2025-11-10 05:22:03,234][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 68 and human policies 1. -[2025-11-10 05:22:03,235][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:22:28,943][__main__][INFO] - Number of regex retries in iteration 690: 0 -[2025-11-10 05:22:28,943][__main__][INFO] - agents played in iteration 690 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:22:29,722][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:22:29,750][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:22:29,778][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:22:29,810][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:22:29,812][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:22:29,813][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:22:30,621][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:22:31,075][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:22:31,579][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:22:32,083][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:22:32,591][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:22:33,088][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:22:33,586][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:22:34,089][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:22:34,591][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:22:35,092][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:22:35,588][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:22:36,086][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:22:36,597][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:22:37,091][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:22:37,585][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:22:38,087][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:22:38,580][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:22:39,089][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:22:39,589][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:22:40,087][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:22:40,586][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:22:41,088][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:22:41,589][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:22:42,087][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:22:42,582][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:22:43,091][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:22:43,588][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:22:44,085][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:22:44,581][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:22:45,077][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:22:45,586][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:22:46,086][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:22:46,584][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:22:47,082][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:22:47,577][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:22:48,073][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:22:48,571][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:22:49,066][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:22:49,574][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:22:50,070][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:22:50,567][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:22:51,066][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:22:51,566][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:22:52,069][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:22:52,592][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:22:53,097][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:22:53,610][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:22:54,109][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:22:54,615][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:22:55,115][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:22:55,615][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:22:56,118][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:22:56,621][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:22:57,124][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:22:57,628][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:22:58,135][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:22:58,635][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:22:59,138][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:23:00,459][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:23:01,169][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:23:01,676][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:23:02,177][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:23:02,692][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:23:03,191][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:23:03,694][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10157 tokens. -[2025-11-10 05:23:04,687][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 55.57%, Block Peak % of device VRAM: 62.21%, ΔTime: 00:00:34 -[2025-11-10 05:23:05,348][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:23:05,350][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:23:05,352][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:23:07,016][__main__][INFO] - Iteration 691 took 1m 3s (40.31% Gen, 57.08% Train). Generation: 25s, Training: 36s. Estimated remaining time: 41h 44m 58s. Estimated total time: 53h 9m 8s. Time estimates for 10 more iterations: 10m 37s, 100 more iterations: 1h 46m 18s, 500 more iterations: 8h 51m 31s. -[2025-11-10 05:23:07,019][__main__][INFO] - Starting iteration 691. -[2025-11-10 05:23:07,492][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 69 and human policies 1. -[2025-11-10 05:23:07,493][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:23:32,805][__main__][INFO] - Number of regex retries in iteration 691: 0 -[2025-11-10 05:23:32,805][__main__][INFO] - agents played in iteration 691 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:23:33,684][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:23:33,713][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:23:33,740][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:23:33,763][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:23:33,764][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:23:33,765][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:23:34,613][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:23:35,066][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:23:35,581][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:23:36,082][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:23:36,585][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:23:37,104][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:23:37,601][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:23:38,100][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:23:38,598][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:23:39,096][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:23:39,610][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:23:40,107][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:23:40,603][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:23:41,104][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:23:41,604][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:23:42,115][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:23:42,614][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:23:43,113][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:23:43,615][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:23:44,124][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:23:44,630][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:23:45,132][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:23:45,631][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:23:46,134][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:23:46,633][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:23:47,134][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:23:47,629][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:23:48,128][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:23:48,631][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:23:49,131][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:23:49,628][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:23:50,126][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:23:50,625][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:23:51,121][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:23:51,624][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:23:52,119][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:23:52,616][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:23:53,112][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:23:53,613][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:23:54,119][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:23:54,616][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:23:55,113][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:23:55,615][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:23:56,112][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:23:56,616][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:23:57,116][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:23:57,620][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:23:58,128][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:23:58,634][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:23:59,139][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:23:59,639][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:24:00,144][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:24:00,648][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:24:01,148][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:24:01,653][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:24:02,160][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:24:02,661][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:24:03,169][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:24:03,671][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:24:04,180][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:24:04,683][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:24:05,190][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:24:05,691][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:24:06,197][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:24:06,699][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10003 tokens. -[2025-11-10 05:24:07,665][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.03%, Current % of VRAM taken: 55.53%, Block Peak % of device VRAM: 62.17%, ΔTime: 00:00:33 -[2025-11-10 05:24:08,430][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:24:08,432][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:24:08,434][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:24:09,509][__main__][INFO] - Iteration 692 took 1m 2s (40.81% Gen, 57.45% Train). Generation: 25s, Training: 35s. Estimated remaining time: 40h 15m 40s. Estimated total time: 51h 40m 52s. Time estimates for 10 more iterations: 10m 20s, 100 more iterations: 1h 43m 21s, 500 more iterations: 8h 36m 48s. -[2025-11-10 05:24:09,511][__main__][INFO] - Starting iteration 692. -[2025-11-10 05:24:09,977][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 69 and human policies 1. -[2025-11-10 05:24:09,978][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:24:37,479][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 05:24:38,335][__main__][INFO] - Number of regex retries in iteration 692: 1 -[2025-11-10 05:24:38,336][__main__][INFO] - agents played in iteration 692 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:24:39,130][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:24:39,157][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:24:39,183][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:24:39,206][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:24:39,207][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:24:39,208][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:24:40,013][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:24:40,471][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:24:40,974][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:24:41,474][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:24:41,977][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:24:42,476][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:24:42,992][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:24:43,489][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:24:43,990][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:24:44,493][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:24:44,993][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:24:45,494][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:24:45,995][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:24:46,498][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:24:46,998][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:24:47,499][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:24:47,997][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:24:48,499][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:24:48,996][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:24:49,514][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:24:50,012][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:24:50,510][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:24:51,008][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:24:51,507][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:24:52,004][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:24:52,498][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:24:52,994][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:24:53,502][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:24:53,996][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:24:54,492][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:24:54,994][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:24:55,491][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:24:55,986][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:24:56,484][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:24:56,978][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:24:57,477][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:24:57,972][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:24:58,467][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:24:58,963][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:24:59,458][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:24:59,958][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:25:00,456][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:25:00,955][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:25:01,455][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:25:01,958][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:25:02,454][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:25:02,958][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:25:03,457][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:25:03,960][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:25:04,463][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:25:05,115][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:25:05,622][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:25:06,122][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:25:06,711][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:25:07,229][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:25:07,733][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:25:08,233][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:25:08,740][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:25:09,245][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:25:09,746][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:25:10,248][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:25:10,752][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:25:11,262][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:25:11,765][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:25:12,271][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10172 tokens. -[2025-11-10 05:25:13,234][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.25%, Current % of VRAM taken: 55.75%, Block Peak % of device VRAM: 62.16%, ΔTime: 00:00:33 -[2025-11-10 05:25:13,894][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:25:13,896][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:25:13,898][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:25:14,791][__main__][INFO] - Iteration 693 took 1m 4s (43.75% Gen, 54.87% Train). Generation: 28s, Training: 35s. Estimated remaining time: 42h 34m 25s. Estimated total time: 54h 0m 43s. Time estimates for 10 more iterations: 10m 48s, 100 more iterations: 1h 48m 1s, 500 more iterations: 9h 0m 7s. -[2025-11-10 05:25:14,795][__main__][INFO] - Starting iteration 693. -[2025-11-10 05:25:15,261][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 69 and human policies 1. -[2025-11-10 05:25:15,262][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:25:46,673][__main__][INFO] - Number of regex retries in iteration 693: 0 -[2025-11-10 05:25:46,674][__main__][INFO] - agents played in iteration 693 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:25:47,445][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:25:47,473][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:25:47,500][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:25:47,523][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:25:47,524][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:25:47,525][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:25:48,398][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:25:48,855][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:25:49,362][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:25:49,862][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:25:50,359][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:25:50,856][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:25:51,356][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:25:51,857][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:25:52,357][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:25:52,857][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:25:53,357][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:25:53,855][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:25:54,352][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:25:54,855][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:25:55,354][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:25:55,853][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:25:56,350][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:25:56,846][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:25:57,346][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:25:57,854][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:25:58,351][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:25:58,851][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:25:59,351][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:25:59,850][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:26:00,345][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:26:00,841][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:26:01,341][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:26:01,845][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:26:02,341][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:26:02,839][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:26:03,343][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:26:03,840][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:26:04,344][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:26:04,865][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:26:05,365][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:26:05,866][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:26:06,363][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:26:06,863][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:26:07,364][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:26:07,867][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:26:08,370][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:26:08,872][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:26:09,375][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:26:09,881][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:26:10,381][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:26:10,881][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:26:11,386][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:26:11,887][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:26:12,400][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:26:12,904][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:26:13,411][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:26:13,915][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:26:14,419][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:26:14,920][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:26:15,427][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:26:15,933][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:26:16,444][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:26:16,954][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:26:17,459][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:26:17,966][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:26:18,467][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:26:18,976][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:26:19,478][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:26:19,981][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:26:20,488][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10169 tokens. -[2025-11-10 05:26:21,477][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 55.64%, Block Peak % of device VRAM: 62.24%, ΔTime: 00:00:33 -[2025-11-10 05:26:22,223][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:26:22,225][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:26:22,228][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:26:23,246][__main__][INFO] - Iteration 694 took 1m 7s (46.20% Gen, 52.30% Train). Generation: 31s, Training: 35s. Estimated remaining time: 45h 11m 49s. Estimated total time: 56h 39m 15s. Time estimates for 10 more iterations: 11m 19s, 100 more iterations: 1h 53m 18s, 500 more iterations: 9h 26m 32s. -[2025-11-10 05:26:23,248][__main__][INFO] - Starting iteration 694. -[2025-11-10 05:26:23,724][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 69 and human policies 1. -[2025-11-10 05:26:23,725][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:26:46,503][__main__][INFO] - Number of regex retries in iteration 694: 0 -[2025-11-10 05:26:46,504][__main__][INFO] - agents played in iteration 694 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:26:47,376][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:26:47,403][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:26:47,430][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:26:47,453][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:26:47,453][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:26:47,455][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:26:48,310][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:26:48,765][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:26:49,267][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:26:49,764][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:26:50,262][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:26:50,758][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:26:51,254][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:26:51,758][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:26:52,254][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:26:52,757][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:26:53,256][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:26:53,752][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:26:54,261][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:26:54,757][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:26:55,256][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:26:55,753][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:26:56,250][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:26:56,752][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:26:57,249][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:26:57,747][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:26:58,246][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:26:58,744][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:26:59,242][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:26:59,738][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:27:00,236][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:27:00,749][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:27:01,246][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:27:01,741][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:27:02,255][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:27:02,752][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:27:03,252][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:27:03,748][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:27:04,244][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:27:04,754][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:27:05,255][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:27:05,751][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:27:06,251][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:27:06,751][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:27:07,255][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:27:07,751][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:27:08,248][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:27:08,745][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:27:09,243][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:27:09,742][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:27:10,240][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:27:10,738][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:27:11,236][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:27:11,731][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:27:12,232][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:27:12,737][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:27:13,270][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:27:13,780][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:27:14,281][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:27:14,781][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:27:15,287][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:27:15,788][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:27:16,290][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:27:16,798][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:27:17,303][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:27:17,810][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:27:18,313][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:27:18,816][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:27:19,320][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:27:19,824][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:27:20,346][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9988 tokens. -[2025-11-10 05:27:21,315][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 55.66%, Block Peak % of device VRAM: 62.08%, ΔTime: 00:00:33 -[2025-11-10 05:27:22,066][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:27:22,068][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:27:22,070][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:27:23,035][__main__][INFO] - Iteration 695 took 59s (38.40% Gen, 59.96% Train). Generation: 22s, Training: 35s. Estimated remaining time: 37h 57m 10s. Estimated total time: 49h 25m 36s. Time estimates for 10 more iterations: 9m 53s, 100 more iterations: 1h 38m 51s, 500 more iterations: 8h 14m 16s. -[2025-11-10 05:27:23,039][__main__][INFO] - Starting iteration 695. -[2025-11-10 05:27:23,504][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 69 and human policies 1. -[2025-11-10 05:27:23,504][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:27:55,244][__main__][INFO] - Number of regex retries in iteration 695: 0 -[2025-11-10 05:27:55,246][__main__][INFO] - agents played in iteration 695 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:27:56,021][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:27:56,044][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:27:56,068][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:27:56,090][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:27:56,091][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:27:56,092][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:27:56,900][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:27:57,373][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:27:57,873][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:27:58,372][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:27:58,870][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:27:59,372][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:27:59,874][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:28:00,371][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:28:00,870][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:28:01,364][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:28:01,861][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:28:02,369][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:28:02,866][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:28:03,363][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:28:03,869][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:28:04,371][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:28:04,870][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:28:05,368][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:28:05,864][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:28:06,365][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:28:06,869][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:28:07,366][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:28:07,871][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:28:08,369][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:28:08,869][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:28:09,368][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:28:09,864][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:28:10,365][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:28:10,860][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:28:11,359][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:28:11,871][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:28:12,373][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:28:12,883][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:28:13,384][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:28:13,883][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:28:14,389][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:28:14,894][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:28:15,395][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:28:15,897][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:28:16,402][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:28:16,912][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:28:17,415][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:28:17,924][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:28:18,430][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:28:18,935][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:28:19,440][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:28:19,943][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:28:20,446][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:28:20,955][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:28:21,458][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:28:21,967][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:28:22,469][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:28:22,975][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:28:23,485][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:28:23,991][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:28:24,493][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:28:24,998][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:28:25,500][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:28:26,011][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:28:26,512][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:28:27,014][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:28:27,517][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:28:28,019][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:28:28,521][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:28:29,022][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10061 tokens. -[2025-11-10 05:28:29,983][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.04%, Current % of VRAM taken: 55.53%, Block Peak % of device VRAM: 62.07%, ΔTime: 00:00:33 -[2025-11-10 05:28:30,632][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:28:30,634][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:28:30,636][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:28:31,491][__main__][INFO] - Iteration 696 took 1m 7s (46.69% Gen, 52.05% Train). Generation: 31s, Training: 35s. Estimated remaining time: 45h 9m 50s. Estimated total time: 56h 39m 24s. Time estimates for 10 more iterations: 11m 19s, 100 more iterations: 1h 53m 18s, 500 more iterations: 9h 26m 34s. -[2025-11-10 05:28:31,494][__main__][INFO] - Starting iteration 696. -[2025-11-10 05:28:31,954][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 69 and human policies 1. -[2025-11-10 05:28:31,955][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:28:56,908][__main__][INFO] - Number of regex retries in iteration 696: 0 -[2025-11-10 05:28:56,909][__main__][INFO] - agents played in iteration 696 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:28:57,682][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:28:57,710][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:28:57,737][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:28:57,760][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:28:57,760][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:28:57,761][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:28:58,565][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:28:59,024][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:28:59,525][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:29:00,020][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:29:00,519][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:29:01,017][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:29:01,512][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:29:02,012][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:29:02,510][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:29:03,022][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:29:03,524][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:29:04,023][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:29:04,523][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:29:05,021][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:29:05,522][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:29:06,020][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:29:06,522][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:29:07,019][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:29:07,522][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:29:08,024][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:29:08,521][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:29:09,027][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:29:09,526][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:29:10,024][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:29:10,521][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:29:11,019][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:29:11,522][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:29:12,023][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:29:12,527][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:29:13,027][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:29:13,528][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:29:14,030][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:29:14,534][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:29:15,029][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:29:15,523][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:29:16,022][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:29:16,517][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:29:17,014][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:29:17,516][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:29:18,012][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:29:18,510][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:29:19,006][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:29:19,505][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:29:20,012][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:29:20,515][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:29:21,014][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:29:21,515][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:29:22,015][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:29:22,515][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:29:23,019][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:29:23,519][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:29:24,023][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:29:24,530][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:29:25,033][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:29:25,538][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:29:26,042][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:29:26,550][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:29:27,053][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:29:27,557][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:29:28,070][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:29:28,584][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:29:29,095][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:29:29,597][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:29:30,097][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:29:30,603][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10059 tokens. -[2025-11-10 05:29:31,592][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.01%, Current % of VRAM taken: 55.50%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:33 -[2025-11-10 05:29:32,318][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:29:32,321][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:29:32,322][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:29:34,561][__main__][INFO] - Iteration 697 took 1m 2s (39.86% Gen, 56.56% Train). Generation: 24s, Training: 35s. Estimated remaining time: 40h 39m 44s. Estimated total time: 52h 10m 22s. Time estimates for 10 more iterations: 10m 26s, 100 more iterations: 1h 44m 20s, 500 more iterations: 8h 41m 43s. -[2025-11-10 05:29:34,570][__main__][INFO] - Starting iteration 697. -[2025-11-10 05:29:35,045][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 69 and human policies 1. -[2025-11-10 05:29:35,046][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:29:58,998][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 05:30:07,018][__main__][INFO] - Number of regex retries in iteration 697: 1 -[2025-11-10 05:30:07,019][__main__][INFO] - agents played in iteration 697 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:30:07,789][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:30:07,812][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:30:07,835][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:30:07,857][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:30:07,857][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:30:07,859][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:30:08,668][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:30:09,127][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:30:09,631][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:30:10,136][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:30:10,634][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:30:11,137][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:30:11,639][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:30:12,140][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:30:12,640][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:30:13,140][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:30:13,638][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:30:14,141][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:30:14,644][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:30:15,142][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:30:15,640][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:30:16,134][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:30:16,636][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:30:17,134][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:30:17,633][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:30:18,132][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:30:18,634][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:30:19,138][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:30:19,639][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:30:20,139][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:30:20,641][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:30:21,140][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:30:21,642][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:30:22,147][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:30:22,647][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:30:23,153][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:30:23,655][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:30:24,159][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:30:24,659][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:30:25,161][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:30:25,664][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:30:26,162][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:30:26,663][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:30:27,200][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:30:27,699][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:30:28,198][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:30:28,698][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:30:29,201][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:30:29,718][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:30:30,225][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:30:30,729][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:30:31,236][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:30:31,741][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:30:32,247][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:30:32,751][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:30:33,252][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:30:33,756][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:30:34,258][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:30:34,759][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:30:35,270][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:30:35,773][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:30:36,300][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:30:36,799][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:30:37,301][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:30:37,806][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:30:38,308][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:30:38,814][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:30:39,314][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:30:39,816][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:30:40,324][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:30:40,826][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10060 tokens. -[2025-11-10 05:30:41,796][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 55.61%, Block Peak % of device VRAM: 62.33%, ΔTime: 00:00:33 -[2025-11-10 05:30:42,419][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:30:42,420][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:30:42,422][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:30:43,522][__main__][INFO] - Iteration 698 took 1m 8s (46.69% Gen, 51.70% Train). Generation: 31s, Training: 35s. Estimated remaining time: 45h 32m 7s. Estimated total time: 57h 3m 54s. Time estimates for 10 more iterations: 11m 24s, 100 more iterations: 1h 54m 7s, 500 more iterations: 9h 30m 39s. -[2025-11-10 05:30:43,524][__main__][INFO] - Starting iteration 698. -[2025-11-10 05:30:44,018][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 69 and human policies 1. -[2025-11-10 05:30:44,019][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:31:08,986][__main__][INFO] - Number of regex retries in iteration 698: 0 -[2025-11-10 05:31:08,988][__main__][INFO] - agents played in iteration 698 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:31:09,756][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:31:09,783][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:31:09,809][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:31:09,832][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:31:09,832][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:31:09,834][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:31:10,661][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:31:11,115][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:31:11,615][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:31:12,126][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:31:12,629][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:31:13,129][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:31:13,623][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:31:14,122][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:31:14,630][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:31:15,127][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:31:15,635][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:31:16,135][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:31:16,639][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:31:17,172][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:31:17,668][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:31:18,164][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:31:18,662][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:31:19,159][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:31:19,658][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:31:20,154][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:31:20,653][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:31:21,156][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:31:21,655][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:31:22,157][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:31:22,662][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:31:23,161][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:31:23,660][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:31:24,162][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:31:24,660][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:31:25,161][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:31:25,661][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:31:26,165][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:31:26,667][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:31:27,172][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:31:27,675][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:31:28,178][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:31:28,680][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:31:29,188][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:31:29,688][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:31:30,191][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:31:30,691][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:31:31,202][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:31:31,710][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:31:32,215][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:31:32,719][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:31:33,235][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:31:33,738][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:31:34,245][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:31:34,750][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:31:35,262][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:31:35,768][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:31:36,272][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:31:36,774][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:31:37,277][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:31:37,782][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:31:38,295][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:31:38,797][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:31:39,299][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:31:39,822][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:31:40,323][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:31:40,825][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:31:41,326][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:31:41,829][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:31:42,336][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:31:42,839][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10187 tokens. -[2025-11-10 05:31:43,860][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 55.58%, Block Peak % of device VRAM: 62.40%, ΔTime: 00:00:33 -[2025-11-10 05:31:46,289][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:31:46,291][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:31:46,293][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:31:47,239][__main__][INFO] - Iteration 699 took 1m 3s (39.49% Gen, 59.01% Train). Generation: 24s, Training: 37s. Estimated remaining time: 41h 8m 15s. Estimated total time: 52h 41m 5s. Time estimates for 10 more iterations: 10m 32s, 100 more iterations: 1h 45m 22s, 500 more iterations: 8h 46m 50s. -[2025-11-10 05:31:47,243][__main__][INFO] - Starting iteration 699. -[2025-11-10 05:31:47,719][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 69 and human policies 1. -[2025-11-10 05:31:47,720][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:32:17,549][__main__][INFO] - Number of regex retries in iteration 699: 0 -[2025-11-10 05:32:17,550][__main__][INFO] - agents played in iteration 699 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:32:18,318][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:32:18,346][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:32:18,372][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:32:18,395][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:32:18,395][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:32:18,396][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:32:19,178][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:32:19,633][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:32:20,140][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:32:20,635][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:32:21,133][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:32:21,637][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:32:22,133][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:32:22,629][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:32:23,136][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:32:23,638][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:32:24,136][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:32:24,635][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:32:25,132][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:32:25,633][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:32:26,136][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:32:26,644][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:32:27,150][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:32:27,652][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:32:28,152][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:32:28,654][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:32:29,155][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:32:29,660][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:32:30,162][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:32:30,661][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:32:31,168][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:32:31,673][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:32:32,176][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:32:32,679][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:32:33,182][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:32:33,683][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:32:34,188][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:32:34,692][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:32:35,195][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:32:35,696][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:32:36,201][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:32:36,702][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:32:37,210][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:32:37,719][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:32:38,220][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:32:38,725][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:32:39,226][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:32:39,729][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:32:40,244][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:32:40,744][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:32:41,248][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:32:41,747][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:32:42,250][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:32:42,749][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:32:43,252][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:32:43,757][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:32:44,259][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:32:44,760][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:32:45,262][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:32:45,766][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:32:46,266][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:32:46,767][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:32:47,270][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:32:47,773][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:32:48,279][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:32:48,779][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:32:49,284][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:32:49,783][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:32:50,284][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:32:50,787][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:32:51,288][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10074 tokens. -[2025-11-10 05:32:52,242][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 55.64%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:33 -[2025-11-10 05:32:53,023][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:32:53,024][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:32:53,026][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:32:53,965][__main__][INFO] - Iteration 700 took 1m 6s (45.03% Gen, 53.55% Train). Generation: 29s, Training: 35s. Estimated remaining time: 43h 38m 22s. Estimated total time: 55h 12m 19s. Time estimates for 10 more iterations: 11m 2s, 100 more iterations: 1h 50m 24s, 500 more iterations: 9h 12m 3s. -[2025-11-10 05:32:53,967][__main__][INFO] - Starting iteration 700. -[2025-11-10 05:32:54,433][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 69 and human policies 1. -[2025-11-10 05:32:54,433][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:33:15,430][__main__][INFO] - Number of regex retries in iteration 700: 0 -[2025-11-10 05:33:15,430][__main__][INFO] - agents played in iteration 700 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:33:16,272][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:33:16,295][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:33:16,318][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:33:16,341][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:33:16,341][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:33:16,342][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:33:17,167][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:33:17,621][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:33:18,130][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:33:18,634][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:33:19,131][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:33:19,645][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:33:20,144][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:33:20,643][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:33:21,138][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:33:21,632][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:33:22,139][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:33:22,636][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:33:23,138][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:33:23,650][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:33:24,147][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:33:24,642][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:33:25,137][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:33:25,636][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:33:26,145][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:33:26,646][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:33:27,145][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:33:27,650][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:33:28,150][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:33:28,650][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:33:29,153][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:33:29,649][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:33:30,145][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:33:30,645][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:33:31,145][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:33:31,643][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:33:32,139][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:33:32,640][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:33:33,137][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:33:33,635][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:33:34,157][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:33:34,656][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:33:35,162][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:33:35,664][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:33:36,163][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:33:36,671][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:33:37,180][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:33:37,685][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:33:38,190][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:33:38,696][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:33:39,203][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:33:39,709][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:33:40,215][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:33:40,726][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:33:41,228][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:33:41,739][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:33:42,247][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:33:42,754][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:33:43,265][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:33:43,772][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:33:44,278][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:33:44,787][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:33:45,295][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:33:45,800][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:33:46,303][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:33:46,808][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:33:47,315][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:33:47,820][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:33:48,324][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:33:48,855][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:33:49,360][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10083 tokens. -[2025-11-10 05:33:50,309][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 55.68%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:33 -[2025-11-10 05:33:51,128][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:33:51,130][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:33:51,132][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:33:53,903][__main__][INFO] - Iteration 701 took 59s (35.30% Gen, 60.03% Train). Generation: 20s, Training: 35s. Estimated remaining time: 37h 58m 37s. Estimated total time: 49h 33m 34s. Time estimates for 10 more iterations: 9m 54s, 100 more iterations: 1h 39m 7s, 500 more iterations: 8h 15m 35s. -[2025-11-10 05:33:53,908][__main__][INFO] - Starting iteration 701. -[2025-11-10 05:33:54,380][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 70 and human policies 1. -[2025-11-10 05:33:54,381][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:34:24,293][__main__][INFO] - Number of regex retries in iteration 701: 0 -[2025-11-10 05:34:24,294][__main__][INFO] - agents played in iteration 701 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:34:25,086][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:34:25,115][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:34:25,144][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:34:25,168][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:34:25,168][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:34:25,169][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:34:25,962][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:34:26,414][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:34:26,916][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:34:27,413][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:34:27,911][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:34:28,407][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:34:28,902][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:34:29,400][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:34:29,895][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:34:30,391][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:34:30,889][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:34:31,382][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:34:31,877][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:34:32,374][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:34:32,875][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:34:33,372][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:34:33,867][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:34:34,363][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:34:34,862][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:34:35,358][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:34:35,854][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:34:36,353][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:34:36,853][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:34:37,356][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:34:37,854][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:34:38,353][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:34:38,853][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:34:39,356][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:34:39,863][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:34:40,369][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:34:40,876][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:34:41,379][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:34:41,887][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:34:42,394][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:34:42,896][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:34:43,399][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:34:43,900][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:34:44,401][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:34:44,902][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:34:45,409][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:34:45,912][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:34:46,418][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:34:46,927][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:34:47,437][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:34:47,940][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:34:48,442][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:34:48,949][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:34:49,453][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:34:49,958][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:34:50,488][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:34:50,998][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:34:51,503][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:34:52,008][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:34:52,512][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:34:53,026][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:34:53,532][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:34:54,068][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:34:54,571][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:34:55,075][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:34:55,577][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:34:56,078][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:34:56,583][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:34:57,085][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:34:57,587][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:34:58,097][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10093 tokens. -[2025-11-10 05:34:59,089][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.09%, Current % of VRAM taken: 55.58%, Block Peak % of device VRAM: 62.35%, ΔTime: 00:00:33 -[2025-11-10 05:34:59,756][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:34:59,758][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:34:59,760][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:35:00,863][__main__][INFO] - Iteration 702 took 1m 6s (44.99% Gen, 53.35% Train). Generation: 29s, Training: 35s. Estimated remaining time: 43h 48m 6s. Estimated total time: 55h 24m 10s. Time estimates for 10 more iterations: 11m 4s, 100 more iterations: 1h 50m 48s, 500 more iterations: 9h 14m 1s. -[2025-11-10 05:35:00,866][__main__][INFO] - Starting iteration 702. -[2025-11-10 05:35:01,360][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 70 and human policies 1. -[2025-11-10 05:35:01,360][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:35:24,455][__main__][INFO] - Number of regex retries in iteration 702: 0 -[2025-11-10 05:35:24,455][__main__][INFO] - agents played in iteration 702 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:35:25,319][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:35:25,342][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:35:25,364][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:35:25,386][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:35:25,387][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:35:25,388][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:35:26,193][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:35:26,660][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:35:27,164][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:35:27,661][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:35:28,174][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:35:28,676][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:35:29,171][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:35:29,669][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:35:30,165][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:35:30,676][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:35:31,171][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:35:31,665][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:35:32,162][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:35:32,658][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:35:33,157][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:35:33,653][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:35:34,152][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:35:34,661][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:35:35,156][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:35:35,657][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:35:36,153][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:35:36,650][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:35:37,154][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:35:37,649][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:35:38,144][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:35:38,653][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:35:39,150][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:35:39,646][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:35:40,141][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:35:40,637][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:35:41,136][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:35:41,632][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:35:42,127][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:35:42,633][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:35:43,131][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:35:43,630][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:35:44,133][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:35:44,632][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:35:45,131][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:35:45,630][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:35:46,133][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:35:46,634][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:35:47,133][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:35:47,637][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:35:48,138][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:35:48,639][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:35:49,140][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:35:49,642][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:35:50,145][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:35:50,648][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:35:51,151][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:35:51,667][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:35:52,168][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:35:52,694][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:35:53,196][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:35:53,698][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:35:54,200][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:35:54,702][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:35:55,208][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:35:55,711][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:35:56,214][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:35:56,717][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:35:57,220][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:35:57,725][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:35:58,225][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10088 tokens. -[2025-11-10 05:35:59,206][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 55.56%, Block Peak % of device VRAM: 62.31%, ΔTime: 00:00:33 -[2025-11-10 05:35:59,956][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:35:59,957][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:35:59,959][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:36:00,988][__main__][INFO] - Iteration 703 took 59s (38.73% Gen, 59.54% Train). Generation: 23s, Training: 35s. Estimated remaining time: 38h 4m 21s. Estimated total time: 49h 41m 25s. Time estimates for 10 more iterations: 9m 56s, 100 more iterations: 1h 39m 22s, 500 more iterations: 8h 16m 54s. -[2025-11-10 05:36:00,995][__main__][INFO] - Starting iteration 703. -[2025-11-10 05:36:01,465][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 70 and human policies 1. -[2025-11-10 05:36:01,466][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:36:31,445][__main__][INFO] - Number of regex retries in iteration 703: 0 -[2025-11-10 05:36:31,447][__main__][INFO] - agents played in iteration 703 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:36:32,224][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:36:32,251][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:36:32,278][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:36:32,301][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:36:32,302][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:36:32,303][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:36:33,094][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:36:33,547][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:36:34,054][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:36:34,551][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:36:35,053][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:36:35,556][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:36:36,051][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:36:36,548][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:36:37,045][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:36:37,540][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:36:38,034][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:36:38,529][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:36:39,025][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:36:39,526][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:36:40,023][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:36:40,522][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:36:41,018][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:36:41,523][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:36:42,018][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:36:42,515][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:36:43,015][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:36:43,511][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:36:44,007][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:36:44,506][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:36:45,002][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:36:45,496][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:36:45,994][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:36:46,490][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:36:46,985][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:36:47,490][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:36:47,987][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:36:48,486][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:36:48,991][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:36:49,492][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:36:49,998][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:36:50,497][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:36:51,000][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:36:51,500][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:36:52,007][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:36:52,511][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:36:53,014][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:36:53,526][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:36:54,027][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:36:54,528][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:36:55,035][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:36:55,537][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:36:56,042][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:36:56,544][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:36:57,046][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:36:57,555][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:36:58,059][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:36:58,562][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:36:59,067][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:36:59,570][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:37:00,077][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:37:00,580][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:37:01,088][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:37:01,597][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:37:02,102][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:37:02,606][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:37:03,109][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:37:03,611][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:37:04,115][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:37:04,619][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:37:05,123][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10110 tokens. -[2025-11-10 05:37:06,115][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.05%, Current % of VRAM taken: 55.54%, Block Peak % of device VRAM: 62.16%, ΔTime: 00:00:33 -[2025-11-10 05:37:06,780][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:37:06,782][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:37:06,784][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:37:07,634][__main__][INFO] - Iteration 704 took 1m 6s (45.31% Gen, 53.41% Train). Generation: 29s, Training: 35s. Estimated remaining time: 43h 30m 16s. Estimated total time: 55h 8m 26s. Time estimates for 10 more iterations: 11m 1s, 100 more iterations: 1h 50m 16s, 500 more iterations: 9h 11m 24s. -[2025-11-10 05:37:07,636][__main__][INFO] - Starting iteration 704. -[2025-11-10 05:37:08,100][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 70 and human policies 1. -[2025-11-10 05:37:08,100][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:37:20,839][mllm.models.large_language_model_local][WARNING] - Response Proposal: x hats, y books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 05:37:31,865][__main__][INFO] - Number of regex retries in iteration 704: 1 -[2025-11-10 05:37:31,865][__main__][INFO] - agents played in iteration 704 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:37:32,628][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:37:32,656][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:37:32,682][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:37:32,706][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.55%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:37:32,706][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:37:32,708][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:37:33,490][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:37:33,946][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:37:34,448][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:37:34,944][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:37:35,441][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:37:35,936][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:37:36,432][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:37:36,931][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:37:37,425][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:37:37,925][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:37:38,418][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:37:38,912][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:37:39,413][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:37:39,906][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:37:40,404][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:37:40,899][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:37:41,394][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:37:41,896][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:37:42,394][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:37:42,892][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:37:43,402][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:37:43,899][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:37:44,398][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:37:44,895][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:37:45,391][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:37:45,885][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:37:46,378][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:37:46,874][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:37:47,381][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:37:47,875][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:37:48,372][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:37:48,882][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:37:49,376][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:37:49,878][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:37:50,372][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:37:50,867][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:37:51,364][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:37:51,862][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:37:52,355][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:37:52,854][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:37:53,350][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:37:53,850][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:37:54,347][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:37:54,842][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:37:55,339][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:37:55,834][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:37:56,330][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:37:56,827][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:37:57,328][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:37:57,827][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:37:58,326][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:37:58,829][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:37:59,330][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:37:59,835][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:38:00,338][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:38:00,838][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:38:01,339][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:38:01,842][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:38:02,347][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:38:02,850][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:38:03,353][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:38:03,857][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:38:04,361][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:38:04,864][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:38:05,366][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10071 tokens. -[2025-11-10 05:38:06,338][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 55.58%, Block Peak % of device VRAM: 62.13%, ΔTime: 00:00:32 -[2025-11-10 05:38:07,085][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:38:07,087][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:38:07,090][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:38:08,051][__main__][INFO] - Iteration 705 took 59s (39.64% Gen, 58.75% Train). Generation: 23s, Training: 35s. Estimated remaining time: 38h 18m 26s. Estimated total time: 49h 57m 37s. Time estimates for 10 more iterations: 9m 59s, 100 more iterations: 1h 39m 55s, 500 more iterations: 8h 19m 36s. -[2025-11-10 05:38:08,054][__main__][INFO] - Starting iteration 705. -[2025-11-10 05:38:08,517][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 70 and human policies 1. -[2025-11-10 05:38:08,517][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:38:28,819][__main__][INFO] - Number of regex retries in iteration 705: 0 -[2025-11-10 05:38:28,822][__main__][INFO] - agents played in iteration 705 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:38:29,665][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:38:29,692][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:38:29,719][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:38:29,742][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:38:29,743][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:38:29,744][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:38:30,581][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:38:31,040][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:38:31,544][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:38:32,053][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:38:32,552][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:38:33,053][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:38:33,550][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:38:34,046][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:38:34,543][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:38:35,043][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:38:35,539][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:38:36,035][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:38:36,534][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:38:37,032][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:38:37,530][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:38:38,031][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:38:38,528][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:38:39,024][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:38:39,525][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:38:40,020][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:38:40,516][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:38:41,011][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:38:41,505][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:38:42,000][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:38:42,501][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:38:42,997][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:38:43,496][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:38:43,995][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:38:44,493][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:38:44,999][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:38:45,496][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:38:45,993][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:38:46,491][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:38:46,989][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:38:47,494][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:38:47,995][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:38:48,494][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:38:49,002][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:38:49,501][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:38:50,013][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:38:50,512][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:38:51,006][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:38:51,505][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:38:52,000][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:38:52,501][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:38:52,996][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:38:53,492][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:38:53,986][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:38:54,480][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:38:54,974][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:38:55,469][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:38:55,966][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:38:56,466][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:38:56,965][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:38:57,464][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:38:57,963][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:38:58,467][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:38:58,973][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:38:59,477][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:38:59,977][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:39:00,481][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:39:00,987][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:39:01,492][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:39:02,003][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:39:02,504][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10058 tokens. -[2025-11-10 05:39:03,554][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.09%, Current % of VRAM taken: 55.59%, Block Peak % of device VRAM: 62.23%, ΔTime: 00:00:32 -[2025-11-10 05:39:04,200][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:39:04,202][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:39:04,204][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:39:05,020][__main__][INFO] - Iteration 706 took 56s (35.93% Gen, 62.62% Train). Generation: 20s, Training: 35s. Estimated remaining time: 35h 25m 2s. Estimated total time: 47h 5m 10s. Time estimates for 10 more iterations: 9m 25s, 100 more iterations: 1h 34m 10s, 500 more iterations: 7h 50m 51s. -[2025-11-10 05:39:05,022][__main__][INFO] - Starting iteration 706. -[2025-11-10 05:39:05,483][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 70 and human policies 1. -[2025-11-10 05:39:05,484][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:39:22,728][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 05:39:34,326][__main__][INFO] - Number of regex retries in iteration 706: 1 -[2025-11-10 05:39:34,327][__main__][INFO] - agents played in iteration 706 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:39:35,128][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:39:35,157][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:39:35,183][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:39:35,206][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.57%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:39:35,206][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:39:35,207][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:39:36,009][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:39:36,462][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:39:36,961][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:39:37,457][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:39:37,957][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:39:38,454][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:39:38,952][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:39:39,454][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:39:39,949][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:39:40,447][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:39:40,947][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:39:41,450][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:39:41,950][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:39:42,450][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:39:42,946][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:39:43,447][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:39:43,944][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:39:44,474][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:39:44,974][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:39:45,471][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:39:45,971][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:39:46,477][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:39:46,978][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:39:47,481][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:39:47,981][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:39:48,476][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:39:48,976][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:39:49,472][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:39:49,968][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:39:50,467][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:39:50,962][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:39:51,458][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:39:51,953][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:39:52,449][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:39:52,951][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:39:53,446][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:39:53,939][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:39:54,439][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:39:54,933][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:39:55,438][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:39:55,933][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:39:56,430][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:39:56,930][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:39:57,426][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:39:57,923][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:39:58,422][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:39:58,918][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:39:59,439][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:39:59,938][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:40:00,434][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:40:00,932][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:40:01,431][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:40:01,930][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:40:02,430][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:40:02,931][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:40:03,431][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:40:03,931][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:40:04,435][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:40:04,936][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:40:05,440][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:40:05,949][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:40:06,449][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:40:06,950][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:40:07,464][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:40:07,974][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10248 tokens. -[2025-11-10 05:40:08,964][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.24%, Current % of VRAM taken: 55.73%, Block Peak % of device VRAM: 62.36%, ΔTime: 00:00:32 -[2025-11-10 05:40:09,712][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:40:09,714][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:40:09,716][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:40:10,683][__main__][INFO] - Iteration 707 took 1m 5s (44.24% Gen, 54.28% Train). Generation: 28s, Training: 35s. Estimated remaining time: 42h 38m 48s. Estimated total time: 54h 20m 2s. Time estimates for 10 more iterations: 10m 52s, 100 more iterations: 1h 48m 40s, 500 more iterations: 9h 3m 20s. -[2025-11-10 05:40:10,687][__main__][INFO] - Starting iteration 707. -[2025-11-10 05:40:11,162][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 70 and human policies 1. -[2025-11-10 05:40:11,163][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:40:27,730][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 05:40:34,758][__main__][INFO] - Number of regex retries in iteration 707: 1 -[2025-11-10 05:40:34,759][__main__][INFO] - agents played in iteration 707 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:40:35,530][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:40:35,556][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:40:35,581][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:40:35,603][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.65%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:40:35,603][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:40:35,605][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:40:36,433][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:40:36,890][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:40:37,395][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:40:37,891][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:40:38,391][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:40:38,892][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:40:39,387][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:40:39,885][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:40:40,387][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:40:40,884][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:40:41,381][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:40:41,881][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:40:42,378][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:40:42,877][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:40:43,373][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:40:43,871][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:40:44,369][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:40:44,868][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:40:45,377][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:40:45,881][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:40:46,386][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:40:46,895][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:40:47,399][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:40:47,895][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:40:48,393][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:40:48,888][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:40:49,384][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:40:49,878][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:40:50,377][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:40:50,873][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:40:51,369][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:40:51,865][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:40:52,362][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:40:52,857][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:40:53,362][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:40:53,859][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:40:54,354][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:40:54,850][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:40:55,347][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:40:55,853][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:40:56,352][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:40:56,847][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:40:57,343][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:40:57,839][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:40:58,338][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:40:58,837][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:40:59,334][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:40:59,851][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:41:00,349][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:41:00,847][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:41:01,345][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:41:01,842][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:41:02,339][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:41:02,840][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:41:03,336][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:41:03,850][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:41:04,348][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:41:04,846][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:41:05,351][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:41:05,853][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:41:06,360][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:41:06,862][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:41:07,364][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:41:07,872][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:41:08,374][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10151 tokens. -[2025-11-10 05:41:09,365][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 55.64%, Block Peak % of device VRAM: 62.28%, ΔTime: 00:00:32 -[2025-11-10 05:41:10,016][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:41:10,019][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:41:10,021][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:41:10,833][__main__][INFO] - Iteration 708 took 59s (39.54% Gen, 59.09% Train). Generation: 23s, Training: 35s. Estimated remaining time: 38h 1m 22s. Estimated total time: 49h 43m 36s. Time estimates for 10 more iterations: 9m 56s, 100 more iterations: 1h 39m 27s, 500 more iterations: 8h 17m 16s. -[2025-11-10 05:41:10,835][__main__][INFO] - Starting iteration 708. -[2025-11-10 05:41:11,302][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 70 and human policies 1. -[2025-11-10 05:41:11,303][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:41:34,370][__main__][INFO] - Number of regex retries in iteration 708: 0 -[2025-11-10 05:41:34,371][__main__][INFO] - agents played in iteration 708 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:41:35,231][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:41:35,254][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:41:35,278][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:41:35,301][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:41:35,302][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:41:35,303][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:41:36,124][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:41:36,578][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:41:37,086][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:41:37,582][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:41:38,080][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:41:38,577][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:41:39,074][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:41:39,583][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:41:40,081][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:41:40,580][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:41:41,079][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:41:41,578][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:41:42,077][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:41:42,576][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:41:43,075][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:41:43,577][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:41:44,076][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:41:44,574][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:41:45,080][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:41:45,583][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:41:46,091][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:41:46,589][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:41:47,087][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:41:47,588][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:41:48,089][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:41:48,588][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:41:49,086][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:41:49,584][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:41:50,082][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:41:50,581][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:41:51,077][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:41:51,580][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:41:52,077][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:41:52,575][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:41:53,076][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:41:53,575][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:41:54,103][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:41:54,606][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:41:55,102][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:41:55,600][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:41:56,097][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:41:56,599][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:41:57,096][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:41:57,595][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:41:58,096][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:41:58,593][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:41:59,093][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:41:59,592][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:42:00,090][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:42:00,591][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:42:01,089][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:42:01,591][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:42:02,088][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:42:02,585][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:42:03,081][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:42:03,579][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:42:04,077][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:42:04,577][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:42:05,081][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:42:05,582][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:42:06,087][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:42:06,589][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:42:07,090][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:42:07,595][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:42:08,098][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10217 tokens. -[2025-11-10 05:42:09,047][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 55.66%, Block Peak % of device VRAM: 62.20%, ΔTime: 00:00:32 -[2025-11-10 05:42:09,808][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:42:09,810][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:42:09,811][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:42:10,754][__main__][INFO] - Iteration 709 took 59s (38.80% Gen, 59.61% Train). Generation: 23s, Training: 35s. Estimated remaining time: 37h 49m 23s. Estimated total time: 49h 32m 37s. Time estimates for 10 more iterations: 9m 54s, 100 more iterations: 1h 39m 5s, 500 more iterations: 8h 15m 26s. -[2025-11-10 05:42:10,757][__main__][INFO] - Starting iteration 709. -[2025-11-10 05:42:11,227][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 70 and human policies 1. -[2025-11-10 05:42:11,228][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:42:29,201][mllm.models.large_language_model_local][WARNING] - Response Proposal: x hats, y books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 05:42:40,215][__main__][INFO] - Number of regex retries in iteration 709: 1 -[2025-11-10 05:42:40,215][__main__][INFO] - agents played in iteration 709 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:42:40,983][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:42:41,006][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:42:41,029][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:42:41,051][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.62%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:42:41,051][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:42:41,052][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:42:41,866][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:42:42,326][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:42:42,828][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:42:43,324][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:42:43,824][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:42:44,329][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:42:44,826][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:42:45,325][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:42:45,822][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:42:46,320][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:42:46,822][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:42:47,320][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:42:47,816][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:42:48,313][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:42:48,809][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:42:49,306][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:42:49,807][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:42:50,305][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:42:50,802][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:42:51,299][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:42:51,797][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:42:52,295][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:42:52,794][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:42:53,292][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:42:53,796][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:42:54,291][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:42:54,790][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:42:55,285][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:42:55,781][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:42:56,278][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:42:56,773][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:42:57,268][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:42:57,764][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:42:58,259][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:42:58,757][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:42:59,254][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:42:59,752][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:43:00,252][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:43:00,751][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:43:01,260][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:43:01,762][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:43:02,262][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:43:02,808][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:43:03,318][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:43:03,817][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:43:04,318][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:43:04,819][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:43:05,327][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:43:05,833][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:43:06,332][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:43:06,832][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:43:07,333][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:43:07,837][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:43:08,339][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:43:08,843][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:43:09,357][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:43:09,861][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:43:10,372][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:43:10,875][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:43:11,376][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:43:11,884][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:43:12,385][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:43:12,891][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:43:13,393][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:43:13,898][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10154 tokens. -[2025-11-10 05:43:14,883][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.09%, Current % of VRAM taken: 55.58%, Block Peak % of device VRAM: 62.45%, ΔTime: 00:00:33 -[2025-11-10 05:43:15,525][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:43:15,527][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:43:15,529][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:43:16,380][__main__][INFO] - Iteration 710 took 1m 5s (44.49% Gen, 54.20% Train). Generation: 28s, Training: 35s. Estimated remaining time: 42h 33m 23s. Estimated total time: 54h 17m 42s. Time estimates for 10 more iterations: 10m 51s, 100 more iterations: 1h 48m 35s, 500 more iterations: 9h 2m 57s. -[2025-11-10 05:43:16,383][__main__][INFO] - Starting iteration 710. -[2025-11-10 05:43:16,851][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 70 and human policies 1. -[2025-11-10 05:43:16,852][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:43:45,495][__main__][INFO] - Number of regex retries in iteration 710: 0 -[2025-11-10 05:43:45,496][__main__][INFO] - agents played in iteration 710 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:43:46,258][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:43:46,286][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:43:46,313][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:43:46,337][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:43:46,337][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:43:46,338][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:43:47,148][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:43:47,606][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:43:48,108][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:43:48,605][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:43:49,101][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:43:49,602][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:43:50,099][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:43:50,595][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:43:51,095][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:43:51,592][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:43:52,090][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:43:52,591][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:43:53,086][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:43:53,588][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:43:54,084][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:43:54,580][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:43:55,075][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:43:55,572][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:43:56,069][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:43:56,566][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:43:57,063][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:43:57,565][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:43:58,062][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:43:58,558][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:43:59,058][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:43:59,553][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:44:00,048][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:44:00,553][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:44:01,050][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:44:01,562][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:44:02,056][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:44:02,550][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:44:03,046][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:44:03,541][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:44:04,039][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:44:04,545][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:44:05,041][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:44:05,537][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:44:06,036][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:44:06,534][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:44:07,034][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:44:07,533][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:44:08,029][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:44:08,532][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:44:09,027][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:44:09,525][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:44:10,024][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:44:10,525][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:44:11,029][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:44:11,529][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:44:12,029][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:44:12,530][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:44:13,030][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:44:13,537][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:44:14,038][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:44:14,540][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:44:15,054][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:44:15,557][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:44:16,062][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:44:16,568][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:44:17,074][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:44:17,577][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:44:18,084][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:44:18,586][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:44:19,090][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10080 tokens. -[2025-11-10 05:44:20,030][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 55.67%, Block Peak % of device VRAM: 62.29%, ΔTime: 00:00:32 -[2025-11-10 05:44:20,769][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:44:20,771][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:44:20,773][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:44:22,702][__main__][INFO] - Iteration 711 took 1m 5s (43.50% Gen, 53.57% Train). Generation: 28s, Training: 35s. Estimated remaining time: 43h 7m 8s. Estimated total time: 54h 52m 34s. Time estimates for 10 more iterations: 10m 58s, 100 more iterations: 1h 49m 45s, 500 more iterations: 9h 8m 45s. -[2025-11-10 05:44:22,704][__main__][INFO] - Starting iteration 711. -[2025-11-10 05:44:23,171][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 71 and human policies 1. -[2025-11-10 05:44:23,172][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:44:44,663][mllm.models.large_language_model_local][WARNING] - Response Proposal: x hats, y books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 05:44:55,294][__main__][INFO] - Number of regex retries in iteration 711: 1 -[2025-11-10 05:44:55,295][__main__][INFO] - agents played in iteration 711 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:44:56,079][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:44:56,108][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:44:56,135][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:44:56,159][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:44:56,159][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:44:56,160][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:44:56,994][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:44:57,452][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:44:57,952][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:44:58,453][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:44:58,949][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:44:59,446][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:44:59,942][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:45:00,438][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:45:00,935][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:45:01,434][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:45:01,934][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:45:02,434][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:45:02,935][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:45:03,433][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:45:03,931][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:45:04,429][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:45:04,936][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:45:05,438][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:45:05,937][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:45:06,437][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:45:06,935][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:45:07,435][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:45:07,931][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:45:08,425][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:45:08,922][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:45:09,445][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:45:09,951][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:45:10,454][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:45:10,952][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:45:11,454][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:45:11,962][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:45:12,468][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:45:12,974][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:45:13,477][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:45:13,981][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:45:14,485][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:45:14,989][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:45:15,494][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:45:15,998][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:45:16,502][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:45:17,009][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:45:17,515][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:45:18,018][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:45:18,521][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:45:19,026][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:45:19,528][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:45:20,032][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:45:20,534][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:45:21,036][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:45:21,541][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:45:22,044][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:45:22,548][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:45:23,053][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:45:23,556][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:45:24,059][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:45:24,560][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:45:25,065][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:45:25,569][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:45:26,069][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:45:26,572][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:45:27,088][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:45:27,590][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:45:28,093][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:45:28,598][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:45:29,098][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10054 tokens. -[2025-11-10 05:45:30,085][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 55.65%, Block Peak % of device VRAM: 62.32%, ΔTime: 00:00:33 -[2025-11-10 05:45:30,735][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:45:30,737][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:45:30,739][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:45:31,550][__main__][INFO] - Iteration 712 took 1m 8s (46.98% Gen, 51.83% Train). Generation: 32s, Training: 35s. Estimated remaining time: 45h 12m 24s. Estimated total time: 56h 58m 59s. Time estimates for 10 more iterations: 11m 23s, 100 more iterations: 1h 53m 57s, 500 more iterations: 9h 29m 49s. -[2025-11-10 05:45:31,552][__main__][INFO] - Starting iteration 712. -[2025-11-10 05:45:32,017][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 71 and human policies 1. -[2025-11-10 05:45:32,018][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:45:53,897][__main__][INFO] - Number of regex retries in iteration 712: 0 -[2025-11-10 05:45:53,898][__main__][INFO] - agents played in iteration 712 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:45:54,791][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:45:54,816][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:45:54,840][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:45:54,862][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.63%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:45:54,862][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:45:54,863][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:45:55,684][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:45:56,144][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:45:56,646][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:45:57,149][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:45:57,648][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:45:58,143][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:45:58,641][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:45:59,140][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:45:59,641][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:46:00,139][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:46:00,641][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:46:01,139][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:46:01,635][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:46:02,133][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:46:02,632][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:46:03,129][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:46:03,630][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:46:04,124][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:46:04,619][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:46:05,116][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:46:05,610][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:46:06,105][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:46:06,604][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:46:07,099][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:46:07,621][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:46:08,121][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:46:08,617][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:46:09,113][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:46:09,609][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:46:10,118][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:46:10,614][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:46:11,113][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:46:11,613][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:46:12,108][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:46:12,605][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:46:13,100][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:46:13,595][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:46:14,118][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:46:14,619][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:46:15,115][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:46:15,612][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:46:16,112][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:46:16,623][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:46:17,124][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:46:17,630][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:46:18,135][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:46:18,639][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:46:19,143][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:46:19,648][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:46:20,151][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:46:20,656][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:46:21,161][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:46:21,662][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:46:22,164][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:46:22,665][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:46:23,167][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:46:23,670][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:46:24,178][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:46:24,691][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:46:25,194][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:46:25,697][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:46:26,201][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:46:26,705][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:46:27,213][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:46:27,716][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10055 tokens. -[2025-11-10 05:46:28,669][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 55.61%, Block Peak % of device VRAM: 62.12%, ΔTime: 00:00:32 -[2025-11-10 05:46:29,394][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:46:29,395][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:46:29,397][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:46:30,335][__main__][INFO] - Iteration 713 took 58s (37.51% Gen, 60.87% Train). Generation: 21s, Training: 35s. Estimated remaining time: 36h 48m 28s. Estimated total time: 48h 36m 1s. Time estimates for 10 more iterations: 9m 43s, 100 more iterations: 1h 37m 12s, 500 more iterations: 8h 6m 0s. -[2025-11-10 05:46:30,338][__main__][INFO] - Starting iteration 713. -[2025-11-10 05:46:30,825][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 71 and human policies 1. -[2025-11-10 05:46:30,825][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:46:52,550][__main__][INFO] - Number of regex retries in iteration 713: 0 -[2025-11-10 05:46:52,552][__main__][INFO] - agents played in iteration 713 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:46:53,388][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:46:53,415][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:46:53,442][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:46:53,465][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.58%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:46:53,465][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:46:53,466][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:46:54,290][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:46:54,745][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:46:55,244][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:46:55,743][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:46:56,239][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:46:56,735][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:46:57,232][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:46:57,729][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:46:58,229][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:46:58,726][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:46:59,226][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:46:59,732][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:47:00,226][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:47:00,721][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:47:01,217][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:47:01,713][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:47:02,217][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:47:02,813][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:47:03,322][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:47:03,819][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:47:04,317][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:47:04,825][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:47:05,321][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:47:05,820][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:47:06,319][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:47:06,814][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:47:07,310][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:47:07,809][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:47:08,303][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:47:08,804][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:47:09,299][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:47:09,794][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:47:10,291][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:47:10,785][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:47:11,280][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:47:11,774][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:47:12,269][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:47:12,765][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:47:13,262][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:47:13,762][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:47:14,257][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:47:14,755][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:47:15,251][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:47:15,748][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:47:16,246][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:47:16,746][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:47:17,241][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:47:17,741][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:47:18,238][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:47:18,740][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:47:19,244][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:47:19,752][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:47:20,253][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:47:20,757][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:47:21,265][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:47:21,770][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:47:22,279][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:47:22,780][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:47:23,283][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:47:23,784][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:47:24,286][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:47:24,787][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:47:25,288][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:47:25,788][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:47:26,300][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10009 tokens. -[2025-11-10 05:47:27,295][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.04%, Current % of VRAM taken: 55.54%, Block Peak % of device VRAM: 62.12%, ΔTime: 00:00:33 -[2025-11-10 05:47:27,940][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:47:27,942][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:47:27,944][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:47:28,749][__main__][INFO] - Iteration 714 took 57s (37.51% Gen, 61.10% Train). Generation: 21s, Training: 35s. Estimated remaining time: 36h 27m 45s. Estimated total time: 48h 16m 17s. Time estimates for 10 more iterations: 9m 39s, 100 more iterations: 1h 36m 32s, 500 more iterations: 8h 2m 42s. -[2025-11-10 05:47:28,752][__main__][INFO] - Starting iteration 714. -[2025-11-10 05:47:29,215][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 71 and human policies 1. -[2025-11-10 05:47:29,215][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:47:54,300][__main__][INFO] - Number of regex retries in iteration 714: 0 -[2025-11-10 05:47:54,300][__main__][INFO] - agents played in iteration 714 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:47:55,069][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:47:55,096][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:47:55,124][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:47:55,147][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:47:55,148][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:47:55,149][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:47:55,963][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:47:56,417][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:47:56,916][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:47:57,415][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:47:57,914][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:47:58,410][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:47:58,906][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:47:59,400][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:47:59,897][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:48:00,402][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:48:00,899][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:48:01,399][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:48:01,896][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:48:02,392][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:48:02,888][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:48:03,387][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:48:03,882][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:48:04,383][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:48:04,877][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:48:05,373][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:48:05,868][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:48:06,364][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:48:06,858][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:48:07,353][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:48:07,853][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:48:08,349][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:48:08,851][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:48:09,347][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:48:09,843][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:48:10,343][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:48:10,847][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:48:11,343][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:48:11,842][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:48:12,365][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:48:12,864][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:48:13,362][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:48:13,858][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:48:14,355][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:48:14,858][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:48:15,354][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:48:15,850][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:48:16,353][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:48:16,849][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:48:17,346][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:48:17,843][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:48:18,340][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:48:18,839][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:48:19,337][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:48:19,843][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:48:20,339][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:48:20,835][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:48:21,333][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:48:21,833][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:48:22,339][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:48:22,838][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:48:23,339][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:48:23,839][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:48:24,339][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:48:24,838][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:48:25,340][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:48:25,841][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:48:26,355][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:48:26,856][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:48:27,361][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:48:27,864][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10087 tokens. -[2025-11-10 05:48:28,805][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 55.67%, Block Peak % of device VRAM: 62.14%, ΔTime: 00:00:32 -[2025-11-10 05:48:29,553][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:48:29,557][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:48:29,559][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:48:30,538][__main__][INFO] - Iteration 715 took 1m 1s (40.91% Gen, 57.50% Train). Generation: 25s, Training: 35s. Estimated remaining time: 39h 16m 38s. Estimated total time: 51h 6m 11s. Time estimates for 10 more iterations: 10m 13s, 100 more iterations: 1h 42m 12s, 500 more iterations: 8h 31m 1s. -[2025-11-10 05:48:30,540][__main__][INFO] - Starting iteration 715. -[2025-11-10 05:48:31,002][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 71 and human policies 1. -[2025-11-10 05:48:31,003][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:48:56,333][__main__][INFO] - Number of regex retries in iteration 715: 0 -[2025-11-10 05:48:56,335][__main__][INFO] - agents played in iteration 715 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:48:57,118][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:48:57,146][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:48:57,174][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:48:57,198][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:48:57,198][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:48:57,199][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:48:58,013][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:48:58,469][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:48:58,972][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:48:59,472][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:48:59,971][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:49:00,468][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:49:00,965][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:49:01,462][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:49:01,957][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:49:02,453][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:49:02,953][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:49:03,449][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:49:03,946][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:49:04,444][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:49:04,940][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:49:05,441][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:49:05,937][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:49:06,434][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:49:06,951][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:49:07,448][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:49:07,949][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:49:08,452][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:49:08,951][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:49:09,449][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:49:09,945][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:49:10,442][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:49:10,953][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:49:11,458][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:49:11,955][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:49:12,452][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:49:12,949][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:49:13,450][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:49:13,949][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:49:14,446][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:49:14,947][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:49:15,443][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:49:15,940][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:49:16,436][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:49:16,931][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:49:17,443][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:49:17,940][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:49:18,438][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:49:18,936][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:49:19,436][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:49:19,943][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:49:20,441][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:49:20,941][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:49:21,446][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:49:21,943][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:49:22,441][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:49:22,941][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:49:23,442][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:49:23,957][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:49:24,462][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:49:24,966][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:49:25,470][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:49:25,975][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:49:26,477][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:49:26,986][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:49:27,487][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:49:27,993][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:49:28,494][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:49:28,995][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:49:29,498][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:49:30,002][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10078 tokens. -[2025-11-10 05:49:31,007][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.05%, Current % of VRAM taken: 55.55%, Block Peak % of device VRAM: 62.22%, ΔTime: 00:00:33 -[2025-11-10 05:49:31,665][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:49:31,667][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:49:31,669][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:49:32,496][__main__][INFO] - Iteration 716 took 1m 1s (41.19% Gen, 57.46% Train). Generation: 25s, Training: 35s. Estimated remaining time: 39h 24m 6s. Estimated total time: 51h 14m 41s. Time estimates for 10 more iterations: 10m 14s, 100 more iterations: 1h 42m 29s, 500 more iterations: 8h 32m 26s. -[2025-11-10 05:49:32,498][__main__][INFO] - Starting iteration 716. -[2025-11-10 05:49:32,968][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 71 and human policies 1. -[2025-11-10 05:49:32,970][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:50:00,073][__main__][INFO] - Number of regex retries in iteration 716: 0 -[2025-11-10 05:50:00,073][__main__][INFO] - agents played in iteration 716 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:50:00,854][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:50:00,882][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:50:00,908][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:50:00,931][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:50:00,932][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:50:00,933][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:50:01,772][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:50:02,234][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:50:02,739][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:50:03,247][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:50:03,751][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:50:04,248][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:50:04,746][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:50:05,246][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:50:05,744][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:50:06,241][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:50:06,739][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:50:07,250][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:50:07,748][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:50:08,245][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:50:08,740][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:50:09,236][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:50:09,743][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:50:10,239][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:50:10,735][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:50:11,233][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:50:11,733][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:50:12,227][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:50:12,723][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:50:13,219][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:50:13,741][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:50:14,236][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:50:14,730][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:50:15,228][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:50:15,721][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:50:16,214][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:50:16,707][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:50:17,200][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:50:17,697][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:50:18,193][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:50:18,689][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:50:19,194][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:50:19,695][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:50:20,194][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:50:20,690][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:50:21,186][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:50:21,684][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:50:22,183][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:50:22,679][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:50:23,178][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:50:23,674][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:50:24,173][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:50:24,670][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:50:25,166][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:50:25,667][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:50:26,163][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:50:26,661][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:50:27,160][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:50:27,659][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:50:28,158][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:50:28,657][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:50:29,159][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:50:29,657][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:50:30,153][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:50:30,662][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:50:31,162][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:50:31,666][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:50:32,176][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:50:32,679][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:50:33,183][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:50:33,689][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10090 tokens. -[2025-11-10 05:50:34,683][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.04%, Current % of VRAM taken: 55.54%, Block Peak % of device VRAM: 62.26%, ΔTime: 00:00:32 -[2025-11-10 05:50:35,442][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:50:35,444][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:50:35,446][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:50:36,333][__main__][INFO] - Iteration 717 took 1m 3s (42.77% Gen, 55.82% Train). Generation: 27s, Training: 35s. Estimated remaining time: 40h 56m 37s. Estimated total time: 52h 48m 16s. Time estimates for 10 more iterations: 10m 33s, 100 more iterations: 1h 45m 36s, 500 more iterations: 8h 48m 2s. -[2025-11-10 05:50:36,336][__main__][INFO] - Starting iteration 717. -[2025-11-10 05:50:36,805][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 71 and human policies 1. -[2025-11-10 05:50:36,806][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:50:55,501][__main__][INFO] - Number of regex retries in iteration 717: 0 -[2025-11-10 05:50:55,503][__main__][INFO] - agents played in iteration 717 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:50:56,485][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:50:56,546][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:50:56,584][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:50:56,616][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.53%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:50:56,617][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:50:56,618][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:50:57,594][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:50:58,141][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:50:58,667][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:50:59,164][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:50:59,663][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:51:00,161][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:51:00,665][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:51:01,168][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:51:01,665][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:51:02,168][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:51:02,675][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:51:03,178][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:51:03,683][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:51:04,186][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:51:04,685][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:51:05,183][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:51:05,705][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:51:06,200][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:51:06,696][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:51:07,193][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:51:07,689][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:51:08,186][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:51:08,682][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:51:09,180][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:51:09,676][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:51:10,173][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:51:10,677][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:51:11,173][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:51:11,671][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:51:12,167][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:51:12,664][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:51:13,159][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:51:13,657][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:51:14,154][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:51:14,667][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:51:15,164][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:51:15,659][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:51:16,156][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:51:16,652][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:51:17,166][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:51:17,662][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:51:18,162][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:51:18,660][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:51:19,154][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:51:19,661][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:51:20,156][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:51:20,653][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:51:21,153][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:51:21,652][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:51:22,152][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:51:22,651][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:51:23,148][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:51:23,660][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:51:24,157][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:51:24,656][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:51:25,159][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:51:25,657][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:51:26,161][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:51:26,659][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:51:27,160][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:51:27,664][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:51:28,160][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:51:28,657][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:51:29,160][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:51:29,666][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10125 tokens. -[2025-11-10 05:51:30,624][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.26%, Current % of VRAM taken: 55.75%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:33 -[2025-11-10 05:51:31,264][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:51:31,266][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:51:31,267][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:51:32,161][__main__][INFO] - Iteration 718 took 55s (33.77% Gen, 64.60% Train). Generation: 18s, Training: 35s. Estimated remaining time: 34h 15m 17s. Estimated total time: 46h 7m 52s. Time estimates for 10 more iterations: 9m 13s, 100 more iterations: 1h 32m 15s, 500 more iterations: 7h 41m 18s. -[2025-11-10 05:51:32,164][__main__][INFO] - Starting iteration 718. -[2025-11-10 05:51:32,625][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 71 and human policies 1. -[2025-11-10 05:51:32,626][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:52:01,394][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 05:52:02,277][__main__][INFO] - Number of regex retries in iteration 718: 1 -[2025-11-10 05:52:02,277][__main__][INFO] - agents played in iteration 718 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:52:03,210][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:52:03,239][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:52:03,266][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:52:03,289][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.64%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:52:03,290][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:52:03,291][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:52:04,118][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:52:04,575][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:52:05,078][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:52:05,581][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:52:06,079][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:52:06,577][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:52:07,076][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:52:07,574][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:52:08,072][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:52:08,581][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:52:09,082][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:52:09,583][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:52:10,078][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:52:10,576][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:52:11,089][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:52:11,591][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:52:12,096][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:52:12,598][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:52:13,095][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:52:13,591][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:52:14,090][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:52:14,589][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:52:15,088][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:52:15,586][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:52:16,081][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:52:16,578][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:52:17,073][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:52:17,572][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:52:18,073][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:52:18,568][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:52:19,082][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:52:19,580][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:52:20,076][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:52:20,571][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:52:21,071][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:52:21,575][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:52:22,071][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:52:22,566][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:52:23,078][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:52:23,575][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:52:24,069][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:52:24,569][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:52:25,064][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:52:25,570][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:52:26,066][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:52:26,562][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:52:27,077][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:52:27,573][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:52:28,071][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:52:28,567][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:52:29,064][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:52:29,580][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:52:30,082][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:52:30,581][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:52:31,085][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:52:31,584][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:52:32,087][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:52:32,591][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:52:33,094][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:52:33,594][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:52:34,095][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:52:34,597][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:52:35,100][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:52:35,600][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:52:36,100][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10077 tokens. -[2025-11-10 05:52:37,089][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.01%, Current % of VRAM taken: 55.50%, Block Peak % of device VRAM: 62.09%, ΔTime: 00:00:32 -[2025-11-10 05:52:37,844][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:52:37,846][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:52:37,848][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:52:38,827][__main__][INFO] - Iteration 719 took 1m 6s (44.79% Gen, 53.73% Train). Generation: 29s, Training: 35s. Estimated remaining time: 43h 16m 23s. Estimated total time: 55h 10m 5s. Time estimates for 10 more iterations: 11m 2s, 100 more iterations: 1h 50m 20s, 500 more iterations: 9h 11m 40s. -[2025-11-10 05:52:38,830][__main__][INFO] - Starting iteration 719. -[2025-11-10 05:52:39,300][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 71 and human policies 1. -[2025-11-10 05:52:39,300][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:53:02,087][__main__][INFO] - Number of regex retries in iteration 719: 0 -[2025-11-10 05:53:02,087][__main__][INFO] - agents played in iteration 719 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:53:03,001][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:53:03,024][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:53:03,050][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:53:03,073][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.51%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:53:03,073][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:53:03,074][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:53:03,956][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:53:04,412][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:53:04,922][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:53:05,421][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:53:05,922][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:53:06,420][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:53:06,921][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:53:07,424][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:53:07,924][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:53:08,422][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:53:08,928][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:53:09,428][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:53:09,928][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:53:10,428][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:53:10,929][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:53:11,435][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:53:11,940][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:53:12,439][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:53:12,939][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:53:13,441][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:53:13,945][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:53:14,443][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:53:14,944][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:53:15,453][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:53:15,952][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:53:16,452][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:53:16,948][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:53:17,446][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:53:17,957][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:53:18,456][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:53:18,955][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:53:19,453][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:53:19,952][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:53:20,471][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:53:20,969][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:53:21,472][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:53:21,974][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:53:22,472][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:53:22,985][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:53:23,483][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:53:23,979][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:53:24,476][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:53:24,973][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:53:25,484][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:53:25,983][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:53:26,479][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:53:26,976][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:53:27,475][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:53:27,976][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:53:28,477][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:53:28,975][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:53:29,493][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:53:29,992][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:53:30,489][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:53:31,019][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:53:31,526][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:53:32,035][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:53:32,542][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:53:33,048][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:53:33,563][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:53:34,062][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:53:34,563][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:53:35,066][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:53:35,571][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:53:36,089][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10102 tokens. -[2025-11-10 05:53:37,096][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 55.65%, Block Peak % of device VRAM: 62.30%, ΔTime: 00:00:33 -[2025-11-10 05:53:37,850][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:53:37,852][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:53:37,854][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:53:38,845][__main__][INFO] - Iteration 720 took 59s (38.27% Gen, 60.07% Train). Generation: 22s, Training: 35s. Estimated remaining time: 37h 42m 37s. Estimated total time: 49h 37m 19s. Time estimates for 10 more iterations: 9m 55s, 100 more iterations: 1h 39m 14s, 500 more iterations: 8h 16m 13s. -[2025-11-10 05:53:38,848][__main__][INFO] - Starting iteration 720. -[2025-11-10 05:53:39,321][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 71 and human policies 1. -[2025-11-10 05:53:39,322][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:54:03,371][__main__][INFO] - Number of regex retries in iteration 720: 0 -[2025-11-10 05:54:03,372][__main__][INFO] - agents played in iteration 720 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:54:04,269][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:54:04,292][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:54:04,317][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:54:04,340][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.61%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:54:04,341][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:54:04,341][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:54:05,225][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:54:05,682][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:54:06,184][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:54:06,682][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:54:07,180][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:54:07,680][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:54:08,178][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:54:08,677][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:54:09,192][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:54:09,689][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:54:10,185][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:54:10,684][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:54:11,182][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:54:11,679][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:54:12,177][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:54:12,674][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:54:13,175][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:54:13,673][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:54:14,175][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:54:14,676][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:54:15,175][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:54:15,677][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:54:16,173][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:54:16,669][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:54:17,168][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:54:17,667][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:54:18,167][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:54:18,663][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:54:19,162][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:54:19,665][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:54:20,162][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:54:20,660][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:54:21,182][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:54:21,679][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:54:22,178][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:54:22,682][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:54:23,177][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:54:23,671][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:54:24,166][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:54:24,661][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:54:25,156][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:54:25,653][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:54:26,149][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:54:26,649][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:54:27,150][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:54:27,649][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:54:28,148][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:54:28,644][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:54:29,141][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:54:29,637][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:54:30,134][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:54:30,635][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:54:31,133][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:54:31,634][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:54:32,131][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:54:32,628][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:54:33,125][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:54:33,621][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:54:34,120][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:54:34,615][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:54:35,112][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:54:35,610][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:54:36,107][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:54:36,610][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:54:37,109][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9938 tokens. -[2025-11-10 05:54:38,063][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 55.56%, Block Peak % of device VRAM: 61.91%, ΔTime: 00:00:32 -[2025-11-10 05:54:38,807][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:54:38,808][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:54:38,810][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:54:40,540][__main__][INFO] - Iteration 721 took 1m 1s (39.29% Gen, 57.89% Train). Generation: 24s, Training: 35s. Estimated remaining time: 39h 5m 13s. Estimated total time: 51h 0m 57s. Time estimates for 10 more iterations: 10m 12s, 100 more iterations: 1h 42m 1s, 500 more iterations: 8h 30m 9s. -[2025-11-10 05:54:40,543][__main__][INFO] - Starting iteration 721. -[2025-11-10 05:54:41,010][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 72 and human policies 1. -[2025-11-10 05:54:41,011][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:55:12,198][__main__][INFO] - Number of regex retries in iteration 721: 0 -[2025-11-10 05:55:12,200][__main__][INFO] - agents played in iteration 721 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:55:13,029][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:55:13,085][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:55:13,125][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:55:13,155][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.56%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:55:13,156][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:55:13,157][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:55:14,009][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:55:14,467][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:55:14,977][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:55:15,477][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:55:15,977][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:55:16,478][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:55:16,981][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:55:17,480][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:55:17,978][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:55:18,479][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:55:18,983][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:55:19,483][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:55:19,985][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:55:20,488][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:55:20,995][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:55:21,500][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:55:21,998][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:55:22,496][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:55:22,998][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:55:23,495][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:55:24,008][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:55:24,506][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:55:25,002][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:55:25,500][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:55:25,997][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:55:26,507][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:55:27,005][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:55:27,501][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:55:28,009][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:55:28,504][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:55:29,001][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:55:29,499][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:55:29,999][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:55:30,507][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:55:31,002][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:55:31,497][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:55:32,014][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:55:32,514][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:55:33,012][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:55:33,509][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:55:34,004][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:55:34,520][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:55:35,020][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:55:35,522][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:55:36,030][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:55:36,531][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:55:37,033][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:55:37,537][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:55:38,044][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:55:38,557][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:55:39,058][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:55:39,559][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:55:40,062][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:55:40,569][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:55:41,069][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:55:41,569][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:55:42,069][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:55:42,576][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:55:43,079][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:55:43,595][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:55:44,097][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:55:44,604][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:55:45,110][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:55:45,616][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:55:46,120][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10139 tokens. -[2025-11-10 05:55:47,084][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 55.57%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:33 -[2025-11-10 05:55:47,745][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:55:47,747][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:55:47,749][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:55:48,559][__main__][INFO] - Iteration 722 took 1m 7s (46.17% Gen, 52.63% Train). Generation: 31s, Training: 35s. Estimated remaining time: 44h 20m 35s. Estimated total time: 56h 17m 26s. Time estimates for 10 more iterations: 11m 15s, 100 more iterations: 1h 52m 34s, 500 more iterations: 9h 22m 54s. -[2025-11-10 05:55:48,561][__main__][INFO] - Starting iteration 722. -[2025-11-10 05:55:49,029][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 72 and human policies 1. -[2025-11-10 05:55:49,029][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:56:16,844][__main__][INFO] - Number of regex retries in iteration 722: 0 -[2025-11-10 05:56:16,845][__main__][INFO] - agents played in iteration 722 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:56:17,614][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:56:17,643][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:56:17,670][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:56:17,693][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.54%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:56:17,694][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:56:17,695][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:56:18,535][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:56:18,995][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:56:19,504][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:56:20,003][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:56:20,501][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:56:21,001][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:56:21,526][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:56:22,026][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:56:22,535][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:56:23,041][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:56:23,543][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:56:24,044][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:56:24,547][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:56:25,051][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:56:25,558][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:56:26,066][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:56:26,566][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:56:27,067][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:56:27,570][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:56:28,072][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:56:28,574][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:56:29,074][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:56:29,575][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:56:30,077][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:56:30,577][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:56:31,073][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:56:31,569][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:56:32,067][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:56:32,563][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:56:33,061][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:56:33,561][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:56:34,058][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:56:34,555][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:56:35,053][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:56:35,550][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:56:36,049][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:56:36,545][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:56:37,046][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:56:37,542][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:56:38,040][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:56:38,543][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:56:39,044][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:56:39,546][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:56:40,052][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:56:40,555][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:56:41,057][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:56:41,560][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:56:42,060][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:56:42,563][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:56:43,066][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:56:43,575][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:56:44,081][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:56:44,583][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:56:45,088][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:56:45,601][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:56:46,107][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:56:46,626][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:56:47,128][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:56:47,628][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:56:48,147][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:56:48,647][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:56:49,148][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:56:49,651][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:56:50,154][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:56:50,660][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10111 tokens. -[2025-11-10 05:56:51,594][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 55.68%, Block Peak % of device VRAM: 62.18%, ΔTime: 00:00:33 -[2025-11-10 05:56:52,365][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:56:52,367][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:56:52,369][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:56:53,344][__main__][INFO] - Iteration 723 took 1m 4s (43.25% Gen, 55.23% Train). Generation: 27s, Training: 35s. Estimated remaining time: 41h 37m 50s. Estimated total time: 53h 35m 46s. Time estimates for 10 more iterations: 10m 43s, 100 more iterations: 1h 47m 11s, 500 more iterations: 8h 55m 57s. -[2025-11-10 05:56:53,346][__main__][INFO] - Starting iteration 723. -[2025-11-10 05:56:53,809][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 72 and human policies 1. -[2025-11-10 05:56:53,810][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:57:20,319][__main__][INFO] - Number of regex retries in iteration 723: 0 -[2025-11-10 05:57:20,320][__main__][INFO] - agents played in iteration 723 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:57:21,090][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:57:21,117][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:57:21,144][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:57:21,167][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.60%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:57:21,168][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:57:21,169][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:57:22,005][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:57:22,466][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:57:22,969][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:57:23,469][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:57:23,971][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:57:24,476][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:57:24,980][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:57:25,483][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:57:25,984][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:57:26,482][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:57:26,976][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:57:27,475][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:57:27,970][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:57:28,465][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:57:28,964][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:57:29,458][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:57:29,957][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:57:30,452][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:57:30,947][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:57:31,442][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:57:31,937][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:57:32,432][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:57:32,927][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 -[2025-11-10 05:57:33,420][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 -[2025-11-10 05:57:33,930][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 -[2025-11-10 05:57:34,430][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 -[2025-11-10 05:57:34,929][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 -[2025-11-10 05:57:35,444][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 -[2025-11-10 05:57:35,944][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 -[2025-11-10 05:57:36,442][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 -[2025-11-10 05:57:36,944][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 -[2025-11-10 05:57:37,469][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 -[2025-11-10 05:57:37,979][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 -[2025-11-10 05:57:38,478][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 -[2025-11-10 05:57:38,979][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 -[2025-11-10 05:57:39,481][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 -[2025-11-10 05:57:39,978][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 -[2025-11-10 05:57:40,480][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 -[2025-11-10 05:57:40,981][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 -[2025-11-10 05:57:41,482][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 -[2025-11-10 05:57:42,373][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 -[2025-11-10 05:57:43,124][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 -[2025-11-10 05:57:43,624][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 -[2025-11-10 05:57:44,125][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 -[2025-11-10 05:57:44,637][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 -[2025-11-10 05:57:45,140][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 -[2025-11-10 05:57:45,651][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 -[2025-11-10 05:57:46,150][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 -[2025-11-10 05:57:46,649][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 -[2025-11-10 05:57:47,155][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 -[2025-11-10 05:57:47,656][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 -[2025-11-10 05:57:48,161][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 -[2025-11-10 05:57:48,662][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 -[2025-11-10 05:57:49,167][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 -[2025-11-10 05:57:49,672][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 -[2025-11-10 05:57:50,173][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 -[2025-11-10 05:57:50,678][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 -[2025-11-10 05:57:51,178][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 -[2025-11-10 05:57:51,680][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 -[2025-11-10 05:57:52,188][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 -[2025-11-10 05:57:52,695][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 -[2025-11-10 05:57:53,193][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 -[2025-11-10 05:57:53,695][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 -[2025-11-10 05:57:54,196][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 -[2025-11-10 05:57:54,697][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10020 tokens. -[2025-11-10 05:57:55,676][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 55.61%, Block Peak % of device VRAM: 62.47%, ΔTime: 00:00:33 -[2025-11-10 05:57:56,388][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/policy_optimizer_state.pt -[2025-11-10 05:57:56,390][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/critic_optimizer_state.pt -[2025-11-10 05:57:56,391][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/m/muqeeth/llm_negotiation/2025_11/no_press_10_1_ties_ad_align_nocurrtimestep_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl -[2025-11-10 05:57:57,237][__main__][INFO] - Iteration 724 took 1m 3s (41.79% Gen, 56.87% Train). Generation: 26s, Training: 36s. Estimated remaining time: 40h 52m 25s. Estimated total time: 52h 51m 25s. Time estimates for 10 more iterations: 10m 34s, 100 more iterations: 1h 45m 42s, 500 more iterations: 8h 48m 34s. -[2025-11-10 05:57:57,241][__main__][INFO] - Starting iteration 724. -[2025-11-10 05:57:57,707][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 72 and human policies 1. -[2025-11-10 05:57:57,708][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 -[2025-11-10 05:58:18,437][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 10 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 -[2025-11-10 05:58:29,079][__main__][INFO] - Number of regex retries in iteration 724: 1 -[2025-11-10 05:58:29,080][__main__][INFO] - agents played in iteration 724 are Alice_buffer, Alice, Bob_buffer, Bob -[2025-11-10 05:58:29,862][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:58:29,890][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:58:29,917][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:58:29,939][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 48.59%, Block Peak % of device VRAM: 19.53%, ΔTime: 00:00:00 -[2025-11-10 05:58:29,940][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. -[2025-11-10 05:58:29,941][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. -[2025-11-10 05:58:30,793][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 -[2025-11-10 05:58:31,250][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 -[2025-11-10 05:58:31,752][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 -[2025-11-10 05:58:32,248][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 -[2025-11-10 05:58:32,746][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 -[2025-11-10 05:58:33,245][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 -[2025-11-10 05:58:33,743][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 -[2025-11-10 05:58:34,265][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 -[2025-11-10 05:58:34,762][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 -[2025-11-10 05:58:35,266][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 -[2025-11-10 05:58:35,767][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 -[2025-11-10 05:58:36,267][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 -[2025-11-10 05:58:36,773][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 -[2025-11-10 05:58:37,277][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 -[2025-11-10 05:58:37,778][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 -[2025-11-10 05:58:38,277][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 -[2025-11-10 05:58:38,781][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 -[2025-11-10 05:58:39,278][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 -[2025-11-10 05:58:39,776][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 -[2025-11-10 05:58:40,272][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 -[2025-11-10 05:58:40,771][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 -[2025-11-10 05:58:41,267][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 -[2025-11-10 05:58:41,763][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +Given the values, I recognize that balls are of significant value to me (10) compared to their low value to Alice (1). I should keep all 10 balls to maximize my points. Hats and books have a low value to me, so I can afford to give them to Alice, keeping as little as possible to ensure the total proposals do not exceed the item quantity. did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 08:45:29,279][__main__][INFO] - Number of regex retries in iteration 748: 2 +[2026-03-26 08:45:29,279][__main__][INFO] - agents played in iteration 748 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:45:30,053][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:45:30,078][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:45:30,101][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:45:30,125][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:45:30,126][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:45:30,126][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:45:30,993][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:45:31,466][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:45:31,974][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:45:32,480][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:45:32,984][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:45:33,488][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:45:33,989][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:45:34,491][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:45:34,992][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:45:35,494][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:45:35,999][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:45:36,500][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:45:37,002][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:45:37,504][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:45:38,008][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:45:38,509][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:45:39,011][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:45:39,511][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:45:40,011][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:45:40,515][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:45:41,017][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:45:41,522][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:45:42,031][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:45:42,531][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:45:43,032][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:45:43,534][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:45:44,038][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:45:44,538][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:45:45,039][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:45:45,540][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:45:46,043][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:45:46,546][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:45:47,048][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:45:47,547][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:45:48,046][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:45:48,545][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:45:49,045][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:45:49,544][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:45:50,042][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:45:50,540][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:45:51,038][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:45:51,535][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:45:52,033][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:45:52,531][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:45:53,033][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:45:53,530][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:45:54,028][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:45:54,528][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:45:55,026][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:45:55,526][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:45:56,026][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:45:56,526][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:45:57,025][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:45:57,525][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:45:58,025][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:45:58,524][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:45:59,023][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:45:59,522][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:46:00,020][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:46:00,518][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:46:01,019][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:46:01,515][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:46:02,013][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:46:02,511][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:46:03,011][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10117 tokens. +[2026-03-26 08:46:03,908][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.18%, Current % of VRAM taken: 60.61%, Block Peak % of device VRAM: 62.22%, ΔTime: 00:00:32 +[2026-03-26 08:46:04,665][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:46:04,667][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:46:04,669][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:46:05,435][__main__][INFO] - Iteration 749 took 51s (30.03% Gen, 68.48% Train). Generation: 15s, Training: 35s. Estimated remaining time: 31h 23m 20s. Estimated total time: 43h 3m 45s. Time estimates for 10 more iterations: 8m 36s, 100 more iterations: 1h 26m 7s, 500 more iterations: 7h 10m 37s. +[2026-03-26 08:46:05,437][__main__][INFO] - Starting iteration 749. +[2026-03-26 08:46:05,853][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 74 and human policies 1. +[2026-03-26 08:46:05,854][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:46:20,745][__main__][INFO] - Number of regex retries in iteration 749: 0 +[2026-03-26 08:46:20,746][__main__][INFO] - agents played in iteration 749 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:46:21,559][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:46:21,583][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:46:21,607][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:46:21,630][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:46:21,631][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:46:21,631][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:46:22,493][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:46:22,945][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:46:23,451][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:46:23,950][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:46:24,449][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:46:24,949][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:46:25,448][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:46:25,946][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:46:26,449][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:46:26,952][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:46:27,451][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:46:27,956][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:46:28,464][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:46:28,966][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:46:29,474][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:46:29,976][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:46:30,480][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:46:30,985][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:46:31,486][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:46:31,991][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:46:32,493][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:46:32,996][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:46:33,498][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:46:34,002][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:46:34,506][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:46:35,004][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:46:35,505][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:46:36,005][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:46:36,506][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:46:37,005][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:46:37,501][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:46:37,999][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:46:38,499][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:46:39,002][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:46:39,505][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:46:40,004][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:46:40,502][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:46:41,004][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:46:41,504][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:46:42,002][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:46:42,505][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:46:43,003][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:46:43,501][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:46:44,007][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:46:44,505][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:46:45,011][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:46:45,511][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:46:46,018][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:46:46,518][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:46:47,017][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:46:47,519][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:46:48,019][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:46:48,518][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:46:49,017][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:46:49,516][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:46:50,016][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:46:50,515][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:46:51,015][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:46:51,513][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:46:52,012][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:46:52,512][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:46:53,011][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:46:53,510][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:46:54,012][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:46:54,513][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10097 tokens. +[2026-03-26 08:46:55,430][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.22%, ΔTime: 00:00:32 +[2026-03-26 08:46:56,179][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:46:56,182][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:46:56,183][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:46:56,913][__main__][INFO] - Iteration 750 took 51s (29.17% Gen, 69.40% Train). Generation: 14s, Training: 35s. Estimated remaining time: 30h 51m 47s. Estimated total time: 42h 33m 4s. Time estimates for 10 more iterations: 8m 30s, 100 more iterations: 1h 25m 6s, 500 more iterations: 7h 5m 30s. +[2026-03-26 08:46:56,916][__main__][INFO] - Starting iteration 750. +[2026-03-26 08:46:57,332][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 74 and human policies 1. +[2026-03-26 08:46:57,333][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:47:11,662][__main__][INFO] - Number of regex retries in iteration 750: 0 +[2026-03-26 08:47:11,662][__main__][INFO] - agents played in iteration 750 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:47:12,434][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:47:12,458][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:47:12,482][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:47:12,505][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:47:12,506][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:47:12,506][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:47:13,349][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:47:13,802][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:47:14,304][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:47:14,806][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:47:15,302][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:47:15,801][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:47:16,302][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:47:16,809][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:47:17,323][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:47:17,822][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:47:18,323][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:47:18,822][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:47:19,320][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:47:19,821][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:47:20,325][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:47:20,838][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:47:21,336][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:47:21,842][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:47:22,343][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:47:22,843][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:47:23,343][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:47:23,842][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:47:24,346][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:47:24,854][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:47:25,359][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:47:25,862][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:47:26,366][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:47:26,875][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:47:27,376][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:47:27,881][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:47:28,383][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:47:28,888][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:47:29,395][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:47:29,902][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:47:30,404][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:47:30,907][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:47:31,408][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:47:31,911][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:47:32,412][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:47:32,913][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:47:33,413][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:47:33,916][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:47:34,419][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:47:34,933][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:47:35,437][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:47:35,941][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:47:36,441][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:47:36,946][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:47:37,448][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:47:37,953][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:47:38,455][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:47:38,955][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:47:39,457][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:47:39,957][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:47:40,461][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:47:40,968][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:47:41,481][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:47:41,982][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:47:42,482][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:47:42,984][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:47:43,487][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:47:43,989][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:47:44,490][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:47:44,991][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:47:45,492][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10099 tokens. +[2026-03-26 08:47:46,407][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 60.50%, Block Peak % of device VRAM: 62.11%, ΔTime: 00:00:33 +[2026-03-26 08:47:47,159][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:47:47,161][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:47:47,163][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:47:48,522][__main__][INFO] - Iteration 751 took 51s (27.99% Gen, 69.35% Train). Generation: 14s, Training: 35s. Estimated remaining time: 30h 57m 21s. Estimated total time: 42h 39m 29s. Time estimates for 10 more iterations: 8m 31s, 100 more iterations: 1h 25m 18s, 500 more iterations: 7h 6m 34s. +[2026-03-26 08:47:48,524][__main__][INFO] - Starting iteration 751. +[2026-03-26 08:47:48,933][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 75 and human policies 1. +[2026-03-26 08:47:48,934][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:48:04,217][__main__][INFO] - Number of regex retries in iteration 751: 0 +[2026-03-26 08:48:04,217][__main__][INFO] - agents played in iteration 751 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:48:04,988][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:48:05,012][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:48:05,036][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:48:05,060][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:48:05,061][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:48:05,062][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:48:05,920][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:48:06,381][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:48:06,883][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:48:07,379][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:48:07,875][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:48:08,372][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:48:08,871][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:48:09,368][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:48:09,867][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:48:10,367][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:48:10,867][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:48:11,365][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:48:11,868][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:48:12,379][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:48:12,885][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:48:13,394][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:48:13,903][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:48:14,407][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:48:14,907][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:48:15,407][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:48:15,908][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:48:16,409][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:48:16,911][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:48:17,410][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:48:17,913][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:48:18,412][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:48:18,912][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:48:19,411][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:48:19,910][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:48:20,412][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:48:20,911][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:48:21,413][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:48:21,911][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:48:22,409][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:48:22,907][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:48:23,407][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:48:23,906][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:48:24,409][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:48:24,914][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:48:25,413][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:48:25,920][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:48:26,422][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:48:26,922][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:48:27,424][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:48:27,926][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:48:28,426][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:48:28,926][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:48:29,427][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:48:29,926][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:48:30,425][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:48:30,922][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:48:31,420][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:48:31,921][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:48:32,420][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:48:32,917][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:48:33,415][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:48:33,912][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:48:34,409][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:48:34,908][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:48:35,406][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:48:35,902][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:48:36,398][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:48:36,896][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:48:37,393][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:48:37,891][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10089 tokens. +[2026-03-26 08:48:38,798][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 60.54%, Block Peak % of device VRAM: 62.25%, ΔTime: 00:00:32 +[2026-03-26 08:48:39,554][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:48:39,556][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:48:39,557][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:48:40,416][__main__][INFO] - Iteration 752 took 51s (29.69% Gen, 68.64% Train). Generation: 15s, Training: 35s. Estimated remaining time: 31h 11m 11s. Estimated total time: 42h 54m 11s. Time estimates for 10 more iterations: 8m 34s, 100 more iterations: 1h 25m 48s, 500 more iterations: 7h 9m 1s. +[2026-03-26 08:48:40,419][__main__][INFO] - Starting iteration 752. +[2026-03-26 08:48:40,838][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 75 and human policies 1. +[2026-03-26 08:48:40,838][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:48:54,763][__main__][INFO] - Number of regex retries in iteration 752: 0 +[2026-03-26 08:48:54,764][__main__][INFO] - agents played in iteration 752 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:48:55,527][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:48:55,551][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:48:55,575][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:48:55,598][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:48:55,599][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:48:55,599][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:48:56,461][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:48:56,915][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:48:57,421][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:48:57,928][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:48:58,435][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:48:58,937][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:48:59,442][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:48:59,942][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:49:00,442][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:49:00,949][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:49:01,461][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:49:01,976][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:49:02,481][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:49:02,984][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:49:03,487][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:49:03,988][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:49:04,491][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:49:04,993][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:49:05,493][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:49:05,994][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:49:06,497][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:49:06,996][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:49:07,499][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:49:07,998][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:49:08,504][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:49:09,008][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:49:09,515][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:49:10,017][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:49:10,519][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:49:11,019][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:49:11,519][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:49:12,021][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:49:12,522][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:49:13,025][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:49:13,525][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:49:14,030][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:49:14,535][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:49:15,047][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:49:15,550][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:49:16,052][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:49:16,554][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:49:17,054][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:49:17,553][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:49:18,054][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:49:18,554][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:49:19,053][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:49:19,552][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:49:20,057][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:49:20,560][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:49:21,064][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:49:21,571][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:49:22,076][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:49:22,577][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:49:23,078][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:49:23,580][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:49:24,080][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:49:24,582][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:49:25,083][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:49:25,584][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:49:26,085][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:49:26,587][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:49:27,090][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:49:27,595][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:49:28,098][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:49:28,604][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10019 tokens. +[2026-03-26 08:49:29,539][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.09%, Current % of VRAM taken: 60.52%, Block Peak % of device VRAM: 62.17%, ΔTime: 00:00:33 +[2026-03-26 08:49:30,290][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:49:30,292][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:49:30,293][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:49:31,134][__main__][INFO] - Iteration 753 took 50s (27.69% Gen, 70.64% Train). Generation: 13s, Training: 35s. Estimated remaining time: 30h 11m 0s. Estimated total time: 41h 54m 51s. Time estimates for 10 more iterations: 8m 22s, 100 more iterations: 1h 23m 49s, 500 more iterations: 6h 59m 8s. +[2026-03-26 08:49:31,136][__main__][INFO] - Starting iteration 753. +[2026-03-26 08:49:31,544][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 75 and human policies 1. +[2026-03-26 08:49:31,544][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:49:45,918][__main__][INFO] - Number of regex retries in iteration 753: 0 +[2026-03-26 08:49:45,919][__main__][INFO] - agents played in iteration 753 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:49:46,692][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:49:46,716][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:49:46,739][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:49:46,763][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:49:46,763][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:49:46,764][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:49:47,635][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:49:48,090][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:49:48,591][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:49:49,091][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:49:49,589][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:49:50,089][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:49:50,587][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:49:51,090][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:49:51,594][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:49:52,095][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:49:52,602][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:49:53,114][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:49:53,615][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:49:54,116][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:49:54,617][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:49:55,118][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:49:55,619][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:49:56,120][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:49:56,622][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:49:57,123][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:49:57,623][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:49:58,124][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:49:58,624][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:49:59,123][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:49:59,623][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:50:00,123][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:50:00,629][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:50:01,132][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:50:01,631][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:50:02,132][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:50:02,630][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:50:03,128][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:50:03,627][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:50:04,129][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:50:04,626][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:50:05,123][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:50:05,621][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:50:06,119][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:50:06,617][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:50:07,117][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:50:07,618][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:50:08,119][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:50:08,619][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:50:09,124][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:50:09,623][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:50:10,129][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:50:10,630][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:50:11,131][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:50:11,631][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:50:12,131][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:50:12,630][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:50:13,129][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:50:13,628][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:50:14,129][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:50:14,631][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:50:15,131][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:50:15,632][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:50:16,132][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:50:16,630][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:50:17,133][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:50:17,634][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:50:18,131][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:50:18,629][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:50:19,127][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:50:19,626][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10077 tokens. +[2026-03-26 08:50:20,548][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 60.56%, Block Peak % of device VRAM: 62.03%, ΔTime: 00:00:32 +[2026-03-26 08:50:21,300][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:50:21,302][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:50:21,304][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:50:22,031][__main__][INFO] - Iteration 754 took 50s (28.47% Gen, 70.09% Train). Generation: 14s, Training: 35s. Estimated remaining time: 30h 19m 40s. Estimated total time: 42h 4m 22s. Time estimates for 10 more iterations: 8m 24s, 100 more iterations: 1h 24m 8s, 500 more iterations: 7h 0m 43s. +[2026-03-26 08:50:22,033][__main__][INFO] - Starting iteration 754. +[2026-03-26 08:50:22,441][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 75 and human policies 1. +[2026-03-26 08:50:22,442][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:50:34,802][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 08:50:41,769][__main__][INFO] - Number of regex retries in iteration 754: 1 +[2026-03-26 08:50:41,770][__main__][INFO] - agents played in iteration 754 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:50:42,534][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:50:42,558][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:50:42,582][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:50:42,605][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:50:42,606][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:50:42,606][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:50:43,459][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:50:43,928][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:50:44,442][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:50:44,943][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:50:45,445][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:50:45,947][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:50:46,452][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:50:46,952][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:50:47,458][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:50:47,960][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:50:48,462][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:50:48,963][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:50:49,467][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:50:49,977][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:50:50,485][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:50:50,986][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:50:51,492][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:50:51,995][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:50:52,498][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:50:52,998][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:50:53,499][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:50:54,004][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:50:54,514][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:50:55,016][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:50:55,522][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:50:56,027][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:50:56,529][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:50:57,034][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:50:57,536][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:50:58,042][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:50:58,553][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:50:59,066][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:50:59,573][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:51:00,073][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:51:00,579][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:51:01,079][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:51:01,579][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:51:02,081][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:51:02,582][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:51:03,081][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:51:03,581][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:51:04,087][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:51:04,595][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:51:05,105][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:51:05,608][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:51:06,111][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:51:06,616][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:51:07,122][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:51:07,622][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:51:08,122][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:51:08,623][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:51:09,124][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:51:09,627][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:51:10,128][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:51:10,631][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:51:11,136][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:51:11,641][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:51:12,148][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:51:12,653][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:51:13,156][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:51:13,659][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:51:14,159][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:51:14,667][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:51:15,170][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:51:15,671][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10173 tokens. +[2026-03-26 08:51:16,620][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 60.49%, Block Peak % of device VRAM: 62.16%, ΔTime: 00:00:33 +[2026-03-26 08:51:17,378][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:51:17,381][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:51:17,382][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:51:18,111][__main__][INFO] - Iteration 755 took 55s (34.72% Gen, 63.97% Train). Generation: 19s, Training: 35s. Estimated remaining time: 34h 37m 52s. Estimated total time: 46h 23m 29s. Time estimates for 10 more iterations: 9m 16s, 100 more iterations: 1h 32m 46s, 500 more iterations: 7h 43m 54s. +[2026-03-26 08:51:18,113][__main__][INFO] - Starting iteration 755. +[2026-03-26 08:51:18,523][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 75 and human policies 1. +[2026-03-26 08:51:18,524][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:51:34,278][__main__][INFO] - Number of regex retries in iteration 755: 0 +[2026-03-26 08:51:34,279][__main__][INFO] - agents played in iteration 755 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:51:35,037][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:51:35,061][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:51:35,085][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:51:35,108][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:51:35,109][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:51:35,109][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:51:35,966][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:51:36,431][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:51:36,943][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:51:37,452][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:51:37,953][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:51:38,458][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:51:38,957][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:51:39,455][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:51:39,960][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:51:40,468][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:51:40,976][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:51:41,479][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:51:41,983][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:51:42,490][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:51:42,993][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:51:43,494][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:51:44,000][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:51:44,516][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:51:45,020][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:51:45,523][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:51:46,030][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:51:46,532][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:51:47,034][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:51:47,537][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:51:48,043][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:51:48,550][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:51:49,061][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:51:49,562][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:51:50,062][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:51:50,564][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:51:51,066][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:51:51,567][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:51:52,067][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:51:52,572][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:51:53,073][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:51:53,574][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:51:54,073][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:51:54,577][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:51:55,079][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:51:55,575][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:51:56,076][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:51:56,581][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:51:57,087][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:51:57,594][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:51:58,096][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:51:58,598][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:51:59,100][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:51:59,603][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:52:00,102][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:52:00,601][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:52:01,102][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:52:01,600][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:52:02,101][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:52:02,599][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:52:03,099][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:52:03,597][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:52:04,095][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:52:04,597][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:52:05,097][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:52:05,596][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:52:06,100][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:52:06,599][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:52:07,099][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:52:07,598][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:52:08,098][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10093 tokens. +[2026-03-26 08:52:09,004][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.11%, Current % of VRAM taken: 60.54%, Block Peak % of device VRAM: 62.28%, ΔTime: 00:00:33 +[2026-03-26 08:52:09,757][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:52:09,759][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:52:09,761][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:52:10,527][__main__][INFO] - Iteration 756 took 52s (30.29% Gen, 68.23% Train). Generation: 15s, Training: 35s. Estimated remaining time: 31h 33m 43s. Estimated total time: 43h 20m 13s. Time estimates for 10 more iterations: 8m 40s, 100 more iterations: 1h 26m 40s, 500 more iterations: 7h 13m 22s. +[2026-03-26 08:52:10,530][__main__][INFO] - Starting iteration 756. +[2026-03-26 08:52:10,944][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 75 and human policies 1. +[2026-03-26 08:52:10,945][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:52:16,036][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 08:52:25,521][__main__][INFO] - Number of regex retries in iteration 756: 1 +[2026-03-26 08:52:25,522][__main__][INFO] - agents played in iteration 756 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:52:26,298][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:52:26,321][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:52:26,345][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:52:26,368][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:52:26,369][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:52:26,370][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:52:27,243][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:52:27,706][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:52:28,221][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:52:28,725][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:52:29,227][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:52:29,728][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:52:30,227][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:52:30,728][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:52:31,233][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:52:31,737][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:52:32,240][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:52:32,741][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:52:33,242][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:52:33,742][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:52:34,243][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:52:34,748][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:52:35,251][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:52:35,753][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:52:36,261][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:52:36,773][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:52:37,274][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:52:37,776][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:52:38,277][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:52:38,776][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:52:39,276][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:52:39,776][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:52:40,283][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:52:40,784][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:52:41,288][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:52:41,788][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:52:42,288][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:52:42,788][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:52:43,289][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:52:43,789][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:52:44,288][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:52:44,787][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:52:45,286][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:52:45,784][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:52:46,286][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:52:46,785][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:52:47,287][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:52:47,786][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:52:48,285][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:52:48,782][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:52:49,284][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:52:49,783][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:52:50,282][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:52:50,780][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:52:51,282][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:52:51,782][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:52:52,281][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:52:52,779][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:52:53,281][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:52:53,780][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:52:54,279][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:52:54,777][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:52:55,275][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:52:55,772][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:52:56,269][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:52:56,766][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:52:57,264][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:52:57,760][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:52:58,255][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:52:58,752][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:52:59,249][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10028 tokens. +[2026-03-26 08:53:00,161][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.01%, Current % of VRAM taken: 60.44%, Block Peak % of device VRAM: 62.08%, ΔTime: 00:00:32 +[2026-03-26 08:53:00,910][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:53:00,912][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:53:00,914][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:53:01,683][__main__][INFO] - Iteration 757 took 50s (28.73% Gen, 69.75% Train). Generation: 14s, Training: 35s. Estimated remaining time: 30h 29m 37s. Estimated total time: 42h 16m 58s. Time estimates for 10 more iterations: 8m 27s, 100 more iterations: 1h 24m 33s, 500 more iterations: 7h 2m 49s. +[2026-03-26 08:53:01,685][__main__][INFO] - Starting iteration 757. +[2026-03-26 08:53:02,099][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 75 and human policies 1. +[2026-03-26 08:53:02,100][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:53:17,214][__main__][INFO] - Number of regex retries in iteration 757: 0 +[2026-03-26 08:53:17,215][__main__][INFO] - agents played in iteration 757 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:53:17,982][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.45%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:53:18,006][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.45%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:53:18,029][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.45%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:53:18,052][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.45%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:53:18,053][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:53:18,054][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:53:18,913][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:53:19,373][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:53:19,876][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:53:20,375][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:53:20,875][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:53:21,375][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:53:21,875][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:53:22,376][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:53:22,879][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:53:23,376][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:53:23,883][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:53:24,383][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:53:24,889][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:53:25,390][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:53:25,898][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:53:26,399][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:53:26,899][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:53:27,405][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:53:27,908][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:53:28,409][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:53:28,911][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:53:29,415][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:53:29,917][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:53:30,418][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:53:30,919][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:53:31,418][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:53:31,921][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:53:32,425][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:53:32,926][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:53:33,425][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:53:33,924][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:53:34,423][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:53:34,922][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:53:35,427][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:53:35,928][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:53:36,429][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:53:36,930][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:53:37,431][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:53:37,933][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:53:38,433][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:53:38,932][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:53:39,432][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:53:39,931][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:53:40,430][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:53:40,931][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:53:41,430][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:53:41,929][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:53:42,430][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:53:42,934][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:53:43,435][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:53:43,942][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:53:44,442][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:53:44,953][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:53:45,454][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:53:45,956][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:53:46,459][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:53:46,961][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:53:47,460][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:53:47,961][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:53:48,461][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:53:48,962][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:53:49,463][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:53:49,966][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:53:50,466][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:53:50,968][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10160 tokens. +[2026-03-26 08:53:51,868][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.16%, Current % of VRAM taken: 60.58%, Block Peak % of device VRAM: 62.36%, ΔTime: 00:00:32 +[2026-03-26 08:53:52,619][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:53:52,621][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:53:52,622][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:53:53,712][__main__][INFO] - Iteration 758 took 51s (29.28% Gen, 68.60% Train). Generation: 15s, Training: 35s. Estimated remaining time: 31h 12m 27s. Estimated total time: 43h 0m 40s. Time estimates for 10 more iterations: 8m 36s, 100 more iterations: 1h 26m 1s, 500 more iterations: 7h 10m 6s. +[2026-03-26 08:53:53,714][__main__][INFO] - Starting iteration 758. +[2026-03-26 08:53:54,120][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 75 and human policies 1. +[2026-03-26 08:53:54,121][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:54:13,208][__main__][INFO] - Number of regex retries in iteration 758: 0 +[2026-03-26 08:54:13,209][__main__][INFO] - agents played in iteration 758 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:54:13,988][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:54:14,015][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:54:14,041][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:54:14,067][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.54%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:54:14,068][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:54:14,068][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:54:14,948][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:54:15,407][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:54:15,913][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:54:16,418][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:54:16,918][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:54:17,418][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:54:17,921][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:54:18,419][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:54:18,924][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:54:19,422][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:54:19,921][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:54:20,422][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:54:20,919][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:54:21,419][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:54:21,920][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:54:22,430][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:54:22,934][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:54:23,437][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:54:23,939][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:54:24,441][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:54:24,943][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:54:25,445][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:54:25,946][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:54:26,449][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:54:26,950][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:54:27,448][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:54:27,947][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:54:28,446][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:54:28,947][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:54:29,450][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:54:29,951][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:54:30,449][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:54:30,948][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:54:31,447][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:54:31,948][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:54:32,446][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:54:32,945][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:54:33,444][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:54:33,943][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:54:34,442][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:54:34,940][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:54:35,437][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:54:35,936][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:54:36,432][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:54:36,930][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:54:37,426][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:54:37,922][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:54:38,419][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:54:38,915][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:54:39,412][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:54:39,909][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:54:40,406][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:54:40,904][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:54:41,400][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:54:41,897][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:54:42,393][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:54:42,890][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:54:43,389][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:54:43,887][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:54:44,386][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:54:44,889][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:54:45,388][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:54:45,886][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:54:46,384][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:54:46,882][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10099 tokens. +[2026-03-26 08:54:47,776][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.12%, Current % of VRAM taken: 60.55%, Block Peak % of device VRAM: 62.06%, ΔTime: 00:00:32 +[2026-03-26 08:54:48,518][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:54:48,521][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:54:48,524][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:54:49,257][__main__][INFO] - Iteration 759 took 55s (34.62% Gen, 64.05% Train). Generation: 19s, Training: 35s. Estimated remaining time: 34h 7m 44s. Estimated total time: 45h 56m 53s. Time estimates for 10 more iterations: 9m 11s, 100 more iterations: 1h 31m 53s, 500 more iterations: 7h 39m 28s. +[2026-03-26 08:54:49,260][__main__][INFO] - Starting iteration 759. +[2026-03-26 08:54:49,664][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 75 and human policies 1. +[2026-03-26 08:54:49,665][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:55:02,352][__main__][INFO] - Number of regex retries in iteration 759: 0 +[2026-03-26 08:55:02,353][__main__][INFO] - agents played in iteration 759 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:55:03,193][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:55:03,218][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:55:03,242][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:55:03,266][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:55:03,267][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:55:03,267][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:55:04,136][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:55:04,602][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:55:05,108][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:55:05,614][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:55:06,118][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:55:06,619][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:55:07,121][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:55:07,625][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:55:08,127][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:55:08,627][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:55:09,132][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:55:09,638][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:55:10,145][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:55:10,646][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:55:11,148][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:55:11,649][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:55:12,152][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:55:12,652][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:55:13,152][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:55:13,658][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:55:14,158][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:55:14,657][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:55:15,157][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:55:15,656][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:55:16,155][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:55:16,657][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:55:17,157][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:55:17,658][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:55:18,163][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:55:18,667][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:55:19,173][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:55:19,685][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:55:20,190][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:55:20,693][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:55:21,195][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:55:21,699][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:55:22,201][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:55:22,702][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:55:23,205][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:55:23,707][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:55:24,210][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:55:24,709][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:55:25,207][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:55:25,706][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:55:26,204][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:55:26,706][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:55:27,206][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:55:27,706][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:55:28,213][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:55:28,715][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:55:29,216][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:55:29,716][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:55:30,216][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:55:30,718][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:55:31,218][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:55:31,718][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:55:32,219][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:55:32,718][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:55:33,215][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:55:33,712][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:55:34,210][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:55:34,709][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:55:35,207][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:55:35,705][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:55:36,204][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10046 tokens. +[2026-03-26 08:55:37,110][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.05%, Current % of VRAM taken: 60.48%, Block Peak % of device VRAM: 62.33%, ΔTime: 00:00:32 +[2026-03-26 08:55:37,857][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:55:37,859][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:55:37,861][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:55:38,599][__main__][INFO] - Iteration 760 took 48s (25.93% Gen, 72.56% Train). Generation: 12s, Training: 35s. Estimated remaining time: 28h 56m 48s. Estimated total time: 40h 46m 46s. Time estimates for 10 more iterations: 8m 9s, 100 more iterations: 1h 21m 33s, 500 more iterations: 6h 47m 47s. +[2026-03-26 08:55:38,601][__main__][INFO] - Starting iteration 760. +[2026-03-26 08:55:39,008][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 75 and human policies 1. +[2026-03-26 08:55:39,009][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:55:51,970][__main__][INFO] - Number of regex retries in iteration 760: 0 +[2026-03-26 08:55:51,971][__main__][INFO] - agents played in iteration 760 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:55:52,739][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:55:52,763][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:55:52,787][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:55:52,811][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.50%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:55:52,811][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:55:52,812][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:55:53,699][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:55:54,166][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:55:54,677][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:55:55,181][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:55:55,683][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:55:56,187][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:55:56,697][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:55:57,205][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:55:57,707][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:55:58,209][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:55:58,711][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:55:59,212][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:55:59,713][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:56:00,215][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:56:00,721][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:56:01,222][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:56:01,722][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:56:02,223][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:56:02,725][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:56:03,232][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:56:03,741][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:56:04,241][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:56:04,743][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:56:05,245][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:56:05,746][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:56:06,246][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:56:06,746][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:56:07,248][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:56:07,749][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:56:08,250][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:56:08,750][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:56:09,253][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:56:09,753][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:56:10,256][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:56:10,758][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:56:11,264][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:56:11,765][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:56:12,266][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:56:12,765][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:56:13,265][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:56:13,768][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:56:14,268][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:56:14,768][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:56:15,267][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:56:15,767][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:56:16,270][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:56:16,770][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:56:17,271][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:56:17,770][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:56:18,270][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:56:18,771][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:56:19,270][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:56:19,770][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:56:20,270][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:56:20,770][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:56:21,271][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:56:21,771][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:56:22,272][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:56:22,771][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:56:23,270][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:56:23,769][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:56:24,267][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:56:24,768][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:56:25,266][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:56:25,764][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9988 tokens. +[2026-03-26 08:56:26,676][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 60.48%, Block Peak % of device VRAM: 62.14%, ΔTime: 00:00:32 +[2026-03-26 08:56:27,437][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:56:27,440][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:56:27,441][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:56:28,775][__main__][INFO] - Iteration 761 took 49s (26.04% Gen, 71.27% Train). Generation: 12s, Training: 35s. Estimated remaining time: 29h 37m 34s. Estimated total time: 41h 28m 22s. Time estimates for 10 more iterations: 8m 17s, 100 more iterations: 1h 22m 56s, 500 more iterations: 6h 54m 43s. +[2026-03-26 08:56:28,778][__main__][INFO] - Starting iteration 761. +[2026-03-26 08:56:29,183][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 76 and human policies 1. +[2026-03-26 08:56:29,183][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:56:44,060][__main__][INFO] - Number of regex retries in iteration 761: 0 +[2026-03-26 08:56:44,061][__main__][INFO] - agents played in iteration 761 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:56:44,837][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:56:44,861][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:56:44,885][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:56:44,908][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:56:44,909][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:56:44,909][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:56:45,781][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:56:46,235][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:56:46,736][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:56:47,233][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:56:47,730][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:56:48,228][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:56:48,732][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:56:49,238][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:56:49,748][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:56:50,251][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:56:50,751][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:56:51,251][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:56:51,752][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:56:52,253][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:56:52,752][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:56:53,255][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:56:53,758][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:56:54,259][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:56:54,759][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:56:55,259][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:56:55,759][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:56:56,258][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:56:56,758][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:56:57,260][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:56:57,765][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:56:58,273][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:56:58,775][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:56:59,275][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:56:59,777][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:57:00,280][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:57:00,780][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:57:01,279][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:57:01,780][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:57:02,282][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:57:02,782][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:57:03,282][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:57:03,782][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:57:04,289][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:57:04,790][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:57:05,293][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:57:05,793][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:57:06,294][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:57:06,795][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:57:07,295][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:57:07,794][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:57:08,293][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:57:08,792][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:57:09,294][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:57:09,795][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:57:10,292][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:57:10,790][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:57:11,287][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:57:11,784][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:57:12,285][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:57:12,781][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:57:13,279][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:57:13,776][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:57:14,276][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:57:14,774][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:57:15,273][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:57:15,772][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:57:16,273][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:57:16,770][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:57:17,271][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:57:17,774][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10021 tokens. +[2026-03-26 08:57:18,708][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 60.51%, Block Peak % of device VRAM: 62.14%, ΔTime: 00:00:32 +[2026-03-26 08:57:19,474][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:57:19,476][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:57:19,478][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:57:20,257][__main__][INFO] - Iteration 762 took 51s (29.13% Gen, 69.34% Train). Generation: 14s, Training: 35s. Estimated remaining time: 30h 42m 7s. Estimated total time: 42h 33m 46s. Time estimates for 10 more iterations: 8m 30s, 100 more iterations: 1h 25m 7s, 500 more iterations: 7h 5m 37s. +[2026-03-26 08:57:20,260][__main__][INFO] - Starting iteration 762. +[2026-03-26 08:57:20,673][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 76 and human policies 1. +[2026-03-26 08:57:20,674][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:57:34,247][__main__][INFO] - Number of regex retries in iteration 762: 0 +[2026-03-26 08:57:34,248][__main__][INFO] - agents played in iteration 762 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:57:35,012][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:57:35,036][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:57:35,060][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:57:35,084][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:57:35,085][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:57:35,085][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:57:35,966][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:57:36,424][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:57:36,934][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:57:37,436][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:57:37,938][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:57:38,439][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:57:38,940][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:57:39,440][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:57:39,944][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:57:40,450][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:57:40,952][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:57:41,454][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:57:41,955][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:57:42,456][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:57:42,955][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:57:43,455][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:57:43,954][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:57:44,459][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:57:44,959][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:57:45,459][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:57:45,966][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:57:46,470][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:57:46,971][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:57:47,474][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:57:47,975][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:57:48,477][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:57:48,976][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:57:49,475][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:57:49,979][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:57:50,479][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:57:50,979][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:57:51,479][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:57:51,979][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:57:52,477][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:57:52,974][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:57:53,472][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:57:53,969][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:57:54,466][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:57:54,963][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:57:55,460][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:57:55,960][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:57:56,460][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:57:56,961][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:57:57,465][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:57:57,970][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:57:58,481][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:57:58,983][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:57:59,487][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:57:59,991][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:58:00,492][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:58:00,993][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:58:01,493][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:58:01,995][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:58:02,495][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:58:02,996][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:58:03,496][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:58:03,998][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:58:04,503][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:58:05,011][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:58:05,513][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:58:06,018][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:58:06,543][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:58:07,044][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:58:07,545][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:58:08,045][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10071 tokens. +[2026-03-26 08:58:08,949][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.17%, Current % of VRAM taken: 60.59%, Block Peak % of device VRAM: 62.21%, ΔTime: 00:00:32 +[2026-03-26 08:58:09,704][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:58:09,706][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:58:09,708][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:58:10,449][__main__][INFO] - Iteration 763 took 49s (27.27% Gen, 71.24% Train). Generation: 13s, Training: 35s. Estimated remaining time: 29h 36m 18s. Estimated total time: 41h 28m 49s. Time estimates for 10 more iterations: 8m 17s, 100 more iterations: 1h 22m 57s, 500 more iterations: 6h 54m 48s. +[2026-03-26 08:58:10,451][__main__][INFO] - Starting iteration 763. +[2026-03-26 08:58:10,858][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 76 and human policies 1. +[2026-03-26 08:58:10,859][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:58:25,124][__main__][INFO] - Number of regex retries in iteration 763: 0 +[2026-03-26 08:58:25,126][__main__][INFO] - agents played in iteration 763 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:58:25,893][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:58:25,917][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:58:25,941][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:58:25,964][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.55%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:58:25,965][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:58:25,965][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:58:26,824][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:58:27,280][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:58:27,785][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:58:28,283][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:58:28,785][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:58:29,283][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:58:29,781][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:58:30,281][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:58:30,783][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:58:31,292][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:58:31,794][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:58:32,300][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:58:32,808][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:58:33,317][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:58:33,821][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:58:34,324][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:58:34,829][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:58:35,329][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:58:35,829][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:58:36,330][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:58:36,832][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:58:37,331][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:58:37,833][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:58:38,335][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:58:38,833][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:58:39,332][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:58:39,835][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:58:40,334][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:58:40,834][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:58:41,335][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:58:41,833][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:58:42,332][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:58:42,834][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:58:43,334][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:58:43,834][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:58:44,332][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:58:44,836][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:58:45,344][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:58:45,845][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:58:46,352][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:58:46,856][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:58:47,359][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:58:47,858][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:58:48,362][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:58:48,861][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:58:49,361][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:58:49,860][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:58:50,363][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:58:50,867][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:58:51,366][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:58:51,868][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:58:52,367][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:58:52,866][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:58:53,364][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:58:53,866][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:58:54,366][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:58:54,863][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:58:55,361][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:58:55,858][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:58:56,356][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:58:56,854][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:58:57,354][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:58:57,852][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:58:58,349][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:58:58,847][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10160 tokens. +[2026-03-26 08:58:59,766][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.05%, Current % of VRAM taken: 60.48%, Block Peak % of device VRAM: 62.24%, ΔTime: 00:00:32 +[2026-03-26 08:59:00,521][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:59:00,524][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:59:00,525][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:59:01,375][__main__][INFO] - Iteration 764 took 50s (28.24% Gen, 70.07% Train). Generation: 14s, Training: 35s. Estimated remaining time: 30h 12m 32s. Estimated total time: 42h 5m 53s. Time estimates for 10 more iterations: 8m 25s, 100 more iterations: 1h 24m 11s, 500 more iterations: 7h 0m 58s. +[2026-03-26 08:59:01,378][__main__][INFO] - Starting iteration 764. +[2026-03-26 08:59:01,782][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 76 and human policies 1. +[2026-03-26 08:59:01,782][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 08:59:17,328][__main__][INFO] - Number of regex retries in iteration 764: 0 +[2026-03-26 08:59:17,329][__main__][INFO] - agents played in iteration 764 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 08:59:18,099][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:59:18,123][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:59:18,147][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:59:18,171][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.47%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 08:59:18,171][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 08:59:18,172][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 08:59:19,030][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 08:59:19,490][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 08:59:19,995][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 08:59:20,495][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 08:59:20,994][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 08:59:21,492][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 08:59:21,992][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 08:59:22,492][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 08:59:22,992][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 08:59:23,491][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 08:59:23,991][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 08:59:24,490][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 08:59:24,990][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 08:59:25,495][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 08:59:25,993][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 08:59:26,497][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 08:59:26,996][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 08:59:27,502][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 08:59:28,004][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 08:59:28,516][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 08:59:29,022][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 08:59:29,531][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 08:59:30,034][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 08:59:30,534][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 08:59:31,039][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 08:59:31,538][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 08:59:32,040][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 08:59:32,541][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 08:59:33,039][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 08:59:33,538][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 08:59:34,036][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 08:59:34,536][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 08:59:35,037][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 08:59:35,536][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 08:59:36,033][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 08:59:36,531][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 08:59:37,029][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 08:59:37,530][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 08:59:38,030][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 08:59:38,529][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 08:59:39,028][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 08:59:39,528][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 08:59:40,030][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 08:59:40,529][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 08:59:41,029][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 08:59:41,531][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 08:59:42,033][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 08:59:42,532][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 08:59:43,032][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 08:59:43,533][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 08:59:44,034][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 08:59:44,532][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 08:59:45,034][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 08:59:45,534][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 08:59:46,032][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 08:59:46,534][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 08:59:47,034][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 08:59:47,537][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 08:59:48,039][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 08:59:48,536][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 08:59:49,039][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 08:59:49,538][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 08:59:50,046][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 08:59:50,548][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 08:59:51,049][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10185 tokens. +[2026-03-26 08:59:51,966][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.06%, Current % of VRAM taken: 60.49%, Block Peak % of device VRAM: 62.33%, ΔTime: 00:00:32 +[2026-03-26 08:59:52,724][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 08:59:52,727][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 08:59:52,728][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 08:59:53,509][__main__][INFO] - Iteration 765 took 51s (30.06% Gen, 68.43% Train). Generation: 15s, Training: 35s. Estimated remaining time: 31h 12m 10s. Estimated total time: 43h 6m 23s. Time estimates for 10 more iterations: 8m 37s, 100 more iterations: 1h 26m 12s, 500 more iterations: 7h 11m 3s. +[2026-03-26 08:59:53,511][__main__][INFO] - Starting iteration 765. +[2026-03-26 08:59:53,918][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 76 and human policies 1. +[2026-03-26 08:59:53,919][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 09:00:08,946][__main__][INFO] - Number of regex retries in iteration 765: 0 +[2026-03-26 09:00:08,947][__main__][INFO] - agents played in iteration 765 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 09:00:09,715][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 09:00:09,741][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 09:00:09,766][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 09:00:09,796][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.49%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 09:00:09,797][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 09:00:09,797][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 09:00:10,695][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 09:00:11,151][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 09:00:11,661][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 09:00:12,163][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 09:00:12,674][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 09:00:13,175][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 09:00:13,677][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 09:00:14,183][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 09:00:14,688][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 09:00:15,189][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 09:00:15,690][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 09:00:16,197][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 09:00:16,700][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 09:00:17,206][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 09:00:17,711][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 09:00:18,212][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 09:00:18,713][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 09:00:19,216][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 09:00:19,719][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 09:00:20,221][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 09:00:20,727][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 09:00:21,233][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 09:00:21,743][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 09:00:22,246][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 09:00:22,751][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 09:00:23,253][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 09:00:23,754][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 09:00:24,256][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 09:00:24,758][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 09:00:25,258][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 09:00:25,759][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 09:00:26,260][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 09:00:26,760][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 09:00:27,263][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 09:00:27,772][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 09:00:28,280][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 09:00:28,793][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 09:00:29,299][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 09:00:29,802][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 09:00:30,306][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 09:00:30,811][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 09:00:31,311][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 09:00:31,813][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 09:00:32,315][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 09:00:32,816][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 09:00:33,317][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 09:00:33,817][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 09:00:34,321][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 09:00:34,820][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 09:00:35,319][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 09:00:35,818][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 09:00:36,315][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 09:00:36,814][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 09:00:37,316][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 09:00:37,814][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 09:00:38,317][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 09:00:38,821][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 09:00:39,331][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 09:00:39,831][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 09:00:40,331][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 09:00:40,830][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 09:00:41,330][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 09:00:41,829][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 09:00:42,329][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 09:00:42,828][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10110 tokens. +[2026-03-26 09:00:43,768][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.03%, Current % of VRAM taken: 60.46%, Block Peak % of device VRAM: 62.18%, ΔTime: 00:00:33 +[2026-03-26 09:00:44,529][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 09:00:44,531][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 09:00:44,535][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 09:00:45,280][__main__][INFO] - Iteration 766 took 51s (29.26% Gen, 69.29% Train). Generation: 15s, Training: 35s. Estimated remaining time: 30h 53m 0s. Estimated total time: 42h 48m 5s. Time estimates for 10 more iterations: 8m 33s, 100 more iterations: 1h 25m 36s, 500 more iterations: 7h 8m 0s. +[2026-03-26 09:00:45,282][__main__][INFO] - Starting iteration 766. +[2026-03-26 09:00:45,688][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 76 and human policies 1. +[2026-03-26 09:00:45,688][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 09:00:59,994][__main__][INFO] - Number of regex retries in iteration 766: 0 +[2026-03-26 09:00:59,995][__main__][INFO] - agents played in iteration 766 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 09:01:00,764][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 09:01:00,788][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 09:01:00,811][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 09:01:00,835][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 09:01:00,836][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 09:01:00,836][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 09:01:01,700][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 09:01:02,156][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 09:01:02,657][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 09:01:03,155][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 09:01:03,652][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 09:01:04,149][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 09:01:04,646][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 09:01:05,144][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 09:01:05,646][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 09:01:06,155][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 09:01:06,655][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 09:01:07,164][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 09:01:07,665][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 09:01:08,165][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 09:01:08,672][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 09:01:09,172][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 09:01:09,673][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 09:01:10,177][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 09:01:10,678][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 09:01:11,178][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 09:01:11,678][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 09:01:12,178][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 09:01:12,678][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 09:01:13,200][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 09:01:13,703][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 09:01:14,202][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 09:01:14,706][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 09:01:15,206][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 09:01:15,705][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 09:01:16,204][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 09:01:16,703][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 09:01:17,201][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 09:01:17,704][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 09:01:18,201][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 09:01:18,698][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 09:01:19,200][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 09:01:19,699][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 09:01:20,196][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 09:01:20,695][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 09:01:21,193][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 09:01:21,693][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 09:01:22,199][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 09:01:22,700][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 09:01:23,214][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 09:01:23,715][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 09:01:24,216][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 09:01:24,717][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 09:01:25,217][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 09:01:25,718][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 09:01:26,218][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 09:01:26,721][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 09:01:27,231][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 09:01:27,731][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 09:01:28,230][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 09:01:28,732][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 09:01:29,231][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 09:01:29,731][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 09:01:30,231][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 09:01:30,730][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 09:01:31,230][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 09:01:31,729][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 09:01:32,231][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 09:01:32,730][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 09:01:33,232][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 09:01:33,730][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10066 tokens. +[2026-03-26 09:01:34,638][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.13%, Current % of VRAM taken: 60.56%, Block Peak % of device VRAM: 62.15%, ΔTime: 00:00:32 +[2026-03-26 09:01:35,394][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 09:01:35,396][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 09:01:35,398][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 09:01:36,166][__main__][INFO] - Iteration 767 took 50s (28.34% Gen, 70.13% Train). Generation: 14s, Training: 35s. Estimated remaining time: 30h 8m 1s. Estimated total time: 42h 3m 57s. Time estimates for 10 more iterations: 8m 24s, 100 more iterations: 1h 24m 7s, 500 more iterations: 7h 0m 39s. +[2026-03-26 09:01:36,169][__main__][INFO] - Starting iteration 767. +[2026-03-26 09:01:36,583][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 76 and human policies 1. +[2026-03-26 09:01:36,584][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 09:01:41,088][mllm.models.large_language_model_local][WARNING] - Response Proposal: 0 hats, 10 books, 10 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 09:01:50,878][__main__][INFO] - Number of regex retries in iteration 767: 1 +[2026-03-26 09:01:50,879][__main__][INFO] - agents played in iteration 767 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 09:01:51,644][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 09:01:51,667][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 09:01:51,691][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 09:01:51,714][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.52%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 09:01:51,715][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 09:01:51,716][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 09:01:52,580][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 09:01:53,043][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 09:01:53,555][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 09:01:54,059][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 09:01:54,568][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 09:01:55,071][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 09:01:55,578][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 09:01:56,084][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 09:01:56,589][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 09:01:57,090][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 09:01:57,587][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 09:01:58,093][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 09:01:58,592][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 09:01:59,093][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 09:01:59,597][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 09:02:00,097][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 09:02:00,603][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 09:02:01,110][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 09:02:01,612][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 09:02:02,116][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 09:02:02,616][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 09:02:03,120][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 09:02:03,620][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 09:02:04,119][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 09:02:04,618][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 09:02:05,119][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 09:02:05,627][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 09:02:06,127][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 09:02:06,627][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 09:02:07,127][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 09:02:07,630][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 09:02:08,131][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 09:02:08,631][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 09:02:09,131][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 09:02:09,632][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 09:02:10,132][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 09:02:10,632][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 09:02:11,133][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 09:02:11,632][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 09:02:12,132][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 09:02:12,631][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 09:02:13,130][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 09:02:13,627][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 09:02:14,124][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 09:02:14,622][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 09:02:15,126][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 09:02:15,624][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 09:02:16,121][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 09:02:16,618][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 09:02:17,116][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 09:02:17,614][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 09:02:18,112][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 09:02:18,611][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 09:02:19,110][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 09:02:19,609][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 09:02:20,107][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 09:02:20,606][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 09:02:21,107][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 09:02:21,607][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 09:02:22,104][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 09:02:22,600][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 09:02:23,097][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 09:02:23,593][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 09:02:24,095][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 09:02:24,593][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10098 tokens. +[2026-03-26 09:02:25,486][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 60.51%, Block Peak % of device VRAM: 62.18%, ΔTime: 00:00:32 +[2026-03-26 09:02:26,246][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 09:02:26,248][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 09:02:26,250][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 09:02:27,025][__main__][INFO] - Iteration 768 took 50s (28.34% Gen, 70.12% Train). Generation: 14s, Training: 35s. Estimated remaining time: 30h 5m 23s. Estimated total time: 42h 2m 10s. Time estimates for 10 more iterations: 8m 24s, 100 more iterations: 1h 24m 4s, 500 more iterations: 7h 0m 21s. +[2026-03-26 09:02:27,028][__main__][INFO] - Starting iteration 768. +[2026-03-26 09:02:27,436][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 76 and human policies 1. +[2026-03-26 09:02:27,437][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 09:02:40,946][__main__][INFO] - Number of regex retries in iteration 768: 0 +[2026-03-26 09:02:40,946][__main__][INFO] - agents played in iteration 768 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 09:02:41,721][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 09:02:41,745][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 09:02:41,769][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 09:02:41,792][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 09:02:41,793][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 09:02:41,793][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 09:02:42,655][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 09:02:43,114][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 09:02:43,614][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 09:02:44,112][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 09:02:44,609][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 09:02:45,112][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 09:02:45,617][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 09:02:46,124][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 09:02:46,625][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 09:02:47,127][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 09:02:47,632][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 09:02:48,133][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 09:02:48,634][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 09:02:49,134][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 09:02:49,637][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 09:02:50,137][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 09:02:50,637][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 09:02:51,139][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 09:02:51,641][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 09:02:52,147][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 09:02:52,654][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 09:02:53,167][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 09:02:53,668][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 09:02:54,172][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 09:02:54,675][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 09:02:55,176][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 09:02:55,677][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 09:02:56,177][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 09:02:56,677][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 09:02:57,177][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 09:02:57,678][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 09:02:58,181][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 09:02:58,681][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 09:02:59,180][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 09:02:59,682][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 09:03:00,180][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 09:03:00,679][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 09:03:01,177][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 09:03:01,676][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 09:03:02,177][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 09:03:02,675][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 09:03:03,176][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 09:03:03,673][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 09:03:04,172][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 09:03:04,672][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 09:03:05,170][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 09:03:05,668][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 09:03:06,166][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 09:03:06,669][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 09:03:07,171][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 09:03:07,670][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 09:03:08,169][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 09:03:08,668][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 09:03:09,168][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 09:03:09,666][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 09:03:10,166][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 09:03:10,663][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 09:03:11,160][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 09:03:11,658][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 09:03:12,155][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 09:03:12,655][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 09:03:13,153][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 09:03:13,655][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 09:03:14,152][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 09:03:14,651][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10139 tokens. +[2026-03-26 09:03:15,562][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.08%, Current % of VRAM taken: 60.51%, Block Peak % of device VRAM: 62.12%, ΔTime: 00:00:32 +[2026-03-26 09:03:16,312][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 09:03:16,315][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 09:03:16,316][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 09:03:17,055][__main__][INFO] - Iteration 769 took 49s (27.23% Gen, 71.28% Train). Generation: 13s, Training: 35s. Estimated remaining time: 29h 23m 21s. Estimated total time: 41h 20m 58s. Time estimates for 10 more iterations: 8m 16s, 100 more iterations: 1h 22m 41s, 500 more iterations: 6h 53m 29s. +[2026-03-26 09:03:17,057][__main__][INFO] - Starting iteration 769. +[2026-03-26 09:03:17,469][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 76 and human policies 1. +[2026-03-26 09:03:17,469][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 09:03:21,906][mllm.models.large_language_model_local][WARNING] - Response Proposal: 10 hats, 0 books, 0 balls did not match regex: (?i)Proposal:\s*((?:\s*(?P(10|[0-9]))\s*(?Phats?|books?|balls?)\s*,?)+), retry 1/3 +[2026-03-26 09:03:31,905][__main__][INFO] - Number of regex retries in iteration 769: 1 +[2026-03-26 09:03:31,906][__main__][INFO] - agents played in iteration 769 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 09:03:32,679][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 09:03:32,703][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 09:03:32,726][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 09:03:32,749][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.48%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 09:03:32,750][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 09:03:32,751][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 09:03:33,620][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 09:03:34,080][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 09:03:34,585][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 09:03:35,083][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 09:03:35,580][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 09:03:36,081][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 09:03:36,592][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 09:03:37,101][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 09:03:37,605][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 09:03:38,105][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 09:03:38,605][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 09:03:39,106][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 09:03:39,607][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 09:03:40,110][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 09:03:40,613][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 09:03:41,112][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 09:03:41,614][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 09:03:42,114][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 09:03:42,615][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 09:03:43,114][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 09:03:43,614][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 09:03:44,114][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 09:03:44,620][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 09:03:45,123][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 09:03:45,622][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 09:03:46,121][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 09:03:46,619][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 09:03:47,121][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 09:03:47,619][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 09:03:48,116][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 09:03:48,614][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 09:03:49,113][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 09:03:49,613][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 09:03:50,119][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 09:03:50,618][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 09:03:51,123][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 09:03:51,632][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 09:03:52,133][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 09:03:52,636][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 09:03:53,137][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 09:03:53,635][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 09:03:54,134][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 09:03:54,637][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 09:03:55,136][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 09:03:55,638][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 09:03:56,139][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 09:03:56,638][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 09:03:57,137][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 09:03:57,635][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 09:03:58,134][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 09:03:58,632][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 09:03:59,131][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 09:03:59,630][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 09:04:00,130][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 09:04:00,628][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 09:04:01,128][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 09:04:01,626][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 09:04:02,129][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 09:04:02,629][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 09:04:03,129][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 09:04:03,629][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 09:04:04,129][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 09:04:04,629][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 09:04:05,130][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 09:04:05,629][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10121 tokens. +[2026-03-26 09:04:06,549][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.15%, Current % of VRAM taken: 60.58%, Block Peak % of device VRAM: 62.14%, ΔTime: 00:00:32 +[2026-03-26 09:04:07,301][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 09:04:07,303][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 09:04:07,305][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 09:04:08,080][__main__][INFO] - Iteration 770 took 50s (28.52% Gen, 69.94% Train). Generation: 14s, Training: 35s. Estimated remaining time: 30h 12m 7s. Estimated total time: 42h 10m 34s. Time estimates for 10 more iterations: 8m 26s, 100 more iterations: 1h 24m 21s, 500 more iterations: 7h 1m 45s. +[2026-03-26 09:04:08,082][__main__][INFO] - Starting iteration 770. +[2026-03-26 09:04:08,490][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 76 and human policies 1. +[2026-03-26 09:04:08,490][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 09:04:21,969][__main__][INFO] - Number of regex retries in iteration 770: 0 +[2026-03-26 09:04:21,970][__main__][INFO] - agents played in iteration 770 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 09:04:22,727][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 09:04:22,750][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 09:04:22,774][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 09:04:22,798][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.51%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 09:04:22,799][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 09:04:22,799][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 09:04:23,665][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 09:04:24,124][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 09:04:24,628][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 09:04:25,127][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 09:04:25,625][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 09:04:26,128][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 09:04:26,626][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 09:04:27,126][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 09:04:27,629][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 09:04:28,127][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 09:04:28,625][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 09:04:29,124][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 09:04:29,621][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 09:04:30,123][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 09:04:30,623][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 09:04:31,121][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 09:04:31,619][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 09:04:32,125][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 09:04:32,624][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 09:04:33,126][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 09:04:33,624][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 09:04:34,127][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 09:04:34,633][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 09:04:35,133][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 09:04:35,633][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 09:04:36,132][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 09:04:36,635][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 09:04:37,135][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 09:04:37,635][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 09:04:38,137][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 09:04:38,638][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 09:04:39,140][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 09:04:39,642][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 09:04:40,141][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 09:04:40,641][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 09:04:41,140][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 09:04:41,638][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 09:04:42,135][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 09:04:42,633][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 09:04:43,132][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 09:04:43,632][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 09:04:44,134][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 09:04:44,635][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 09:04:45,136][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 09:04:45,638][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 09:04:46,137][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 09:04:46,639][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 09:04:47,137][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 09:04:47,635][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 09:04:48,133][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 09:04:48,632][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 09:04:49,131][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 09:04:49,630][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 09:04:50,127][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 09:04:50,623][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 09:04:51,124][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 09:04:51,620][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 09:04:52,124][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 09:04:52,624][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 09:04:53,123][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 09:04:53,628][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 09:04:54,129][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 09:04:54,640][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 09:04:55,141][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 09:04:55,641][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 10108 tokens. +[2026-03-26 09:04:56,550][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.14%, Current % of VRAM taken: 60.56%, Block Peak % of device VRAM: 62.06%, ΔTime: 00:00:32 +[2026-03-26 09:04:57,306][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 09:04:57,308][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 09:04:57,310][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 09:04:58,630][__main__][INFO] - Iteration 771 took 50s (26.88% Gen, 70.48% Train). Generation: 13s, Training: 35s. Estimated remaining time: 29h 47m 46s. Estimated total time: 41h 47m 4s. Time estimates for 10 more iterations: 8m 21s, 100 more iterations: 1h 23m 34s, 500 more iterations: 6h 57m 50s. +[2026-03-26 09:04:58,633][__main__][INFO] - Starting iteration 771. +[2026-03-26 09:04:59,044][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 77 and human policies 1. +[2026-03-26 09:04:59,044][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 +[2026-03-26 09:05:13,851][__main__][INFO] - Number of regex retries in iteration 771: 0 +[2026-03-26 09:05:13,852][__main__][INFO] - agents played in iteration 771 are Alice_buffer, Bob_buffer, Bob, Alice +[2026-03-26 09:05:14,626][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 09:05:14,649][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 09:05:14,673][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 09:05:14,696][mllm.training.trainer_ad_align][INFO] - For task: Get advantages with critic gradient accumulation, ΔVRAM % (total): 0.00%, Current % of VRAM taken: 53.53%, Block Peak % of device VRAM: 19.52%, ΔTime: 00:00:00 +[2026-03-26 09:05:14,697][mllm.training.trainer_ad_align][INFO] - Sharing advantage alignment data. +[2026-03-26 09:05:14,697][mllm.training.trainer_ad_align][INFO] - Receiving advantage packets. +[2026-03-26 09:05:15,568][mllm.training.trainer_common][INFO] - Processing mini-batch 0 of 64 +[2026-03-26 09:05:16,023][mllm.training.trainer_common][INFO] - Processing mini-batch 1 of 64 +[2026-03-26 09:05:16,522][mllm.training.trainer_common][INFO] - Processing mini-batch 2 of 64 +[2026-03-26 09:05:17,023][mllm.training.trainer_common][INFO] - Processing mini-batch 3 of 64 +[2026-03-26 09:05:17,519][mllm.training.trainer_common][INFO] - Processing mini-batch 4 of 64 +[2026-03-26 09:05:18,014][mllm.training.trainer_common][INFO] - Processing mini-batch 5 of 64 +[2026-03-26 09:05:18,510][mllm.training.trainer_common][INFO] - Processing mini-batch 6 of 64 +[2026-03-26 09:05:19,007][mllm.training.trainer_common][INFO] - Processing mini-batch 7 of 64 +[2026-03-26 09:05:19,504][mllm.training.trainer_common][INFO] - Processing mini-batch 8 of 64 +[2026-03-26 09:05:20,008][mllm.training.trainer_common][INFO] - Processing mini-batch 9 of 64 +[2026-03-26 09:05:20,523][mllm.training.trainer_common][INFO] - Processing mini-batch 10 of 64 +[2026-03-26 09:05:21,028][mllm.training.trainer_common][INFO] - Processing mini-batch 11 of 64 +[2026-03-26 09:05:21,529][mllm.training.trainer_common][INFO] - Processing mini-batch 12 of 64 +[2026-03-26 09:05:22,031][mllm.training.trainer_common][INFO] - Processing mini-batch 13 of 64 +[2026-03-26 09:05:22,533][mllm.training.trainer_common][INFO] - Processing mini-batch 14 of 64 +[2026-03-26 09:05:23,033][mllm.training.trainer_common][INFO] - Processing mini-batch 15 of 64 +[2026-03-26 09:05:23,533][mllm.training.trainer_common][INFO] - Processing mini-batch 16 of 64 +[2026-03-26 09:05:24,035][mllm.training.trainer_common][INFO] - Processing mini-batch 17 of 64 +[2026-03-26 09:05:24,537][mllm.training.trainer_common][INFO] - Processing mini-batch 18 of 64 +[2026-03-26 09:05:25,038][mllm.training.trainer_common][INFO] - Processing mini-batch 19 of 64 +[2026-03-26 09:05:25,542][mllm.training.trainer_common][INFO] - Processing mini-batch 20 of 64 +[2026-03-26 09:05:26,061][mllm.training.trainer_common][INFO] - Processing mini-batch 21 of 64 +[2026-03-26 09:05:26,564][mllm.training.trainer_common][INFO] - Processing mini-batch 22 of 64 +[2026-03-26 09:05:27,065][mllm.training.trainer_common][INFO] - Processing mini-batch 23 of 64 +[2026-03-26 09:05:27,567][mllm.training.trainer_common][INFO] - Processing mini-batch 24 of 64 +[2026-03-26 09:05:28,071][mllm.training.trainer_common][INFO] - Processing mini-batch 25 of 64 +[2026-03-26 09:05:28,572][mllm.training.trainer_common][INFO] - Processing mini-batch 26 of 64 +[2026-03-26 09:05:29,075][mllm.training.trainer_common][INFO] - Processing mini-batch 27 of 64 +[2026-03-26 09:05:29,577][mllm.training.trainer_common][INFO] - Processing mini-batch 28 of 64 +[2026-03-26 09:05:30,078][mllm.training.trainer_common][INFO] - Processing mini-batch 29 of 64 +[2026-03-26 09:05:30,584][mllm.training.trainer_common][INFO] - Processing mini-batch 30 of 64 +[2026-03-26 09:05:31,084][mllm.training.trainer_common][INFO] - Processing mini-batch 31 of 64 +[2026-03-26 09:05:31,583][mllm.training.trainer_common][INFO] - Processing mini-batch 32 of 64 +[2026-03-26 09:05:32,082][mllm.training.trainer_common][INFO] - Processing mini-batch 33 of 64 +[2026-03-26 09:05:32,580][mllm.training.trainer_common][INFO] - Processing mini-batch 34 of 64 +[2026-03-26 09:05:33,077][mllm.training.trainer_common][INFO] - Processing mini-batch 35 of 64 +[2026-03-26 09:05:33,575][mllm.training.trainer_common][INFO] - Processing mini-batch 36 of 64 +[2026-03-26 09:05:34,072][mllm.training.trainer_common][INFO] - Processing mini-batch 37 of 64 +[2026-03-26 09:05:34,570][mllm.training.trainer_common][INFO] - Processing mini-batch 38 of 64 +[2026-03-26 09:05:35,067][mllm.training.trainer_common][INFO] - Processing mini-batch 39 of 64 +[2026-03-26 09:05:35,564][mllm.training.trainer_common][INFO] - Processing mini-batch 40 of 64 +[2026-03-26 09:05:36,066][mllm.training.trainer_common][INFO] - Processing mini-batch 41 of 64 +[2026-03-26 09:05:36,563][mllm.training.trainer_common][INFO] - Processing mini-batch 42 of 64 +[2026-03-26 09:05:37,069][mllm.training.trainer_common][INFO] - Processing mini-batch 43 of 64 +[2026-03-26 09:05:37,568][mllm.training.trainer_common][INFO] - Processing mini-batch 44 of 64 +[2026-03-26 09:05:38,069][mllm.training.trainer_common][INFO] - Processing mini-batch 45 of 64 +[2026-03-26 09:05:38,567][mllm.training.trainer_common][INFO] - Processing mini-batch 46 of 64 +[2026-03-26 09:05:39,066][mllm.training.trainer_common][INFO] - Processing mini-batch 47 of 64 +[2026-03-26 09:05:39,569][mllm.training.trainer_common][INFO] - Processing mini-batch 48 of 64 +[2026-03-26 09:05:40,069][mllm.training.trainer_common][INFO] - Processing mini-batch 49 of 64 +[2026-03-26 09:05:40,568][mllm.training.trainer_common][INFO] - Processing mini-batch 50 of 64 +[2026-03-26 09:05:41,070][mllm.training.trainer_common][INFO] - Processing mini-batch 51 of 64 +[2026-03-26 09:05:41,568][mllm.training.trainer_common][INFO] - Processing mini-batch 52 of 64 +[2026-03-26 09:05:42,070][mllm.training.trainer_common][INFO] - Processing mini-batch 53 of 64 +[2026-03-26 09:05:42,568][mllm.training.trainer_common][INFO] - Processing mini-batch 54 of 64 +[2026-03-26 09:05:43,066][mllm.training.trainer_common][INFO] - Processing mini-batch 55 of 64 +[2026-03-26 09:05:43,564][mllm.training.trainer_common][INFO] - Processing mini-batch 56 of 64 +[2026-03-26 09:05:44,063][mllm.training.trainer_common][INFO] - Processing mini-batch 57 of 64 +[2026-03-26 09:05:44,561][mllm.training.trainer_common][INFO] - Processing mini-batch 58 of 64 +[2026-03-26 09:05:45,058][mllm.training.trainer_common][INFO] - Processing mini-batch 59 of 64 +[2026-03-26 09:05:45,557][mllm.training.trainer_common][INFO] - Processing mini-batch 60 of 64 +[2026-03-26 09:05:46,056][mllm.training.trainer_common][INFO] - Processing mini-batch 61 of 64 +[2026-03-26 09:05:46,554][mllm.training.trainer_common][INFO] - Processing mini-batch 62 of 64 +[2026-03-26 09:05:47,052][mllm.training.trainer_common][INFO] - Processing mini-batch 63 of 64 +[2026-03-26 09:05:47,551][mllm.training.trainer_common][INFO] - Accumulated the policy gradient loss for 9996 tokens. +[2026-03-26 09:05:48,467][mllm.training.trainer_common][INFO] - For task: Apply reinforce step, ΔVRAM % (total): 14.07%, Current % of VRAM taken: 60.50%, Block Peak % of device VRAM: 62.10%, ΔTime: 00:00:32 +[2026-03-26 09:05:49,212][mllm.training.trainer_common][INFO] - Saved main optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/policy_optimizer_state.pt +[2026-03-26 09:05:49,215][mllm.training.trainer_common][INFO] - Saved critic optimizer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/critic_optimizer_state.pt +[2026-03-26 09:05:49,216][mllm.training.trainer_common][INFO] - Saved trainer state to /scratch/muqeeth/llm_negotiation/2026_03/split_no_comm_vanilla_ad_align_seed42/seed_42/agent_trainer/trainer_annealing_state.pkl +[2026-03-26 09:05:49,977][__main__][INFO] - Iteration 772 took 50s (29.07% Gen, 69.43% Train). Generation: 14s, Training: 35s. Estimated remaining time: 30h 26m 31s. Estimated total time: 42h 26m 41s. Time estimates for 10 more iterations: 8m 29s, 100 more iterations: 1h 24m 53s, 500 more iterations: 7h 4m 26s. +[2026-03-26 09:05:49,990][__main__][INFO] - Starting iteration 772. +[2026-03-26 09:05:50,405][__main__][INFO] - Inference policies count is regular policies 2 and buffer policies 77 and human policies 1. +[2026-03-26 09:05:50,406][__main__][INFO] - Hard coded buffer agents are set to False with prob 0 diff --git a/src_code_for_reproducibility/__pycache__/__init__.cpython-312.pyc b/src_code_for_reproducibility/__pycache__/__init__.cpython-312.pyc index e1bb76ab4b9c5bb44943e5eabe2ca3dadde518d2..1e0ce601325acab2a3935c9b22e10e27fcccc961 100644 Binary files a/src_code_for_reproducibility/__pycache__/__init__.cpython-312.pyc and b/src_code_for_reproducibility/__pycache__/__init__.cpython-312.pyc differ diff --git a/src_code_for_reproducibility/chat_utils/apply_template.py b/src_code_for_reproducibility/chat_utils/apply_template.py index f143ab15e6b30637700df3381b7de686153017cd..6bbdc32dbb1df0407ff24ae90395dba0d162bf7d 100644 --- a/src_code_for_reproducibility/chat_utils/apply_template.py +++ b/src_code_for_reproducibility/chat_utils/apply_template.py @@ -1,10 +1,17 @@ +""" +File: mllm/chat_utils/apply_template.py +Summary: Applies tokenizer-specific chat templates and stitches chat token IDs. +""" + import torch from mllm.chat_utils.chat_turn import ChatTurn from mllm.chat_utils.template_specific import ( + custom_gemma3_template, custom_llama3_template, custom_qwen2_template, custom_qwen3_template, + gemma3_assistant_postfix, qwen2_assistant_postfix, qwen3_assistant_postfix, ) @@ -20,6 +27,8 @@ def get_custom_chat_template(tokenizer) -> str: return custom_llama3_template elif "qwen3" in tokenizer.name_or_path.lower(): return custom_qwen3_template + elif "gemma" in tokenizer.name_or_path.lower(): + return custom_gemma3_template else: raise ValueError(f"Tokenizer {tokenizer.name_or_path} not supported") @@ -32,13 +41,15 @@ def get_custom_assistant_postfix(tokenizer) -> torch.Tensor: return qwen2_assistant_postfix elif "qwen3" in tokenizer.name_or_path.lower(): return qwen3_assistant_postfix + elif "gemma" in tokenizer.name_or_path.lower(): + return gemma3_assistant_postfix return torch.tensor([], dtype=torch.long) def tokenize_chats(chats: list[ChatTurn], tokenizer, enable_thinking) -> None: """ Set the chat_template_token_ids for each chat turn. - # TODO: use engine tokens if available + We rely on tokenizer-side templates because engine-provided cached tokens are not exposed yet. """ custom_template = get_custom_chat_template(tokenizer) custom_assistant_postfix: torch.Tensor = get_custom_assistant_postfix(tokenizer) diff --git a/src_code_for_reproducibility/chat_utils/chat_turn.py b/src_code_for_reproducibility/chat_utils/chat_turn.py index d22c1b853def374fa0ee21eb523f2b9d104ed35b..cfc0d9422a6070c86b1da8abce17ad28816fb2eb 100644 --- a/src_code_for_reproducibility/chat_utils/chat_turn.py +++ b/src_code_for_reproducibility/chat_utils/chat_turn.py @@ -1,3 +1,8 @@ +""" +File: mllm/chat_utils/chat_turn.py +Summary: Defines the ChatTurn schema plus helpers for serialization and validation. +""" + from __future__ import annotations import json diff --git a/src_code_for_reproducibility/chat_utils/template_specific.py b/src_code_for_reproducibility/chat_utils/template_specific.py index 5d8f8e5f1be6bcafd1b9e1a465c446253b985e3e..c22328455c55f0b0a02439efdacf6b09234d7981 100644 --- a/src_code_for_reproducibility/chat_utils/template_specific.py +++ b/src_code_for_reproducibility/chat_utils/template_specific.py @@ -1,3 +1,8 @@ +""" +File: mllm/chat_utils/template_specific.py +Summary: Stores chat template variants and assistant postfix tensors per tokenizer. +""" + import huggingface_hub import torch from transformers import AutoTokenizer @@ -25,6 +30,11 @@ qwen3_assistant_postfix = ( .encode("\n", return_tensors="pt") .flatten() ) +gemma3_assistant_postfix = ( + AutoTokenizer.from_pretrained("google/gemma-3-4b-it") + .encode("\n", return_tensors="pt") + .flatten() +) custom_qwen2_template = """ {%- if add_system_prompt %} {{- '<|im_start|>system\nYou are Qwen, created by Alibaba Cloud. You are a helpful assistant.<|im_end|>\n' }} @@ -85,3 +95,20 @@ custom_qwen3_template = """ {%- endif %} {%- endif %} """ + +custom_gemma3_template = """ +{%- if add_system_prompt %} +{{- bos_token -}} +{%- endif %} +{%- for message in messages -%} +{%- if message['role'] == 'assistant' -%} +{%- set role = 'model' -%} +{%- else -%} +{%- set role = message['role'] -%} +{%- endif -%} +{{ '' + role + '\n' + message['content'] | trim + '\n' }} +{%- endfor -%} +{%- if add_generation_prompt -%} +{{ 'model\n' }} +{%- endif -%} +""" diff --git a/src_code_for_reproducibility/markov_games/__init__.py b/src_code_for_reproducibility/markov_games/__init__.py index e69de29bb2d1d6434b8b29ae775ad8c2e48c5391..2a7015344d8ac9b53f2660d4e837f709908213db 100644 --- a/src_code_for_reproducibility/markov_games/__init__.py +++ b/src_code_for_reproducibility/markov_games/__init__.py @@ -0,0 +1,4 @@ +""" +File: mllm/markov_games/__init__.py +Summary: Makes Markov-game subpackages importable from the top-level namespace. +""" diff --git a/src_code_for_reproducibility/markov_games/agent.py b/src_code_for_reproducibility/markov_games/agent.py index 9961d860fecca7eaa4d13cb195b50dc658be5bec..56406ae2695ce97ad7fa4fc436908904ee11be9f 100644 --- a/src_code_for_reproducibility/markov_games/agent.py +++ b/src_code_for_reproducibility/markov_games/agent.py @@ -1,11 +1,6 @@ """ -In simple RL paradise, where the action dimensions are constant and well defined, -Agent classes are not necessary. But in MARL, with LLM's, there isn't always -a direct path from policy to action. For instance, from the observation of the environment, -a prompt must be created. Then, the outputs of the policy might be incorrect, so a second -request to the LLM must be sent before the action is well defined. This is why this Agent class exists. -It acts as a mini environment, bridging the gap between the core simulation and -the LLM policies. +File: mllm/markov_games/agent.py +Summary: Declares the base Agent interface connecting simulations to policy calls. """ from abc import ABC, abstractmethod @@ -18,6 +13,8 @@ from mllm.markov_games.rollout_tree import AgentActLog class Agent(ABC): + """Abstract policy wrapper that bridges simulations with arbitrary backends.""" + @abstractmethod def __init__( self, @@ -29,7 +26,10 @@ class Agent(ABC): **kwargs, ): """ - Initialize the agent state. + Initialize the agent state and seed its RNG. + + Subclasses typically store extra handles (tokenizers, inference clients, etc.) + but they should always call ``super().__init__`` so sampling remains reproducible. """ self.seed = seed self.agent_id = agent_id @@ -40,37 +40,33 @@ class Agent(ABC): async def act(self, observation) -> Tuple[Any, AgentActLog]: """ - Query (possibly multiple times) a policy (or possibly a pool of policies) to - obtain the action of the agent. + Produce the next action (and associated chat log) given an environment observation. - Example: - action = None - prompt = self.observation_to_prompt(observation) - while not self.valid(action): - output = await self.policy.generate(prompt) - action = self.policy_output_to_action(output) - return action - - Returns: - action - step_info + Implementations can iterate with rejection sampling, multi-call deliberation, etc. + Returns both the chosen action and an `AgentActLog` describing how it was produced. """ raise NotImplementedError def get_safe_copy(self): """ - Return copy of the agent object that is decorrelated from the original object. + Return a deep copy whose future calls do not mutate the original agent. + + Needed for branch exploration/reruns with alternative actions. """ raise NotImplementedError def reset(self): + """Reset any internal state between rollouts.""" raise NotImplementedError def render(self): + """Optional human-readable visualization of the agent (CLI/UI).""" raise NotImplementedError def close(self): + """Release any external resources (network sockets, subprocesses, etc.).""" raise NotImplementedError def get_agent_info(self): + """Return diagnostic metadata to embed inside rollout logs.""" raise NotImplementedError diff --git a/src_code_for_reproducibility/markov_games/alternative_actions_runner.py b/src_code_for_reproducibility/markov_games/alternative_actions_runner.py index c64db2deda539a1a71e045309cfdf257d2cbc614..d5165a2552019aefdf281c2bd41e50d204713921 100644 --- a/src_code_for_reproducibility/markov_games/alternative_actions_runner.py +++ b/src_code_for_reproducibility/markov_games/alternative_actions_runner.py @@ -1,3 +1,8 @@ +""" +File: mllm/markov_games/alternative_actions_runner.py +Summary: Generates rollout branches by replaying trajectories with unilateral action changes. +""" + import asyncio import copy import json @@ -16,7 +21,6 @@ from mllm.markov_games.rollout_tree import ( AgentId = str - async def run_with_unilateral_alt_action( markov_game: MarkovGame, agent_id: AgentId, @@ -25,7 +29,11 @@ async def run_with_unilateral_alt_action( max_depth: int, ): """ - This function is used to generate a new branch for a given agent. + Roll out a counterfactual branch where ``agent_id`` deviates unilaterally. + + Starting from ``branch_node`` (which already contains the main trajectory), + we replay the simulation with the deviating agent's action while freezing + all other agents/actions, then continue for ``max_depth`` steps. """ # Generate alternative action and take a step @@ -65,20 +73,20 @@ async def AlternativeActionsRunner( branch_only_on_new_round: bool = False, ): """ - This method generates a trajectory with partially completed branches, - where the branching comes from taking unilateraly different actions. - The resulting data is used to estimate the updated advantage alignment policy gradient terms. - Let k := nb_sub_steps. Then the number of steps generated is O(Tk), where T is - the maximum trajectory length. + Generate a rollout tree containing the main path plus unilateral deviation branches. + + For each timestep we: + 1. Cache agent actions without side effects. + 2. Advance the main trajectory. + 3. Spawn ``nb_alternative_actions`` asynchronous deviations per agent, + each replaying up to ``max_depth`` steps from the cached pre-action state. + The resulting branches feed advantage-alignment estimators. """ tasks = [] time_step = 0 terminated = False - root = RolloutTreeRootNode( - id=markov_game.get_id(), - crn_id=markov_game.get_crn_id() - ) + root = RolloutTreeRootNode(id=markov_game.get_id(), crn_id=markov_game.get_crn_id()) previous_node = root while not terminated: diff --git a/src_code_for_reproducibility/markov_games/group_timesteps.py b/src_code_for_reproducibility/markov_games/group_timesteps.py index dad5271c500f539f2719110bd676f183746e51e4..48b5882a632ba858787befaac306195af959b376 100644 --- a/src_code_for_reproducibility/markov_games/group_timesteps.py +++ b/src_code_for_reproducibility/markov_games/group_timesteps.py @@ -1,6 +1,8 @@ """ -This module contains the logic for grouping time steps. +File: mllm/markov_games/group_timesteps.py +Summary: Provides timestep-grouping utilities for rollout trees and training. """ + import copy from typing import Callable @@ -84,25 +86,6 @@ def group_time_steps( raise Exception( "Grouping timesteps by round is not supported for branching trajectories yet." ) - # Special recursive case for branches - # if isinstance(current_node, RolloutTreeBranchNode): - # branches = {} - # for agent_id, branch_nodes in current_node.branches.items(): - # branch_group_nodes = [] - # for branch_node in branch_nodes: - # branch_group_node = group_time_steps_rec( - # current_node=branch_node, - # group_time_step=group_time_step, - # accumulation_step_logs=copy.deepcopy(accumulation_step_logs)) - # branch_group_nodes.append(branch_group_node) - # branches[agent_id] = branch_group_nodes - - # main_child_group_node = group_time_steps_rec( - # current_node=current_node.main_child, - # group_time_step=group_time_step, - # accumulation_step_logs=copy.deepcopy(accumulation_step_logs)) - - # return RolloutTreeBranchNode(main_child=main_child_group_node, branches=branches) # Accumulate accumulation_step_logs.append(current_node.step_log) diff --git a/src_code_for_reproducibility/markov_games/linear_runner.py b/src_code_for_reproducibility/markov_games/linear_runner.py index f81ab6d37ae9c680b2f2a53388988117d37f8a47..e3e14a3eda72cf4e620db5ab8ed0d3f7d552e9fe 100644 --- a/src_code_for_reproducibility/markov_games/linear_runner.py +++ b/src_code_for_reproducibility/markov_games/linear_runner.py @@ -1,3 +1,8 @@ +""" +File: mllm/markov_games/linear_runner.py +Summary: Simulates a single unbranched Markov-game rollout and records it. +""" + import asyncio import json import os.path @@ -10,7 +15,14 @@ async def LinearRunner( markov_game: MarkovGame, output_folder: str ) -> RolloutTreeRootNode: """ - This method generates a trajectory without branching. + Generate a single main-path rollout (no branching) for the provided Markov game. + + Parameters + ---------- + markov_game: + Initialized ``MarkovGame`` with agents + simulation ready to step. + output_folder: + Unused placeholder in the legacy API (kept for compatibility). """ time_step = 0 terminated = False diff --git a/src_code_for_reproducibility/markov_games/markov_game.py b/src_code_for_reproducibility/markov_games/markov_game.py index 73a48213bddcf0a59976fa0870eec19f59ae47d9..7964fd69d24f617c76e36f852491b1e6141f6c48 100644 --- a/src_code_for_reproducibility/markov_games/markov_game.py +++ b/src_code_for_reproducibility/markov_games/markov_game.py @@ -1,18 +1,8 @@ """ -This class unifies a simulation, and the agents acting in it (see `simulation.py` & `agent.py`). -In a MarkovGame step, - 1) each agent takes an action, - 2) the state transitions with respect to these actions, - 3) all relevant data of the step is appended to the historical data list - -In order to perform 3), the agents and the simulation are expected, at each time step, -to return a log of the state transition (from their perspective). -For instance, the Simulation might send rewards and the agents might send prompting contexts to be used later to generate the training data. -A different approach would be to simply have the agents keep their data private and log it upon completion of a trajectory. -The approach we use here centralizes the data gathering aspect, -making it easy to create sub-trajectories (in the `runners` defined in `runners.py`) descriptions that -only log information for step transitions occuring after the branching out. +File: mllm/markov_games/markov_game.py +Summary: Defines the MarkovGame base class plus shared simulation interfaces. """ + import asyncio import copy import json @@ -31,6 +21,8 @@ AgentId = str @dataclass class AgentAndActionSafeCopy: + """Snapshot of an agent, its action, and metadata used for branch replay.""" + action: Any action_info: AgentActLog agent_after_action: type[Agent] @@ -45,12 +37,18 @@ class MarkovGame(object): crn_id: int, ): """ - Args: - agents: - output_path: - Path where the step infos are saved. - simulation: - Simulation object. Example: IPDSimulation + Initialize the Markov game wrapper. + + Parameters + ---------- + id: + Unique rollout identifier (logged into rollout trees). + agents: + Mapping of agent_id -> Agent instance. + simulation: + Environment implementing the ``Simulation`` interface (IPD, TAS, etc.). + crn_id: + Identifier for the common random number stream used by this rollout. """ self.agents = agents self.agent_ids = self.agents.keys() @@ -131,7 +129,7 @@ class MarkovGame(object): async def set_action_of_agent(self, agent_id: AgentId): """ - TOWRITE + Query a single agent for its next action and store the result locally. """ agent = self.agents[agent_id] obs = self.simulation.get_obs_agent(agent_id) @@ -141,7 +139,7 @@ class MarkovGame(object): async def set_actions(self): """ - TOWRITE + Query every agent concurrently and populate the cached actions/logs. """ # background_tasks = set() tasks = [] @@ -152,16 +150,27 @@ class MarkovGame(object): def take_simulation_step(self): """ - TOWRITE + Advance the simulation by one step using the cached actions. """ terminated, self.simulation_step_log = self.simulation.step(self.actions) return terminated def get_step_log(self) -> StepLog: """ - TOWRITE - TODO: assert actions and simulation have taken step + Package the most recent simulation step and agent logs into a StepLog. """ + if self.simulation_step_log is None: + raise RuntimeError( + "Simulation step log is empty; call take_simulation_step() first." + ) + missing_logs = [ + agent_id for agent_id, log in self.agent_step_logs.items() if log is None + ] + if missing_logs: + raise RuntimeError( + f"Agent action logs missing for: {', '.join(missing_logs)}. " + "Ensure set_actions() ran before requesting the step log." + ) step_log = StepLog( simulation_step_log=self.simulation_step_log, action_logs=self.agent_step_logs, @@ -170,7 +179,7 @@ class MarkovGame(object): async def step(self) -> Tuple[bool, StepLog]: """ - TOWRITE + Convenience step that collects actions, advances the simulation, and returns the log. """ await self.set_actions() terminated = self.take_simulation_step() @@ -179,7 +188,7 @@ class MarkovGame(object): def get_safe_copy(self): """ - TOWRITE + Create a shallow copy of the game with deep-copied agents/simulation for branching. """ new_markov_game = copy.copy(self) diff --git a/src_code_for_reproducibility/markov_games/mg_utils.py b/src_code_for_reproducibility/markov_games/mg_utils.py index 27e0eb606ea6cfd5682c3b21a53e783c23e9b43f..4fc406cd1f0cba593daad1108de2746b6a1d7678 100644 --- a/src_code_for_reproducibility/markov_games/mg_utils.py +++ b/src_code_for_reproducibility/markov_games/mg_utils.py @@ -1,9 +1,18 @@ +""" +File: mllm/markov_games/mg_utils.py +Summary: Holds miscellaneous helpers shared across Markov-game modules. +""" + import asyncio import copy from collections.abc import Callable from dataclasses import dataclass from mllm.markov_games.ipd.ipd_agent import IPDAgent +from mllm.markov_games.ipd.Ipd_hard_coded_agents import ( + AlwaysCooperateIPDAgent, + AlwaysDefectIPDAgent, +) from mllm.markov_games.ipd.ipd_simulation import IPD from mllm.markov_games.markov_game import MarkovGame from mllm.markov_games.negotiation.dond_agent import DealNoDealAgent @@ -12,17 +21,10 @@ from mllm.markov_games.negotiation.nego_hard_coded_policies import ( HardCodedNegoGreedyPolicy, HardCodedNegoWelfareMaximizingPolicy, ) -from mllm.markov_games.ipd.Ipd_hard_coded_agents import AlwaysCooperateIPDAgent, AlwaysDefectIPDAgent from mllm.markov_games.negotiation.no_press_nego_agent import NoPressAgent from mllm.markov_games.negotiation.no_press_nego_simulation import NoPressSimulation -from mllm.markov_games.negotiation.tas_agent import TrustAndSplitAgent from mllm.markov_games.negotiation.tas_rps_agent import TrustAndSplitRPSAgent from mllm.markov_games.negotiation.tas_rps_simulation import TrustAndSplitRPSSimulation -from mllm.markov_games.negotiation.tas_simple_agent import TrustAndSplitSimpleAgent -from mllm.markov_games.negotiation.tas_simple_simulation import ( - TrustAndSplitSimpleSimulation, -) -from mllm.markov_games.negotiation.tas_simulation import TrustAndSplitSimulation from mllm.markov_games.rollout_tree import ( AgentActLog, RolloutTreeBranchNode, @@ -37,6 +39,8 @@ AgentId = str @dataclass class AgentConfig: + """Configuration blob describing one agent in a Markov game spec.""" + agent_id: str agent_name: str agent_class_name: str @@ -46,6 +50,8 @@ class AgentConfig: @dataclass class MarkovGameConfig: + """Top-level config that ties together simulation settings and agent configs.""" + id: int seed: int simulation_class_name: str @@ -57,7 +63,9 @@ def init_markov_game_components( config: MarkovGameConfig, policies: dict[str, Callable[[list[dict]], str]] ): """ - TOWRITE + Materialize Agents and the Simulation described by ``config`` and return a MarkovGame. + + `policies` is a mapping of policy_id -> callable retrieved from the hosting trainer. """ agents = {} agent_names = [] diff --git a/src_code_for_reproducibility/markov_games/negotiation/dond_simulation.py b/src_code_for_reproducibility/markov_games/negotiation/dond_simulation.py index a27d6ce2cf7e31a0cddd341db39ae7898b086115..d941fff41dd68d0d675dcffc31720ba20a7efb4f 100644 --- a/src_code_for_reproducibility/markov_games/negotiation/dond_simulation.py +++ b/src_code_for_reproducibility/markov_games/negotiation/dond_simulation.py @@ -1,30 +1,45 @@ +""" +File: mllm/markov_games/negotiation/dond_simulation.py +Summary: Simulates Deal-or-No-Deal negotiation games and logs rollouts. +""" + import copy from dataclasses import dataclass from typing import Any, Dict, List, Tuple from numpy.random import default_rng +from mllm.markov_games.negotiation.nego_simulation import ( + NegotiationObs, + NegotiationSimulation, + NegotiationState, + Split, +) from mllm.markov_games.rollout_tree import SimulationStepLog -from mllm.markov_games.negotiation.nego_simulation import Split, NegotiationState, NegotiationObs, NegotiationSimulation from mllm.utils.get_coagent_id import get_coagent_id - AgentId = str @dataclass class DealNoDealState(NegotiationState): + """NegotiationState with per-agent value tables and item taxonomy.""" + item_types: List[str] values: Dict[AgentId, Dict[str, int]] + @dataclass class DealNoDealObs(NegotiationObs): + """Observation that reveals own values and (lagged) opponent values.""" + my_values: Dict[str, int] item_types: List[str] previous_values_coagent: Dict[str, int] | None def random_partition_integer(rng, total: int, parts: int) -> List[int]: + """Sample non-negative integers summing to ``total`` across ``parts`` buckets.""" if parts <= 0: return [] if total <= 0: @@ -37,7 +52,9 @@ def random_partition_integer(rng, total: int, parts: int) -> List[int]: prev = c return vals + class DealNoDealSimulation(NegotiationSimulation): + """NegotiationSimulation variant implementing the Rubinstein-style Deal-or-No-Deal.""" def __init__( self, @@ -75,7 +92,9 @@ class DealNoDealSimulation(NegotiationSimulation): if ok1 and ok2: return {self.agent_ids[0]: a, self.agent_ids[1]: b} - def _is_valid_allocation(self, allocation: Dict[str, int], stock: Dict[str, int]) -> bool: + def _is_valid_allocation( + self, allocation: Dict[str, int], stock: Dict[str, int] + ) -> bool: for t in self.item_types: v = allocation.get(t) if v is None: @@ -85,16 +104,18 @@ class DealNoDealSimulation(NegotiationSimulation): if v < 0 or v > int(stock.get(t, 0)): return False return True - + def set_new_round_of_variant(self): # Keep same values, resample stock self.state.quantities = self._sample_stock() - def get_info_of_variant(self, state: NegotiationState, actions: Dict[AgentId, Any]) -> Dict[str, Any]: + def get_info_of_variant( + self, state: NegotiationState, actions: Dict[AgentId, Any] + ) -> Dict[str, Any]: return { "quantities": copy.deepcopy(state.quantities), "values": copy.deepcopy(state.values), - 'splits': copy.deepcopy(state.splits), + "splits": copy.deepcopy(state.splits), } def get_rewards(self, splits: Dict[AgentId, Split]) -> Dict[AgentId, float]: @@ -105,11 +126,15 @@ class DealNoDealSimulation(NegotiationSimulation): split_b = splits[self.agent_ids[1]].items_given_to_self rewards = {self.agent_ids[0]: 0, self.agent_ids[1]: 0} for t in self.item_types: - # If not complementary, return 0! + # If not complementary, return 0! if not split_a[t] + split_b[t] == self.state.quantities[t]: return {self.agent_ids[0]: 0, self.agent_ids[1]: 0} - rewards[self.agent_ids[0]] += split_a[t] * self.state.values[self.agent_ids[0]][t] - rewards[self.agent_ids[1]] += split_b[t] * self.state.values[self.agent_ids[1]][t] + rewards[self.agent_ids[0]] += ( + split_a[t] * self.state.values[self.agent_ids[0]][t] + ) + rewards[self.agent_ids[1]] += ( + split_b[t] * self.state.values[self.agent_ids[1]][t] + ) return rewards def get_obs(self): @@ -149,5 +174,3 @@ class DealNoDealSimulation(NegotiationSimulation): item_types=list(self.item_types), ) return self.get_obs() - - diff --git a/src_code_for_reproducibility/markov_games/negotiation/nego_simulation.py b/src_code_for_reproducibility/markov_games/negotiation/nego_simulation.py index 6a4d18532ab0472cd8f83414d52cf6df589fe126..2172aa345aba163ed8d84b477f54c3fbde5ce249 100644 --- a/src_code_for_reproducibility/markov_games/negotiation/nego_simulation.py +++ b/src_code_for_reproducibility/markov_games/negotiation/nego_simulation.py @@ -1,7 +1,8 @@ """ -Negotiation simulation environment -other agent is set at the start of every round. Even though current agent changes over message turns in a round. +File: mllm/markov_games/negotiation/nego_simulation.py +Summary: Simulation harness for general negotiation environments. """ + import copy from abc import abstractmethod from dataclasses import dataclass @@ -18,16 +19,22 @@ AgentId = str @dataclass class Split: + """Structured proposal describing how many units of each item an agent keeps.""" + items_given_to_self: Dict[str, int] @dataclass class Message: + """Single chat utterance exchanged during the negotiation phase.""" + message: str @dataclass # gets extended by variants class NegotiationState: + """Full simulator state snapshot shared by all negotiation variants.""" + round_nb: int last_message: str current_agent: AgentId @@ -44,6 +51,8 @@ class NegotiationState: @dataclass # gets extended by variants class NegotiationObs: + """Observation presented to agents each turn (base fields; variants extend).""" + round_nb: int last_message: str quota_messages_per_agent_per_round: int @@ -134,12 +143,14 @@ class NegotiationSimulation(Simulation): @abstractmethod def set_new_round_of_variant(self): + """Variant hook: sample new private values / stock before each round.""" pass @abstractmethod def get_info_of_variant( self, state: NegotiationState, actions: Dict[AgentId, Any] ) -> Dict[str, Any]: + """Variant hook: populate SimulationStepLog.info with custom diagnostics.""" pass def step(self, actions: Any) -> Tuple[bool, SimulationStepLog]: @@ -190,7 +201,7 @@ class NegotiationSimulation(Simulation): is_last_timestep_in_round = True done = self.state.round_nb >= self.nb_of_rounds - # Message phase + # Message phase: roll the conversation forward a single turn. elif isinstance(action, Message): self.state.last_message = action.message self.state.nb_messages_sent[current_agent] += 1 diff --git a/src_code_for_reproducibility/markov_games/negotiation/tas_agent.py b/src_code_for_reproducibility/markov_games/negotiation/tas_agent.py index 002160873969ab7292f0f62a091e12ec376022c6..9e0ba17b1fcf2be321966898af7a024c31af9e08 100644 --- a/src_code_for_reproducibility/markov_games/negotiation/tas_agent.py +++ b/src_code_for_reproducibility/markov_games/negotiation/tas_agent.py @@ -1,9 +1,16 @@ +""" +File: mllm/markov_games/negotiation/tas_agent.py +Summary: Agent implementation for Take-and-Split negotiations. +""" + from mllm.markov_games.negotiation.nego_agent import NegotiationAgent from mllm.markov_games.negotiation.nego_simulation import Split from mllm.markov_games.negotiation.tas_simulation import TrustAndSplitObs class TrustAndSplitAgent(NegotiationAgent): + """Prompt/template wrapper for the classic multi-item Take-and-Split benchmark.""" + def __init__(self, num_message_chars, *args, **kwargs): self.num_message_chars = num_message_chars super().__init__(*args, **kwargs) @@ -58,12 +65,14 @@ class TrustAndSplitAgent(NegotiationAgent): self.send_message_prompt = f"Send your message now in ... (<={self.num_message_chars} chars)." def get_message_regex(self, observation: TrustAndSplitObs) -> str: + """Constrain chat to bounded XML tags for stable parsing.""" return rf"[\s\S]{{0,{self.num_message_chars}}}" # def get_message_regex(self, observation: TrustAndSplitObs) -> str: # return rf"(?s).{{0,{self.num_message_chars}}}" def get_split_regex(self, observation: TrustAndSplitObs) -> str: + """Allow natural-language item names while still returning machine-parsable XML.""" items = list(observation.quantities.keys()) # Accept both singular and plural forms item_pattern = "|".join( @@ -75,6 +84,7 @@ class TrustAndSplitAgent(NegotiationAgent): def get_split_action( self, policy_output: str, observation: TrustAndSplitObs ) -> Split: + """Convert human-readable allocation text back into canonical item IDs.""" items = list(observation.quantities.keys()) import re as _re diff --git a/src_code_for_reproducibility/markov_games/negotiation/tas_rps_agent.py b/src_code_for_reproducibility/markov_games/negotiation/tas_rps_agent.py index e711c2a65d336e4d9b991c68662069e96b4dfee8..5bd9dbe3c8fb43f59cc2b73f1f0227d4f32eacf8 100644 --- a/src_code_for_reproducibility/markov_games/negotiation/tas_rps_agent.py +++ b/src_code_for_reproducibility/markov_games/negotiation/tas_rps_agent.py @@ -1,3 +1,8 @@ +""" +File: mllm/markov_games/negotiation/tas_rps_agent.py +Summary: Agent logic for TAS Rock-Paper-Scissors blended game. +""" + import copy from collections.abc import Callable from dataclasses import dataclass @@ -15,6 +20,8 @@ from mllm.markov_games.rollout_tree import AgentActLog, ChatTurn class TrustAndSplitRPSAgent(NegotiationAgent): + """NegotiationAgent that reasons about hidden hands before submitting TAS splits.""" + def __init__( self, num_message_chars: int, @@ -88,6 +95,7 @@ class TrustAndSplitRPSAgent(NegotiationAgent): self.send_message_prompt = f"Send your message now in ... (<={self.num_message_chars} chars)." def get_message_regex(self, observation: TrustAndSplitRPSObs) -> str: + """Switch between ... and <> formats on demand.""" if self.message_start_end_format: return ( rf"<>[\s\S]{{0,{self.num_message_chars}}}<>" @@ -96,6 +104,7 @@ class TrustAndSplitRPSAgent(NegotiationAgent): return rf"[\s\S]{{0,{self.num_message_chars}}}" def get_split_regex(self, observation: TrustAndSplitRPSObs) -> str: + """Force single-number proposals inside whichever tag style the config selected.""" if self.proposal_start_end_format: return r"<> ?(10|[0-9]) ?<>" else: @@ -104,6 +113,7 @@ class TrustAndSplitRPSAgent(NegotiationAgent): def get_split_action( self, policy_output: str, observation: TrustAndSplitRPSObs ) -> Split: + """Parse the proposal tag (or raw integer fallback) into a Split.""" import re as _re if self.proposal_start_end_format: diff --git a/src_code_for_reproducibility/markov_games/rollout_tree.py b/src_code_for_reproducibility/markov_games/rollout_tree.py index 5e90fa6d7a0e0b644ccaf3533a1513568d21868f..c9feb0e92f3bcf19255d80c6ff2dcd9a045c6c21 100644 --- a/src_code_for_reproducibility/markov_games/rollout_tree.py +++ b/src_code_for_reproducibility/markov_games/rollout_tree.py @@ -1,5 +1,6 @@ """ -TODO: add parent to nodes so that some verification can be done. For instance, to ensure that node reward keys match the parent node. +File: mllm/markov_games/rollout_tree.py +Summary: Defines rollout tree data structures and serialization helpers. """ from __future__ import annotations @@ -18,11 +19,15 @@ AgentId = str class SimulationStepLog(BaseModel): + """Minimal snapshot of environment-side rewards and auxiliary info.""" + rewards: dict[AgentId, float] info: Any = None class AgentActLog(BaseModel): + """LLM-side provenance for an action (chat turns + metadata).""" + chat_turns: list[ChatTurn] | None info: Any = None @@ -55,6 +60,8 @@ class StepLog(BaseModel): class RolloutTreeNode(BaseModel): + """Single timestep of the main trajectory (or a branch) plus linkage.""" + step_log: StepLog time_step: int child: RolloutTreeNode | RolloutTreeBranchNode | None = None @@ -70,6 +77,8 @@ class RolloutTreeBranchNode(BaseModel): class RolloutTreeRootNode(BaseModel): + """Entry point for serialized rollouts (main path plus optional branches).""" + id: int crn_id: int # ID of the rng used to generate this rollout tree child: RolloutTreeNode | RolloutTreeBranchNode | None = None diff --git a/src_code_for_reproducibility/markov_games/run_markov_games.py b/src_code_for_reproducibility/markov_games/run_markov_games.py index 08b84024668ed4375453d1fd515a78eb9bb23414..86b7f2ef0ac2e19afdd61a8a32342aaf8120e7de 100644 --- a/src_code_for_reproducibility/markov_games/run_markov_games.py +++ b/src_code_for_reproducibility/markov_games/run_markov_games.py @@ -1,3 +1,8 @@ +""" +File: mllm/markov_games/run_markov_games.py +Summary: CLI entry point for running configured Markov-game experiments. +""" + import asyncio from collections.abc import Callable from dataclasses import dataclass @@ -14,6 +19,12 @@ async def run_markov_games( output_folder: str, markov_games: list[MarkovGame], ) -> list[RolloutTreeRootNode]: + """ + Kick off multiple Markov game rollouts concurrently and return their trees. + + Parameters mirror the Hydra configs (runner callable + kwargs) so callers can + choose ``LinearRunner``, ``AlternativeActionsRunner`` or future variants. + """ tasks = [] for mg in markov_games: tasks.append( diff --git a/src_code_for_reproducibility/markov_games/simulation.py b/src_code_for_reproducibility/markov_games/simulation.py index b0b804e2aa4c288b3d98cc8106cfd727f1cc1e1a..dd0a2e61924f9a79aee3229ed8d7aa20827ae859 100644 --- a/src_code_for_reproducibility/markov_games/simulation.py +++ b/src_code_for_reproducibility/markov_games/simulation.py @@ -1,8 +1,6 @@ """ -A Simulation is the environment of a Markov Game. -The Simulation is not responsible for properly checking / formatting the responses of LLM's. -This is the job of the `Agent` class. -Simulations expect clean actions, and are defined similarly to `gymnasium` environments, except that they are adapted for the Multi-agent setting. +File: mllm/markov_games/simulation.py +Summary: Core simulation loop utilities and step logging for Markov games. """ from abc import ABC, abstractmethod @@ -22,59 +20,68 @@ class Simulation(ABC): @abstractmethod def step(self, actions: Any) -> Tuple[bool, SimulationStepLog]: """ - Returns terminated, info + Advance the environment by one logical tick using ``actions``. + + Returns + ------- + terminated: bool + Whether the episode has finished. + SimulationStepLog + Reward/info bundle describing this transition. """ raise NotImplementedError def get_obs(self): - """Returns all agent observations in dict - - Returns: - observations - """ + """Return a dict mapping agent_id -> observation for *all* agents.""" raise NotImplementedError def get_obs_agent(self, agent_id): - """Returns observation for agent_id""" + """Return the observation for a single agent.""" raise NotImplementedError def get_obs_size(self): - """Returns the shape of the observation""" + """Describe the observation tensor shape (useful for critic heads).""" raise NotImplementedError def get_state(self): + """Return the privileged simulator state if available.""" raise NotImplementedError def get_state_size(self): - """Returns the shape of the state""" + """Describe the state tensor shape.""" raise NotImplementedError def get_avail_actions(self): + """Return the global action mask/tensor if the space is discrete.""" raise NotImplementedError def get_avail_agent_actions(self, agent_id): - """Returns the available actions for agent_id""" + """Return the available action mask for a given agent.""" raise NotImplementedError def get_total_actions(self): - """Returns the total number of actions an agent could ever take""" - # TODO: This is only suitable for a discrete 1 dimensional action space for each agent + """Returns the total number of actions an agent could ever take. + + Implementations currently assume a discrete, one-dimensional action space per agent. + """ raise NotImplementedError def get_safe_copy(self): """ - Return copy of the agent object that is decorrelated from the original object. + Return copy of the simulator that shares no mutable state with the original. """ raise NotImplementedError def reset(self): - """Returns initial observations and states""" + """Reset to the initial state and return the starting observations.""" raise NotImplementedError def render(self): + """Optional human-facing visualization.""" raise NotImplementedError def close(self): + """Release any owned resources (files, processes, etc.).""" raise NotImplementedError # def seed(self): diff --git a/src_code_for_reproducibility/markov_games/statistics_runner.py b/src_code_for_reproducibility/markov_games/statistics_runner.py index eb30800f4aa4e5f919292023df77fa29e43d9c24..e58131fc505806a758936978a46e4f8faefacad3 100644 --- a/src_code_for_reproducibility/markov_games/statistics_runner.py +++ b/src_code_for_reproducibility/markov_games/statistics_runner.py @@ -1,3 +1,8 @@ +""" +File: mllm/markov_games/statistics_runner.py +Summary: Executes multiple rollouts to compute experiment statistics. +""" + from __future__ import annotations import gc @@ -36,17 +41,20 @@ def _iterate_main_nodes(root: RolloutTreeRootNode) -> Iterator[RolloutTreeNode]: def iterate_main_simulation_logs( root: RolloutTreeRootNode, ) -> Iterator[SimulationStepLog]: + """Yield ``SimulationStepLog`` objects along the main (non-branch) path.""" for node in _iterate_main_nodes(root): yield node.step_log.simulation_step_log def stream_rollout_files(iteration_folder: Path) -> Iterator[Path]: + """Iterate over every ``*.rt.pkl`` file under an iteration directory.""" for p in iteration_folder.rglob("*.rt.pkl"): if p.is_file(): yield p def load_root(path: Path) -> RolloutTreeRootNode: + """Load and validate a rollout tree from disk.""" with open(path, "rb") as f: data = pickle.load(f) return RolloutTreeRootNode.model_validate(data) @@ -54,6 +62,8 @@ def load_root(path: Path) -> RolloutTreeRootNode: @dataclass class StatRecord: + """Convenience container for serialized stat rows.""" + mgid: int crn_id: Optional[int] iteration: str diff --git a/src_code_for_reproducibility/models/__init__.py b/src_code_for_reproducibility/models/__init__.py index e69de29bb2d1d6434b8b29ae775ad8c2e48c5391..46d40ee71acb4c10a596d4107d18fd3e890df610 100644 --- a/src_code_for_reproducibility/models/__init__.py +++ b/src_code_for_reproducibility/models/__init__.py @@ -0,0 +1,4 @@ +""" +File: mllm/models/__init__.py +Summary: Exports model-layer utilities from the models package. +""" diff --git a/src_code_for_reproducibility/models/__pycache__/human_policy.cpython-312.pyc b/src_code_for_reproducibility/models/__pycache__/human_policy.cpython-312.pyc index e3f4a2685a9cd79a8828cb18f6a3d32f03e63769..e8250e46883c9a18d29e02c6917ea0a528366fed 100644 Binary files a/src_code_for_reproducibility/models/__pycache__/human_policy.cpython-312.pyc and b/src_code_for_reproducibility/models/__pycache__/human_policy.cpython-312.pyc differ diff --git a/src_code_for_reproducibility/models/__pycache__/inference_backend.cpython-312.pyc b/src_code_for_reproducibility/models/__pycache__/inference_backend.cpython-312.pyc index c77324d067890d56a074a4e427cb5e9b6c59df01..405731ad14e4fa8cc427a89e8f23062f05f28482 100644 Binary files a/src_code_for_reproducibility/models/__pycache__/inference_backend.cpython-312.pyc and b/src_code_for_reproducibility/models/__pycache__/inference_backend.cpython-312.pyc differ diff --git a/src_code_for_reproducibility/models/__pycache__/inference_backend_dummy.cpython-312.pyc b/src_code_for_reproducibility/models/__pycache__/inference_backend_dummy.cpython-312.pyc index 6a89b2fde9bcdd46de8e225fa9ee9e67ae9f5b9c..9b0fb935c37b2dce88065a0da5188c58fcb2b762 100644 Binary files a/src_code_for_reproducibility/models/__pycache__/inference_backend_dummy.cpython-312.pyc and b/src_code_for_reproducibility/models/__pycache__/inference_backend_dummy.cpython-312.pyc differ diff --git a/src_code_for_reproducibility/models/__pycache__/inference_backend_vllm.cpython-312.pyc b/src_code_for_reproducibility/models/__pycache__/inference_backend_vllm.cpython-312.pyc index da0a8fd5d1dceb7ea6f095412c40fd34b8534e9b..555da444de4de381e84ad3e9999f1b84aab9dd37 100644 Binary files a/src_code_for_reproducibility/models/__pycache__/inference_backend_vllm.cpython-312.pyc and b/src_code_for_reproducibility/models/__pycache__/inference_backend_vllm.cpython-312.pyc differ diff --git a/src_code_for_reproducibility/models/__pycache__/large_language_model_api.cpython-312.pyc b/src_code_for_reproducibility/models/__pycache__/large_language_model_api.cpython-312.pyc index 7de6daa0ac310a269a94692e05457e6c5a8ab925..250bd3412ee0a5198f76e824ef265ae2d0bb5732 100644 Binary files a/src_code_for_reproducibility/models/__pycache__/large_language_model_api.cpython-312.pyc and b/src_code_for_reproducibility/models/__pycache__/large_language_model_api.cpython-312.pyc differ diff --git a/src_code_for_reproducibility/models/__pycache__/large_language_model_local.cpython-312.pyc b/src_code_for_reproducibility/models/__pycache__/large_language_model_local.cpython-312.pyc index e1f8ad83ac621376945ef4f69083606ae9d8bca4..0a2d8e4d055374deea50f4647881338c04779a13 100644 Binary files a/src_code_for_reproducibility/models/__pycache__/large_language_model_local.cpython-312.pyc and b/src_code_for_reproducibility/models/__pycache__/large_language_model_local.cpython-312.pyc differ diff --git a/src_code_for_reproducibility/models/__pycache__/scalar_critic.cpython-312.pyc b/src_code_for_reproducibility/models/__pycache__/scalar_critic.cpython-312.pyc index 650e0c11acbe423288c4a4a4cd005fe2be810eea..8a8922c3e90b6faf918756a152fe55910127d33b 100644 Binary files a/src_code_for_reproducibility/models/__pycache__/scalar_critic.cpython-312.pyc and b/src_code_for_reproducibility/models/__pycache__/scalar_critic.cpython-312.pyc differ diff --git a/src_code_for_reproducibility/models/adapter_training_wrapper.py b/src_code_for_reproducibility/models/adapter_training_wrapper.py index ccb41c9fb2c314db09a3e8a370ffa50c3648a198..f99c768935cca8203f7a9482d25aeade0dee0d59 100644 --- a/src_code_for_reproducibility/models/adapter_training_wrapper.py +++ b/src_code_for_reproducibility/models/adapter_training_wrapper.py @@ -1,11 +1,14 @@ -import torch -import torch.nn as nn +""" +File: mllm/models/adapter_training_wrapper.py +Summary: Wraps a shared LLM with adapter-specific PEFT handling for training. +""" + import logging from typing import Union -from peft import ( - LoraConfig, - get_peft_model, -) + +import torch +import torch.nn as nn +from peft import LoraConfig, get_peft_model logger = logging.getLogger(__name__) @@ -18,13 +21,14 @@ class AdapterWrapper(nn.Module): • exposes only the parameters that should be trained for that adapter (plus whatever extra modules you name). """ + def __init__( self, shared_llm: nn.Module, adapter_id: str, lora_config: dict, path: Union[str, None] = None, - ): + ): super().__init__() self.shared_llm = shared_llm self.adapter_id = adapter_id @@ -47,7 +51,9 @@ class AdapterWrapper(nn.Module): adapter_name=adapter_id, ) loaded_from = path - except Exception as exc: # noqa: BLE001 - want to log any load failure context + except ( + Exception + ) as exc: # noqa: BLE001 - want to log any load failure context logger.warning( f"Adapter '{adapter_id}': failed to load from '{path}': {exc}" ) diff --git a/src_code_for_reproducibility/models/human_policy.py b/src_code_for_reproducibility/models/human_policy.py index 00076a9e592cb29f8e6f33d852fc51be3e47ed13..699c2d817c17abfdd0e144e99e14fb2b3ba06872 100644 --- a/src_code_for_reproducibility/models/human_policy.py +++ b/src_code_for_reproducibility/models/human_policy.py @@ -1,3 +1,8 @@ +""" +File: mllm/models/human_policy.py +Summary: Implements an interactive human-in-the-loop policy for experiments. +""" + import asyncio import os import re diff --git a/src_code_for_reproducibility/models/inference_backend.py b/src_code_for_reproducibility/models/inference_backend.py index 2fd31f0f51e00dbcbf1db954e8eab42550e05d4e..c204482170d5a4418870805b620295cab294fab6 100644 --- a/src_code_for_reproducibility/models/inference_backend.py +++ b/src_code_for_reproducibility/models/inference_backend.py @@ -1,3 +1,8 @@ +""" +File: mllm/models/inference_backend.py +Summary: Declares the inference backend interface and shared dataclasses. +""" + from abc import ABC, abstractmethod from dataclasses import dataclass from typing import Any, Optional diff --git a/src_code_for_reproducibility/models/inference_backend_dummy.py b/src_code_for_reproducibility/models/inference_backend_dummy.py index ac7d4bd5566220377913e6af012545df492acad9..22dd123f5fbcf9a976282b0657097edc680c6ac3 100644 --- a/src_code_for_reproducibility/models/inference_backend_dummy.py +++ b/src_code_for_reproducibility/models/inference_backend_dummy.py @@ -1,3 +1,8 @@ +""" +File: mllm/models/inference_backend_dummy.py +Summary: Stub inference backend that returns synthetic completions for tests. +""" + import asyncio from typing import Optional diff --git a/src_code_for_reproducibility/models/inference_backend_vllm.py b/src_code_for_reproducibility/models/inference_backend_vllm.py index fed694711057ba831e57a75b7b53d019026d835a..a4a7fc73287cb676ce56beea5de77cf03fc24555 100644 --- a/src_code_for_reproducibility/models/inference_backend_vllm.py +++ b/src_code_for_reproducibility/models/inference_backend_vllm.py @@ -1,3 +1,8 @@ +""" +File: mllm/models/inference_backend_vllm.py +Summary: Connects to in-process vLLM instances for batched generation. +""" + import asyncio import re from typing import Optional @@ -23,19 +28,12 @@ class VLLMAsyncBackend(LLMInferenceBackend): sampling_params: dict = {}, ): self.model_name = model_name - # self.adapter_paths = adapter_paths or {} - # self.current_adapter = None - # self.vllm_adapter_ids = { - # adapter_id: generate_short_id() for adapter_id in adapter_paths.keys() - # } self.vllm_adapter_ids = {} ea = dict(model=model_name, **engine_init_kwargs) - # ea["enable_lora"] = True - # ea["max_loras"] = len(self.vllm_adapter_ids) - # ea["enable_sleep_mode"] = True self.engine = AsyncLLMEngine.from_engine_args(AsyncEngineArgs(**ea)) self.sampling_params = sampling_params + self.tokenizer = tokenizer def prepare_adapter( self, @@ -43,7 +41,6 @@ class VLLMAsyncBackend(LLMInferenceBackend): adapter_path: Optional[str], weights_got_updated: bool, ) -> None: - # self.current_adapter = adapter_id if weights_got_updated: self.vllm_adapter_ids[adapter_id] = generate_short_id() self.current_lora_request = LoRARequest( @@ -96,9 +93,6 @@ class VLLMAsyncBackend(LLMInferenceBackend): ] log_probs = torch.tensor(log_probs) out_token_ids = torch.tensor(out_token_ids, dtype=torch.long) - # for out_token_id, logprob_dict in zip(out_token_ids, res.outputs[0].logprobs): - # if logprob_dict[out_token_id].logprob < -1: - # print(f"High negative logprob {logprob_dict[out_token_id].logprob} for {logprob_dict}") content = raw_text reasoning_content = None diff --git a/src_code_for_reproducibility/models/large_language_model_api.py b/src_code_for_reproducibility/models/large_language_model_api.py index 1afedf4e19f6b00c40c0127b18843eab03a89648..d687c85d5386ec909ab140c7166886725d64f97d 100644 --- a/src_code_for_reproducibility/models/large_language_model_api.py +++ b/src_code_for_reproducibility/models/large_language_model_api.py @@ -1,3 +1,8 @@ +""" +File: mllm/models/large_language_model_api.py +Summary: Implements API-based large-language-model inference adapters. +""" + from __future__ import annotations import asyncio @@ -13,7 +18,7 @@ from openai import AsyncOpenAI, OpenAIError from mllm.markov_games.rollout_tree import ChatTurn from mllm.models.inference_backend import LLMInferenceOutput -# TODO: Get this automatically from OpenAI +# Static list copied from the public OpenAI docs until a discovery endpoint is exposed. reasoning_models = [ "gpt-5-nano", "gpt-5-mini", @@ -119,9 +124,7 @@ class LargeLanguageModelOpenAI: agent_id: str, regex: Optional[str] = None, ) -> LLMInferenceOutput: - # Remove any non-role/content keys from the prompt else openai will error - - # TODO: + # Remove any non-role/content keys from the prompt else openai will error. prompt = [{"role": p.role, "content": p.content} for p in state] # if self.sleep_between_requests: diff --git a/src_code_for_reproducibility/models/large_language_model_local.py b/src_code_for_reproducibility/models/large_language_model_local.py index 7eac1c32c0233cf04106fa12be333ebf74319c2a..4475b51d8fa895895df6d6dcaff4ed430cda74ca 100644 --- a/src_code_for_reproducibility/models/large_language_model_local.py +++ b/src_code_for_reproducibility/models/large_language_model_local.py @@ -1,5 +1,6 @@ """ -TODO: Figure out how to tweak SGlang not to go OOM when batch size is 32. See https://github.com/sgl-project/sglang/issues/6309. +File: mllm/models/large_language_model_local.py +Summary: Provides a local large language model wrapper over inference backends. """ import logging @@ -16,23 +17,14 @@ import httpx import requests import torch import torch.nn as nn - -# from sglang.utils import ( -# launch_server_cmd, -# print_highlight, -# terminate_process, -# wait_for_server, -# ) from torch.optim import SGD, Adam, AdamW, RMSprop from transformers import AutoModelForCausalLM, AutoTokenizer -from trl import AutoModelForCausalLMWithValueHead from mllm.chat_utils.apply_template import chat_turns_to_token_ids from mllm.markov_games.rollout_tree import ChatTurn from mllm.models.adapter_training_wrapper import AdapterWrapper from mllm.models.inference_backend import LLMInferenceOutput from mllm.models.inference_backend_dummy import DummyInferenceBackend -from mllm.models.inference_backend_sglang import SGLangOfflineBackend from mllm.models.inference_backend_vllm import VLLMAsyncBackend logger = logging.getLogger(__name__) @@ -44,7 +36,7 @@ PolicyID = str class LeanLocalLLM: """ - TOWRITE + Wrapper that manages local HuggingFace models, adapters, and inference backends. """ def __init__( @@ -55,7 +47,7 @@ class LeanLocalLLM: hf_kwargs: dict = {}, adapter_configs: dict = {}, output_directory: str = "./models/", - inference_backend: Literal["vllm", "sglang", "dummy"] = "vllm", + inference_backend: Literal["vllm", "dummy"] = "vllm", inference_backend_sampling_params: dict = {}, inference_backend_init_kwargs: dict = {}, initial_adapter_paths: dict[str, str] | None = None, @@ -180,15 +172,7 @@ class LeanLocalLLM: # Init inference inference_backend # --------------------------------------------------------- - if inference_backend == "sglang": - self.inference_backend = SGLangOfflineBackend( - model_name=self.model_name, - save_path=self.save_path, - adapter_paths=self.adapter_paths, - tokenizer=self.tokenizer, - kwargs=inference_backend_init_kwargs, - ) - elif inference_backend == "vllm": + if inference_backend == "vllm": self.inference_backend = VLLMAsyncBackend( model_name=self.model_name, # adapter_paths=self.adapter_paths, @@ -206,7 +190,7 @@ class LeanLocalLLM: def get_inference_policies(self) -> dict[PolicyID, Callable]: """ - TOWRITE + Build async policy callables keyed by adapter id for inference-only usage. """ policies = {} for adapter_id in self.adapter_ids: @@ -242,8 +226,8 @@ class LeanLocalLLM: """ Returns wrappers over the adapters which allows them be interfaced like regular PyTorch models. - # TODO: create the adapter wrappers here - See adapter_wrapper.py + AdapterWrapper lives in adapter_wrapper.py; the huggingface modules already wrap + parameters here, so we surface them directly until an extra shim is required. """ trainable_objects = {an: self.hf_adapters[an] for an in self.adapter_ids} return trainable_objects @@ -297,13 +281,11 @@ class LeanLocalLLM: tokenizer=self.tokenizer, enable_thinking=self.enable_thinking, ) - # print(f"context is {self.tokenizer.decode(context_token_ids)}") policy_output = await self.inference_backend.generate( input_token_ids=context_token_ids.tolist(), extract_thinking=(self.max_thinking_characters > 0), regex=current_regex, ) - # print(f"generated: {self.tokenizer.decode(policy_output.out_token_ids)}") if ( pattern is None or (pattern.fullmatch(policy_output.content)) @@ -347,11 +329,6 @@ class LeanLocalLLM: for adapter_id in self.past_agent_adapter_ids: self.weights_got_updated[adapter_id] = True - # import random - # self.save_path = self.save_path + str(random.randint(1,500)) - # print(f"Save path: {self.save_path}") - # self.adapter_paths = {adapter_id:os.path.join(self.save_path, adapter_id) for adapter_id in self.adapter_ids} - adapter_id = self.adapter_ids[0] self.hf_adapters[adapter_id].save_pretrained(self.save_path) diff --git a/src_code_for_reproducibility/models/scalar_critic.py b/src_code_for_reproducibility/models/scalar_critic.py index b0cabf6acfb6757db2871778a397bdbe38b813dd..0b704dcc78fdfbed1c68b1ac469e9c7b51758211 100644 --- a/src_code_for_reproducibility/models/scalar_critic.py +++ b/src_code_for_reproducibility/models/scalar_critic.py @@ -1,6 +1,13 @@ -import torch, torch.nn as nn, torch.optim as optim -from transformers import AutoModelForCausalLM, AutoTokenizer +""" +File: mllm/models/scalar_critic.py +Summary: Defines a scalar critic network and helper utilities. +""" + +import torch +import torch.nn as nn +import torch.optim as optim from peft import LoraConfig, get_peft_model +from transformers import AutoModelForCausalLM, AutoTokenizer from mllm.models.adapter_training_wrapper import AdapterWrapper @@ -11,18 +18,16 @@ class ScalarCritic(nn.Module): V_φ(s) = wᵀ h_last + b Only LoRA adapters (inside critic_adapter) and the value head are trainable. """ + def __init__(self, critic_adapter: AdapterWrapper): super().__init__() self.critic_adapter = critic_adapter hidden_size = self.critic_adapter.shared_llm.config.hidden_size self.value_head = nn.Linear(hidden_size, 1).to( - dtype=critic_adapter.dtype, - device=critic_adapter.device) + dtype=critic_adapter.dtype, device=critic_adapter.device + ) - def forward(self, - input_ids, - attention_mask=None, - **kwargs): + def forward(self, input_ids, attention_mask=None, **kwargs): # AdapterWrapper activates its own adapter internally outputs = self.critic_adapter( input_ids=input_ids, @@ -30,7 +35,7 @@ class ScalarCritic(nn.Module): output_hidden_states=True, **kwargs, ) - h_last = outputs.hidden_states[-1] # (B, S, H) + h_last = outputs.hidden_states[-1] # (B, S, H) values = self.value_head(h_last).squeeze(-1) # (B, S) return values diff --git a/src_code_for_reproducibility/training/__init__.py b/src_code_for_reproducibility/training/__init__.py index e69de29bb2d1d6434b8b29ae775ad8c2e48c5391..188fde562de5d8f658ef12708df9967f45cb2a7a 100644 --- a/src_code_for_reproducibility/training/__init__.py +++ b/src_code_for_reproducibility/training/__init__.py @@ -0,0 +1,4 @@ +""" +File: mllm/training/__init__.py +Summary: Exposes training submodules through the package namespace. +""" diff --git a/src_code_for_reproducibility/training/__pycache__/__init__.cpython-312.pyc b/src_code_for_reproducibility/training/__pycache__/__init__.cpython-312.pyc index 6d68eb043e149c6005bbcdfe2fbfb1e1d1d61819..eb3d8d4bee48323b7b756ccf91dc2a191e067e3b 100644 Binary files a/src_code_for_reproducibility/training/__pycache__/__init__.cpython-312.pyc and b/src_code_for_reproducibility/training/__pycache__/__init__.cpython-312.pyc differ diff --git a/src_code_for_reproducibility/training/__pycache__/annealing_methods.cpython-312.pyc b/src_code_for_reproducibility/training/__pycache__/annealing_methods.cpython-312.pyc index 73991155f0a6d13f03b43dfcd777a49610c88c0e..72478465a6de321e0dfd6a137aeb563c85a97f62 100644 Binary files a/src_code_for_reproducibility/training/__pycache__/annealing_methods.cpython-312.pyc and b/src_code_for_reproducibility/training/__pycache__/annealing_methods.cpython-312.pyc differ diff --git a/src_code_for_reproducibility/training/__pycache__/credit_methods.cpython-312.pyc b/src_code_for_reproducibility/training/__pycache__/credit_methods.cpython-312.pyc index 549ee7cccb7bf669178129df112ae091e2389518..46eae35640116412e52dc6fbbc7f7038c07dd49d 100644 Binary files a/src_code_for_reproducibility/training/__pycache__/credit_methods.cpython-312.pyc and b/src_code_for_reproducibility/training/__pycache__/credit_methods.cpython-312.pyc differ diff --git a/src_code_for_reproducibility/training/__pycache__/tally_metrics.cpython-312.pyc b/src_code_for_reproducibility/training/__pycache__/tally_metrics.cpython-312.pyc index 6c2ee27df0f569f1b270d0b7e2977afa9c954181..d0bfb113d53be2b81539815445c95d5e83d1dcc6 100644 Binary files a/src_code_for_reproducibility/training/__pycache__/tally_metrics.cpython-312.pyc and b/src_code_for_reproducibility/training/__pycache__/tally_metrics.cpython-312.pyc differ diff --git a/src_code_for_reproducibility/training/__pycache__/tally_rollout.cpython-312.pyc b/src_code_for_reproducibility/training/__pycache__/tally_rollout.cpython-312.pyc index c27a1b8c25697b30b7642894b4ac32e5f5b41531..224f9e383d51bd946401cd4aa01919574b577bad 100644 Binary files a/src_code_for_reproducibility/training/__pycache__/tally_rollout.cpython-312.pyc and b/src_code_for_reproducibility/training/__pycache__/tally_rollout.cpython-312.pyc differ diff --git a/src_code_for_reproducibility/training/__pycache__/tally_tokenwise.cpython-312.pyc b/src_code_for_reproducibility/training/__pycache__/tally_tokenwise.cpython-312.pyc index afe36bd9eb6fa41bb714296a853748846986ed3c..31df01b89cad4eab566667f599501f06d824190e 100644 Binary files a/src_code_for_reproducibility/training/__pycache__/tally_tokenwise.cpython-312.pyc and b/src_code_for_reproducibility/training/__pycache__/tally_tokenwise.cpython-312.pyc differ diff --git a/src_code_for_reproducibility/training/__pycache__/tokenize_chats.cpython-312.pyc b/src_code_for_reproducibility/training/__pycache__/tokenize_chats.cpython-312.pyc index bb3cd1234f6a70b27bef4626f66ec30d5ab52517..f8572c9625f8725bad9404192a5ee14e5d5f1cc6 100644 Binary files a/src_code_for_reproducibility/training/__pycache__/tokenize_chats.cpython-312.pyc and b/src_code_for_reproducibility/training/__pycache__/tokenize_chats.cpython-312.pyc differ diff --git a/src_code_for_reproducibility/training/__pycache__/trainer_ad_align.cpython-312.pyc b/src_code_for_reproducibility/training/__pycache__/trainer_ad_align.cpython-312.pyc index 7a1cc78df9edd45b4e52fc1f2685eb3a99d48d23..9ef8e960c579b57b42e743902666edb071a55883 100644 Binary files a/src_code_for_reproducibility/training/__pycache__/trainer_ad_align.cpython-312.pyc and b/src_code_for_reproducibility/training/__pycache__/trainer_ad_align.cpython-312.pyc differ diff --git a/src_code_for_reproducibility/training/__pycache__/trainer_common.cpython-312.pyc b/src_code_for_reproducibility/training/__pycache__/trainer_common.cpython-312.pyc index cc7e9a6c2510ae34731aa3d46ca42829a483bba4..12a65f20e21690751f077533e91cb7f4fc300f14 100644 Binary files a/src_code_for_reproducibility/training/__pycache__/trainer_common.cpython-312.pyc and b/src_code_for_reproducibility/training/__pycache__/trainer_common.cpython-312.pyc differ diff --git a/src_code_for_reproducibility/training/__pycache__/trainer_independent.cpython-312.pyc b/src_code_for_reproducibility/training/__pycache__/trainer_independent.cpython-312.pyc index 45d61bbfffaa2299ef7c915156ead08105ef64af..854620cf63cea8ec419bcb2a5f63788abcb01bf6 100644 Binary files a/src_code_for_reproducibility/training/__pycache__/trainer_independent.cpython-312.pyc and b/src_code_for_reproducibility/training/__pycache__/trainer_independent.cpython-312.pyc differ diff --git a/src_code_for_reproducibility/training/__pycache__/trainer_sum_rewards.cpython-312.pyc b/src_code_for_reproducibility/training/__pycache__/trainer_sum_rewards.cpython-312.pyc index ad600757f4ff3da914a94f33768dba90070148cb..14c926fe983ca27a29fdfb8afc569d9ff83d60c7 100644 Binary files a/src_code_for_reproducibility/training/__pycache__/trainer_sum_rewards.cpython-312.pyc and b/src_code_for_reproducibility/training/__pycache__/trainer_sum_rewards.cpython-312.pyc differ diff --git a/src_code_for_reproducibility/training/__pycache__/training_data_utils.cpython-312.pyc b/src_code_for_reproducibility/training/__pycache__/training_data_utils.cpython-312.pyc index 01210498226a9a10c4c6675edd9108c376cad74b..6b5a388e13e83425d7cf9aa370b5715d8b01701e 100644 Binary files a/src_code_for_reproducibility/training/__pycache__/training_data_utils.cpython-312.pyc and b/src_code_for_reproducibility/training/__pycache__/training_data_utils.cpython-312.pyc differ diff --git a/src_code_for_reproducibility/training/annealing_methods.py b/src_code_for_reproducibility/training/annealing_methods.py index 1d46d6fe04d9482ab4f2c5763f4735390d7ecdf6..591d91f7720880fc202b116b27b15b996c256dc4 100644 --- a/src_code_for_reproducibility/training/annealing_methods.py +++ b/src_code_for_reproducibility/training/annealing_methods.py @@ -1,6 +1,20 @@ +""" +File: mllm/training/annealing_methods.py +Summary: Implements annealing schedules used across training loops. +""" + import numpy as np def sigmoid_annealing(step: int, temperature: float) -> float: - return 2 / (1 + np.exp(-step / temperature)) - 1 + """ + Smoothly ramp a scalar from 0 → 1 using a temperature-controlled sigmoid. + Args: + step: Current training step or iteration. + temperature: Controls how sharp the transition is; larger values flatten the curve. + + Returns: + Float in [-1, 1] that can be rescaled for annealing schedules. + """ + return 2 / (1 + np.exp(-step / temperature)) - 1 diff --git a/src_code_for_reproducibility/training/credit_methods.py b/src_code_for_reproducibility/training/credit_methods.py index f2b639121f84439b9cbceac4e4a9b09a1a5f7371..c29032630d06536e7efe6ceaae04092a616ce53a 100644 --- a/src_code_for_reproducibility/training/credit_methods.py +++ b/src_code_for_reproducibility/training/credit_methods.py @@ -1,9 +1,16 @@ +""" +File: mllm/training/credit_methods.py +Summary: Holds credit-assignment routines for reinforcement learning updates. +""" + import torch def whiten_advantages(advantages: torch.Tensor) -> torch.Tensor: """ - Whitens the advantages. + Normalize a vector of advantages to zero mean / unit variance (global). + + Useful for variance reduction before computing gradients. """ whitened_advantages = (advantages - torch.mean(advantages)) / ( torch.std(advantages) + 1e-9 @@ -15,7 +22,9 @@ def whiten_advantages_time_step_wise( advantages: torch.Tensor, # (B, T) ) -> torch.Tensor: """ - Whitens the advantages. + Whiten advantages independently per timestep (column-wise mean/std). + + Helps when rollout lengths differ or certain positions have very different scales. """ assert advantages.dim() == 2, "Wrong dimensions." whitened_advantages_time_step_wise = ( @@ -28,7 +37,9 @@ def get_discounted_state_visitation_credits( credits: torch.Tensor, discount_factor: float # (B, T) ) -> torch.Tensor: """ - Computes discounted state visitation credits for a sequence of credits. + Apply geometric discounting to credits so earlier visits count less. + + Equivalent to per-timestep multiplication by ``gamma^t``. """ return credits * ( discount_factor ** torch.arange(credits.shape[1], device=credits.device) @@ -59,6 +70,7 @@ def get_discounted_returns( def get_rloo_credits(credits: torch.Tensor): # (B, S) + """Compute leave-one-out baselines for a batch of credits.""" assert credits.dim() == 2, "Wrong dimensions." rloo_baselines = torch.zeros_like(credits) n = credits.shape[0] @@ -76,12 +88,9 @@ def get_generalized_advantage_estimates( lambda_coef: float, ) -> torch.Tensor: """ - Computes Generalized Advantage Estimates (GAE) for a sequence of rewards and value estimates. - See https://arxiv.org/pdf/1506.02438 for details. - + Compute Generalized Advantage Estimates (GAE). - Returns: - torch.Tensor: Array of GAE values. + See https://arxiv.org/pdf/1506.02438 for derivation. """ assert rewards.dim() == value_estimates.dim() == 2, "Wrong dimensions." @@ -106,6 +115,7 @@ def get_advantage_alignment_weights( advantages: torch.Tensor, # (B, T) exclude_k_equals_t: bool, gamma: float, + discount_t: bool, ) -> torch.Tensor: """ The advantage alignment credit is calculated as @@ -127,7 +137,6 @@ def get_advantage_alignment_weights( sub = torch.eye(T, device=advantages.device) else: sub = torch.zeros((T, T), device=advantages.device) - # Identity is for \( k < t \), remove for \( k \leq t \) ad_align_weights = discounted_advantages @ ( torch.triu(torch.ones((T, T), device=advantages.device)) - sub @@ -136,6 +145,11 @@ def get_advantage_alignment_weights( torch.arange(0, T, 1, device=advantages.device) ) ad_align_weights = t_discounts * ad_align_weights + if discount_t: + time_discounted_advantages = advantages * ( + gamma * torch.ones((1, T), device=advantages.device) + ) ** (torch.arange(0, T, 1, device=advantages.device)) + ad_align_weights = ad_align_weights - advantages + time_discounted_advantages return ad_align_weights @@ -157,6 +171,7 @@ def get_advantage_alignment_credits( mean_normalize_ad_align: bool = False, whiten_adalign_advantages: bool = False, whiten_adalign_advantages_time_step_wise: bool = False, + discount_t: bool = False, ) -> torch.Tensor: """ Calculate the advantage alignment credits with vectorization, as described in https://arxiv.org/abs/2406.14662. @@ -202,7 +217,10 @@ def get_advantage_alignment_credits( if use_old_ad_align: ad_align_weights = get_advantage_alignment_weights( - advantages=a1, exclude_k_equals_t=exclude_k_equals_t, gamma=gamma + advantages=a1, + exclude_k_equals_t=exclude_k_equals_t, + gamma=gamma, + discount_t=discount_t, ) sub_tensors["ad_align_weights_prev"] = ad_align_weights if exclude_k_equals_t: @@ -212,7 +230,6 @@ def get_advantage_alignment_credits( if rloo_branch: a1_alternative = torch.cat([a1.unsqueeze(2), a1_alternative], dim=2) a1_alternative = a1_alternative.mean(dim=2) - # print(f"a1_alternative: {a1_alternative}, a1: {a1}\n") a1, baseline = get_rloo_credits(a1) if reuse_baseline: a1_alternative = a1_alternative - baseline @@ -265,11 +282,6 @@ def get_advantage_alignment_credits( if force_coop_first_step: ad_align_weights[:, 0] = 1 sub_tensors["coop_first_step_ad_align_weights"] = ad_align_weights - # # Normalize alignment terms (across same time step) - # if use_variance_regularization_in_ad_align: - # # TODO: verify - # reg_coef = torch.std(a1[:, -1]) / (torch.std(opp_shaping_terms[:, -1]) + 1e-9) - # opp_shaping_terms *= reg_coef #################################### # Compose elements together diff --git a/src_code_for_reproducibility/training/tally_metrics.py b/src_code_for_reproducibility/training/tally_metrics.py index 9d3675c78a66b65a7d538975586c5882ee11b84c..a0073ab3be86bbaf448b6a898341a4f5a3087b24 100644 --- a/src_code_for_reproducibility/training/tally_metrics.py +++ b/src_code_for_reproducibility/training/tally_metrics.py @@ -1,3 +1,8 @@ +""" +File: mllm/training/tally_metrics.py +Summary: Transforms tally files into aggregated metric summaries. +""" + import os from numbers import Number from typing import Union @@ -16,9 +21,11 @@ class Tally: self.stats = {} def reset(self): + """Reset all recorded metrics back to an empty dictionary.""" self.stats = {} def _coerce_scalar(self, value: Union[int, float]) -> Union[int, float]: + """Ensure ``value`` is a plain Python scalar (detach tensors, etc.).""" if hasattr(value, "item") and callable(getattr(value, "item")): try: value = value.item() @@ -29,6 +36,7 @@ class Tally: raise AssertionError("Metric must be a scalar number") def add_metric(self, path: str, metric: Union[int, float]): + """Accumulate a metric under ``path`` (scalar on first add, list thereafter).""" metric = float(metric) assert isinstance(path, str), "Path must be a string." assert isinstance(metric, float), "Metric must be a scalar number." @@ -43,6 +51,7 @@ class Tally: self.stats[path] = [existing, scalar] def save(self, identifier: str, folder: str): + """Persist the tally as a pickle file under ``folder``.""" os.makedirs(name=folder, exist_ok=True) try: import pickle diff --git a/src_code_for_reproducibility/training/tally_rollout.py b/src_code_for_reproducibility/training/tally_rollout.py index 18bbff345510b55f36aafc88f3f8be9ef46657cc..04bb4f36d7f1c6759c3fb0f0102f29b117ea57c1 100644 --- a/src_code_for_reproducibility/training/tally_rollout.py +++ b/src_code_for_reproducibility/training/tally_rollout.py @@ -1,3 +1,8 @@ +""" +File: mllm/training/tally_rollout.py +Summary: Serializes rollout data into tallies for downstream processing. +""" + import json import os from copy import deepcopy @@ -10,16 +15,14 @@ from transformers import AutoTokenizer class RolloutTallyItem: - def __init__(self, crn_ids: list[str], rollout_ids: list[str], agent_ids: list[str], metric_matrix: torch.Tensor): - """ - Initializes the RolloutTallyItem object. - - Args: - crn_ids (list[str]): List of CRN IDs. - rollout_ids (list[str]): List of rollout IDs. - agent_ids (list[str]): List of agent IDs. - metric_matrix (torch.Tensor): Metric matrix. - """ + def __init__( + self, + crn_ids: list[str], + rollout_ids: list[str], + agent_ids: list[str], + metric_matrix: torch.Tensor, + ): + """Lightweight data container that keeps rollout-aligned metric matrices.""" if isinstance(crn_ids, torch.Tensor): crn_ids = crn_ids.detach().cpu().numpy() if isinstance(rollout_ids, torch.Tensor): @@ -30,7 +33,9 @@ class RolloutTallyItem: self.rollout_ids = rollout_ids self.agent_ids = agent_ids metric_matrix = metric_matrix.detach().cpu() - assert 0 < metric_matrix.ndim <= 2, "Metric matrix must have less than or equal to 2 dimensions" + assert ( + 0 < metric_matrix.ndim <= 2 + ), "Metric matrix must have less than or equal to 2 dimensions" if metric_matrix.ndim == 1: metric_matrix = metric_matrix.reshape(1, -1) # Convert to float32 if tensor is in BFloat16 format (not supported by numpy) @@ -38,6 +43,7 @@ class RolloutTallyItem: metric_matrix = metric_matrix.float() self.metric_matrix = metric_matrix.numpy() + class RolloutTally: """ Tally is a utility class for collecting and storing training metrics. @@ -57,44 +63,23 @@ class RolloutTally: # Global ordered list of sample identifiers (crn_id, rollout_id) added in the order samples are processed def reset(self): - """ - Resets the base and contextualized tallies to empty dictionaries. - """ + """Reset the tally to an empty dict.""" self.metrics = {} def get_from_nested_dict(self, dictio: dict, path: str): - """ - Retrieves the value at a nested path in a dictionary. - - Args: - dictio (dict): The dictionary to search. - path (list): List of keys representing the path. - - Returns: - Any: The value at the specified path, or None if not found. - """ + """Retrieve a nested entry, creating intermediate dicts as needed.""" assert isinstance(path, list), "Path must be list." for sp in path[:-1]: dictio = dictio.setdefault(sp, {}) return dictio.get(path[-1], None) def set_at_path(self, dictio: dict, path: str, value): - """ - Sets a value at a nested path in a dictionary, creating intermediate dictionaries as needed. - - Args: - dictio (dict): The dictionary to modify. - path (list): List of keys representing the path. - value (Any): The value to set at the specified path. - """ + """Store ``value`` at ``path``; helper used by ``add_metric``.""" for sp in path[:-1]: dictio = dictio.setdefault(sp, {}) dictio[path[-1]] = value - - def add_metric( - self, path: list[str], rollout_tally_item: RolloutTallyItem - ): + def add_metric(self, path: list[str], rollout_tally_item: RolloutTallyItem): """ Adds a metric to the base tally at the specified path. @@ -111,14 +96,8 @@ class RolloutTally: else: array_list.append(rollout_tally_item) - def save(self, identifier: str, folder: str): - """ - Saves the base and contextualized tallies to disk as JSON files, and also saves contextualized tallies as CSV files for each game/rollout. - - Args: - path (str): Directory path where the metrics will be saved. - """ + """Persist the tally as a pickle (metrics only) under ``folder``.""" os.makedirs(name=folder, exist_ok=True) from datetime import datetime diff --git a/src_code_for_reproducibility/training/tally_tokenwise.py b/src_code_for_reproducibility/training/tally_tokenwise.py index 5c04f480dedc113e4004c1143ff46ff5cc34785b..b7770e0cb79d5ed4e56a3f66b6982582c72e0bb7 100644 --- a/src_code_for_reproducibility/training/tally_tokenwise.py +++ b/src_code_for_reproducibility/training/tally_tokenwise.py @@ -1,3 +1,8 @@ +""" +File: mllm/training/tally_tokenwise.py +Summary: Converts token-level tallies into per-token statistics. +""" + import json import os from typing import Any, Dict, List, Tuple, Union @@ -47,6 +52,7 @@ class ContextualizedTokenwiseTally: return self.tokenizer.convert_ids_to_tokens(tids) def _ensure_ready(self): + """Validate that action mask and range are configured prior to writes.""" assert self.action_mask is not None, "call set_action_mask(mask) first" assert self.range is not None, "call set_range((start, end)) first" @@ -71,15 +77,11 @@ class ContextualizedTokenwiseTally: # --------- Setters --------- def set_action_mask(self, action_mask: torch.Tensor): - """ - action_mask: (B, S) bool or 0/1 indicating valid steps - """ + """Register the (B, S) mask indicating which tokens correspond to actions.""" self.action_mask = action_mask def set_range(self, range: Tuple[int, int]): - """ - range: slice (start, end) into self.paths for current batch - """ + """Record which subset of ``paths`` the current mini-batch corresponds to.""" self.range = range # --------- Column builders --------- diff --git a/src_code_for_reproducibility/training/tokenize_chats.py b/src_code_for_reproducibility/training/tokenize_chats.py index 5e6017cdbb805c3002c254eebd151d7192eec4ad..94da0030ec2afe19d5e5cd8a9a9e39b595d19975 100644 --- a/src_code_for_reproducibility/training/tokenize_chats.py +++ b/src_code_for_reproducibility/training/tokenize_chats.py @@ -1,3 +1,8 @@ +""" +File: mllm/training/tokenize_chats.py +Summary: Tokenizes chat datasets and prepares tensors for training. +""" + import logging import sys @@ -11,11 +16,6 @@ logger = logging.getLogger(__name__) logger.addHandler(logging.StreamHandler(sys.stdout)) -# def get_chat_dicts(chat: list[TrainingChatTurn]) -> list[dict]: -# chat_dicts = [chat_turn.dict() for chat_turn in chat] -# return chat_dicts - - def process_training_chat( tokenizer: AutoTokenizer, chat_history: list[TrainingChatTurn], diff --git a/src_code_for_reproducibility/training/trainer_ad_align.py b/src_code_for_reproducibility/training/trainer_ad_align.py index 88f2f35788885cb1ff2eaa981eb1e74951930f1a..14e18e51480e594355b3416555011223ff0e8f36 100644 --- a/src_code_for_reproducibility/training/trainer_ad_align.py +++ b/src_code_for_reproducibility/training/trainer_ad_align.py @@ -1,3 +1,8 @@ +""" +File: mllm/training/trainer_ad_align.py +Summary: Trainer specialized for the advantage-alignment objective. +""" + import copy import logging import sys @@ -40,6 +45,8 @@ AgentId = str @dataclass class AdAlignTrainingData: + """Holds tensorized rollouts plus precomputed advantages for one agent.""" + agent_id: str main_data: TrajectoryBatch # list-of-tensors: per rollout advantages with length jT @@ -53,12 +60,14 @@ def get_alternative_chat_histories( agent_id: str, root: RolloutTreeRootNode ) -> list[list[TrainingChatTurn], list[torch.FloatTensor]]: """ - args: - agent_id: The agent we want to get the chat history for. - root: The root of the rollout tree. - returns: - alternative_chats: list[list[TrainingChatTurn]] (jT*A, jS') - alternative_rewards: list[torch.FloatTensor] (jT*A, jT') + Traverse every unilateral branch under ``root`` and collect chat/reward histories. + + Returns + ------- + alternative_chats: + Flattened list of chat turns for each branch (ordered by branch depth). + alternative_rewards: + Matching list of reward tensors aligned with the chat history. """ current_node = root.child branches = current_node.branches @@ -123,6 +132,7 @@ class TrainerAdAlign(BaseTrainer): mean_normalize_ad_align: bool = False, whiten_adalign_advantages: bool = False, whiten_adalign_advantages_time_step_wise: bool = False, + ad_align_discount_t: bool = False, *args, **kwargs, ): @@ -156,6 +166,7 @@ class TrainerAdAlign(BaseTrainer): self.whiten_adalign_advantages_time_step_wise = ( whiten_adalign_advantages_time_step_wise ) + self.ad_align_discount_t = ad_align_discount_t self.training_data: dict[AgentId, AdAlignTrainingData] = {} self.debug_path_list: list[str] = [] @@ -163,8 +174,7 @@ class TrainerAdAlign(BaseTrainer): self, agent_id: str, roots: list[RolloutTreeRootNode] ): """ - TOWRITE - Set the advantage alignment data for the trainer. + Materialize main and alternative trajectory tensors used by the advantage-alignment trainer. """ B = len(roots) # Number of rollouts @@ -227,7 +237,9 @@ class TrainerAdAlign(BaseTrainer): batch_state_ends_mask.append(state_ends_mask) batch_engine_log_probs.append(engine_log_probs) batch_rewards.append(main_rewards) - jT = main_rewards.numel() # TODO: better than this + jT = ( + main_rewards.numel() + ) # Number of timesteps inferred from reward tensor length. jT_list.append(jT) if A > 0: # We get the branching time steps for each of the `jT` time steps in the main trajectory. @@ -455,6 +467,7 @@ class TrainerAdAlign(BaseTrainer): mean_normalize_ad_align=self.mean_normalize_ad_align, whiten_adalign_advantages=self.whiten_adalign_advantages, whiten_adalign_advantages_time_step_wise=self.whiten_adalign_advantages_time_step_wise, + discount_t=self.ad_align_discount_t, ) for key, value in sub_tensors.items(): self.rollout_tally.add_metric( diff --git a/src_code_for_reproducibility/training/trainer_common.py b/src_code_for_reproducibility/training/trainer_common.py index 8da913096e80194d87d4f6f69e0a5ebc3b4c9cb6..0a0fb6f64707df0314ca4ed53611c2c592cc4377 100644 --- a/src_code_for_reproducibility/training/trainer_common.py +++ b/src_code_for_reproducibility/training/trainer_common.py @@ -1,8 +1,8 @@ """ -TODO: Add coefficients for losses (depend on total number of tokens or batch) -TODO: adapt reinforce step for torch.compile -TODO: add lr schedulers support +File: mllm/training/trainer_common.py +Summary: Shared trainer utilities, base classes, and gradient helpers. """ + import logging import os import pickle @@ -53,7 +53,10 @@ class TrainerAnnealingState: class BaseTrainer(ABC): """ - Trainer + Shared scaffolding for policy-gradient trainers (optimizer wiring, logging, etc.). + + Subclasses implement `set_agent_trajectory_data` / `share_advantage_data` + to plug in algorithm-specific behavior. """ def __init__( @@ -97,6 +100,7 @@ class BaseTrainer(ABC): importance_sampling_strategy: Literal[ "per_token", "per_sequence" ] = "per_token", + no_rloo_grouping: bool = False, ): """ Initialize the REINFORCE trainer with reward shaping for multi-agent or single-agent training. @@ -215,6 +219,7 @@ class BaseTrainer(ABC): truncated_importance_sampling_ratio_cap ) self.importance_sampling_strategy = importance_sampling_strategy + self.no_rloo_grouping = no_rloo_grouping def mask_non_restricted_token_logits(self, logits: torch.Tensor) -> torch.Tensor: """ @@ -228,8 +233,7 @@ class BaseTrainer(ABC): Returns: torch.Tensor: The masked logits tensor. """ - # TODO: verify. Not sure what we do here is differentiable - # also, we recompute for nothing + # Gradients flow only through the kept logits; masking is recomputed per batch for clarity. if self.restrict_tokens is not None: allowed_token_ids = [] @@ -251,28 +255,6 @@ class BaseTrainer(ABC): return logits - # def get_gradient_magnitude(self, loss_term: torch.Tensor) -> float: - # """ - # Computes the L2 norm of the gradients of the given loss term with respect to the model parameters. - - # Args: - # loss_term (torch.Tensor): The loss tensor to compute gradients for. - - # Returns: - # float: The L2 norm of the gradients, or 0.0 if no gradients are present. - # """ - # with torch.no_grad(): - # grads = torch.autograd.grad( - # loss_term, - # [p for p in self.policy.parameters() if p.requires_grad], - # retain_graph=True, - # allow_unused=True, - # ) - # grads = [g for g in grads if g is not None] - # if not grads: - # return torch.tensor(0.0, device=loss_term.device) - # return torch.norm(torch.stack([g.norm(2) for g in grads])).item() - def apply_reinforce_step( self, training_batch: TrainingBatch, @@ -378,8 +360,8 @@ class BaseTrainer(ABC): metric_id="next_token_credit", metrics=credits_mb ) - # Forward pass + cast to FP-32 for higher prec. - # TODO: create attention mask if not relying on default (assume causal llm) + # Forward pass + cast to FP-32 for higher prec. Causal LM attention masks are implicit; + # wire up a custom mask here only if the policy deviates from standard autoregressive behavior. logits = self.policy(input_ids=contexts_mb)[0] # (B, S, V) # Mask non-restricted tokens @@ -650,8 +632,7 @@ class BaseTrainer(ABC): for key, value in running_mean_logs.items(): self.tally.add_metric(path=key, metric=value) - # Clear - # TODO: verify + # Clear accelerator state so we do not accumulate references between optimizer steps. self.accelerator.clear(self.policy, self.policy_optimizer) import gc @@ -663,7 +644,7 @@ class BaseTrainer(ABC): self, trajectories: TrajectoryBatch, critic_loss_scaling_factor: float = 2.0 ) -> torch.FloatTensor: """ - TOWRITE + Compute (and optionally whiten) advantages while training the critic in mini-batches. Uses GAE if enabled, otherwise uses Monte Carlo returns. Optionally trains the critic if GAE is used. Returns: @@ -862,7 +843,7 @@ class BaseTrainer(ABC): trajectories.crn_ids.unique().shape[0] != trajectories.crn_ids.shape[0] ) - if is_grouped_by_rng: + if is_grouped_by_rng and not self.no_rloo_grouping: for crn_id in trajectories.crn_ids.unique(): rng_mask = trajectories.crn_ids == crn_id rng_advantages = padded_advantages[rng_mask] @@ -920,7 +901,7 @@ class BaseTrainer(ABC): self, agent_id: str, roots: list[RolloutTreeRootNode] ) -> None: """ - TOWRITE + Populate self.training_data for a single agent using the provided rollout trees. """ pass @@ -928,7 +909,7 @@ class BaseTrainer(ABC): self, roots: list[RolloutTreeRootNode], agent_ids: list[str] ) -> None: """ - TOWRITE + Convenience wrapper to ingest trajectory data for every training agent. """ for agent_id in agent_ids: self.set_agent_trajectory_data(agent_id, roots) @@ -943,12 +924,9 @@ class BaseTrainer(ABC): def set_policy_gradient_data(self, agent_ids: list[str]) -> None: """ - Already set earlier # TODO: make it separate and clean + Reset and rebuild the policy-gradient minibatches before iterating through agents. """ self.policy_gradient_data = None - # for agent_id, trajectory_batch in self.training_data.items(): - # if "buffer" in agent_id: - # continue for agent_id in agent_ids: assert "buffer" not in agent_id, "Buffer agents do not train policy" trajectory_batch = self.training_data[agent_id] @@ -978,7 +956,7 @@ class BaseTrainer(ABC): def train(self) -> None: """ - TOWRITE + Entry point for policy updates: prepare batches, compute gradients, and update parameters. """ assert self.policy_gradient_data is not None, "Policy gradient data is not set" if self.critic_optimizer is not None: diff --git a/src_code_for_reproducibility/training/trainer_independent.py b/src_code_for_reproducibility/training/trainer_independent.py index 30c1505a7fa07fcbe7044c4e83a6b7f556fe3817..807c69b3e903024028d8255ae3e79eb6537e609b 100644 --- a/src_code_for_reproducibility/training/trainer_independent.py +++ b/src_code_for_reproducibility/training/trainer_independent.py @@ -1,6 +1,8 @@ """ - +File: mllm/training/trainer_independent.py +Summary: Trainer for independently optimizing each agent. """ + import logging import os import sys @@ -41,6 +43,8 @@ logger.addHandler(logging.StreamHandler(sys.stdout)) @dataclass class TrainingData: + """Caches per-agent trajectory tensors plus their computed advantages.""" + agent_id: str main_data: TrajectoryBatch # list-of-tensors: per rollout advantages with length jT @@ -52,9 +56,9 @@ class TrainerNaive(BaseTrainer): self, agent_id: str, roots: list[RolloutTreeRootNode] ) -> None: """ - TOWRITE + Tokenize rollouts for a given agent and cache the tensors used for training. """ - # TODO: append to current batch data instead, else we will only train for one agent! + # Reset per-agent buffers; extend this logic if joint training batches are needed. self.policy_gradient_data = None # Tensorize Chats diff --git a/src_code_for_reproducibility/training/trainer_sum_rewards.py b/src_code_for_reproducibility/training/trainer_sum_rewards.py index 8f2ca9d16bf3b3990f01aeb99b6941f6a1df87d2..336a542bbf13691a9041bcf15da063f3183db4fe 100644 --- a/src_code_for_reproducibility/training/trainer_sum_rewards.py +++ b/src_code_for_reproducibility/training/trainer_sum_rewards.py @@ -1,6 +1,8 @@ """ - +File: mllm/training/trainer_sum_rewards.py +Summary: Trainer that optimizes the sum-of-rewards objective. """ + import logging import os import sys @@ -20,7 +22,7 @@ from mllm.training.credit_methods import ( get_discounted_returns, get_discounted_state_visitation_credits, get_generalized_advantage_estimates, - get_rloo_credits, + get_rloo_credits, ) from mllm.training.tally_metrics import Tally from mllm.training.tally_rollout import RolloutTally, RolloutTallyItem @@ -44,9 +46,7 @@ logger.addHandler(logging.StreamHandler(sys.stdout)) class TrainerSumRewards(TrainerNaive): def receive_advantage_data(self, advantage_packets: list[AdvantagePacket]): - """ - Sums the advantages of the other trainers - """ + """Sum peer advantages onto this agent's advantages to optimize joint reward.""" logger.info(f"Receiving advantage packets.") assert ( diff --git a/src_code_for_reproducibility/training/training_data_utils.py b/src_code_for_reproducibility/training/training_data_utils.py index deb59e5c575c8d18e8f49d348c8aa342f1c40d22..2685ee93b27a1208c9ee36a3605bd3321f4fe9fc 100644 --- a/src_code_for_reproducibility/training/training_data_utils.py +++ b/src_code_for_reproducibility/training/training_data_utils.py @@ -1,3 +1,8 @@ +""" +File: mllm/training/training_data_utils.py +Summary: Utilities for loading, filtering, and batching training data. +""" + from dataclasses import dataclass from typing import Literal, Optional, Tuple @@ -14,6 +19,8 @@ from mllm.markov_games.rollout_tree import ( @dataclass class AdvantagePacket: + """Message used by trainers to share per-rollout advantages.""" + agent_id: str rollout_ids: torch.IntTensor # (B,) # list-of-tensors @@ -21,10 +28,8 @@ class AdvantagePacket: class TrainingChatTurn: - # TODO: simplify by making this a child of ChatTurn """ - This class contains the chat turns for a single agent. - It is like ChatTurn, but with the time step added. + Lightweight ChatTurn variant that records the timestep index alongside role/content. """ def __init__( @@ -70,7 +75,7 @@ def get_main_chat_list_and_rewards( This method traverses a rollout tree and returns a the list of ChatTurn for an agent. If it encounters a branch node, it follows the main path. """ - # TODO; extend for all trees, not just linear + # Currently follows only the main branch; extend if side branches must be included. if isinstance(root, RolloutTreeRootNode): current_node = root.child else: @@ -101,9 +106,9 @@ def get_tokenwise_credits( batch_credits: torch.FloatTensor | torch.Tensor, # (B, jT) ) -> torch.FloatTensor | torch.Tensor: # (B, jS) """ - TOWRITE + Expand per-state credits so every token at that timestep receives the same value. """ - # TODO vectorize this code + # The explicit loops keep jagged tensor semantics simple; optimize later if profiling warrants it. batch_token_credits = [] for credits, timesteps in zip(batch_credits, batch_timesteps): token_credits = torch.zeros_like( @@ -260,6 +265,8 @@ timestep = int @dataclass class PaddedTensorTrainingBatch: + """Helper struct returned by ``TrainingBatch.get_padded_tensors``.""" + batch_input_ids: torch.LongTensor | torch.Tensor batch_action_mask: torch.BoolTensor | torch.Tensor batch_entropy_mask: Optional[torch.BoolTensor | torch.Tensor] @@ -290,11 +297,6 @@ class TrainingBatch: batch_timesteps: list[torch.IntTensor] # List[(jS,)] def __post_init__(self): - # Put everything in the right device - # self.rollout_ids = self.rollout_ids.to("cuda" if torch.cuda.is_available() else "cpu") - # self.batch_input_ids = self.batch_input_ids.to("cuda" if torch.cuda.is_available() else "cpu") - # self.batch_action_mask = self.batch_action_mask.to("cuda" if torch.cuda.is_available() else "cpu") - # self.batch_credits = self.batch_credits.to("cuda" if torch.cuda.is_available() else "cpu") # Ensure batch dimension is present assert ( len(self.batch_input_ids) @@ -348,8 +350,7 @@ class TrainingBatch: def get_padded_tensors(self, padding: float = 0.0): """ - TOWRITE - Always pad to the right. + Materialize right-padded tensors so PyTorch ops can run on uniform shapes. """ padded_batch_input_ids = pad_sequence( self.batch_input_ids, batch_first=True, padding_value=int(padding) diff --git a/src_code_for_reproducibility/utils/__init__.py b/src_code_for_reproducibility/utils/__init__.py index e69de29bb2d1d6434b8b29ae775ad8c2e48c5391..f8f86a6250af7b404a43dc38a7d58aef50dbeb6d 100644 --- a/src_code_for_reproducibility/utils/__init__.py +++ b/src_code_for_reproducibility/utils/__init__.py @@ -0,0 +1,4 @@ +""" +File: mllm/utils/__init__.py +Summary: Utility package exposing helper modules. +""" diff --git a/src_code_for_reproducibility/utils/__pycache__/__init__.cpython-312.pyc b/src_code_for_reproducibility/utils/__pycache__/__init__.cpython-312.pyc index 32c4e0f8b0b09f12cba13137dbdc2a6f1d367f0f..e63b428c00db55d1fd97a390ba95c8e0e2b8f6a7 100644 Binary files a/src_code_for_reproducibility/utils/__pycache__/__init__.cpython-312.pyc and b/src_code_for_reproducibility/utils/__pycache__/__init__.cpython-312.pyc differ diff --git a/src_code_for_reproducibility/utils/__pycache__/dict_get_path.cpython-312.pyc b/src_code_for_reproducibility/utils/__pycache__/dict_get_path.cpython-312.pyc index ec38a350f3ab2827c91067693e5aaed5cd6e1385..bf519a8c2dd0e9288fbeec449932cb32f838bea2 100644 Binary files a/src_code_for_reproducibility/utils/__pycache__/dict_get_path.cpython-312.pyc and b/src_code_for_reproducibility/utils/__pycache__/dict_get_path.cpython-312.pyc differ diff --git a/src_code_for_reproducibility/utils/__pycache__/get_coagent_id.cpython-312.pyc b/src_code_for_reproducibility/utils/__pycache__/get_coagent_id.cpython-312.pyc index 05691ab328cc6ccf7a54fc828757fc12582c8ec5..a954fa00312e4c613fe3feb2fe2248988dc6f00d 100644 Binary files a/src_code_for_reproducibility/utils/__pycache__/get_coagent_id.cpython-312.pyc and b/src_code_for_reproducibility/utils/__pycache__/get_coagent_id.cpython-312.pyc differ diff --git a/src_code_for_reproducibility/utils/__pycache__/resource_context.cpython-312.pyc b/src_code_for_reproducibility/utils/__pycache__/resource_context.cpython-312.pyc index 6b9763f1e29c04e7debd256efe7bcfc0cdcfefc8..084c8810b7f0ced9d7ecc96db6933f4bd0a8931d 100644 Binary files a/src_code_for_reproducibility/utils/__pycache__/resource_context.cpython-312.pyc and b/src_code_for_reproducibility/utils/__pycache__/resource_context.cpython-312.pyc differ diff --git a/src_code_for_reproducibility/utils/__pycache__/rollout_tree_gather_utils.cpython-312.pyc b/src_code_for_reproducibility/utils/__pycache__/rollout_tree_gather_utils.cpython-312.pyc index fdd056e5ba1aceb1536f2eab49f1c63fe7fd647d..348e542fbd081961b48ba11f57ac98c99d11e41c 100644 Binary files a/src_code_for_reproducibility/utils/__pycache__/rollout_tree_gather_utils.cpython-312.pyc and b/src_code_for_reproducibility/utils/__pycache__/rollout_tree_gather_utils.cpython-312.pyc differ diff --git a/src_code_for_reproducibility/utils/__pycache__/rollout_tree_stats.cpython-312.pyc b/src_code_for_reproducibility/utils/__pycache__/rollout_tree_stats.cpython-312.pyc index 5eb56226df7b241e081728dc406c4f0162cba1fa..34a48d6500c7c7e62e47a7def5f05e8824eaac0c 100644 Binary files a/src_code_for_reproducibility/utils/__pycache__/rollout_tree_stats.cpython-312.pyc and b/src_code_for_reproducibility/utils/__pycache__/rollout_tree_stats.cpython-312.pyc differ diff --git a/src_code_for_reproducibility/utils/__pycache__/short_id_gen.cpython-312.pyc b/src_code_for_reproducibility/utils/__pycache__/short_id_gen.cpython-312.pyc index 10898f798ab2178b950e57100739d421c01f00f7..6f20495b7f4d59902c0ff611c84ba21a2311e77b 100644 Binary files a/src_code_for_reproducibility/utils/__pycache__/short_id_gen.cpython-312.pyc and b/src_code_for_reproducibility/utils/__pycache__/short_id_gen.cpython-312.pyc differ diff --git a/src_code_for_reproducibility/utils/__pycache__/stat_pack.cpython-312.pyc b/src_code_for_reproducibility/utils/__pycache__/stat_pack.cpython-312.pyc index 65e397d163b96f1a0b90c7541734bd729bb82c9e..9a454b258bff6fdfbd14a21b1b02cf2e957747f8 100644 Binary files a/src_code_for_reproducibility/utils/__pycache__/stat_pack.cpython-312.pyc and b/src_code_for_reproducibility/utils/__pycache__/stat_pack.cpython-312.pyc differ diff --git a/src_code_for_reproducibility/utils/__pycache__/update_start_epoch.cpython-312.pyc b/src_code_for_reproducibility/utils/__pycache__/update_start_epoch.cpython-312.pyc index 3ed6ba62682410f9b2fe69fed3538a96d8fda887..6f9d3c6bca8b73cbb23dff3263b3a5f12f598599 100644 Binary files a/src_code_for_reproducibility/utils/__pycache__/update_start_epoch.cpython-312.pyc and b/src_code_for_reproducibility/utils/__pycache__/update_start_epoch.cpython-312.pyc differ diff --git a/src_code_for_reproducibility/utils/__pycache__/wandb_utils.cpython-312.pyc b/src_code_for_reproducibility/utils/__pycache__/wandb_utils.cpython-312.pyc index 09bd9177e48a4140805ea2695cb567ffe987d70a..e4da2d0a7c94f30a7b0dcb724c0e4eee16bebd6d 100644 Binary files a/src_code_for_reproducibility/utils/__pycache__/wandb_utils.cpython-312.pyc and b/src_code_for_reproducibility/utils/__pycache__/wandb_utils.cpython-312.pyc differ diff --git a/src_code_for_reproducibility/utils/dict_get_path.py b/src_code_for_reproducibility/utils/dict_get_path.py index cad0acbc4ecd3003dc3ece71a586b4796c0796af..16b91ec7ec8ecf4e5ed96af29945f44d27bd0276 100644 --- a/src_code_for_reproducibility/utils/dict_get_path.py +++ b/src_code_for_reproducibility/utils/dict_get_path.py @@ -1,5 +1,10 @@ +""" +File: mllm/utils/dict_get_path.py +Summary: Retrieves nested dictionary values using dotted key paths. +""" -def get_from_nested_dict(a:dict, path) -> any: + +def get_from_nested_dict(a: dict, path) -> any: # path is string or list of string try: if isinstance(path, str): diff --git a/src_code_for_reproducibility/utils/rollout_tree_stats.py b/src_code_for_reproducibility/utils/rollout_tree_stats.py index 1ac3cd0e34212e7fdbeba19e501b7d96a5f128e1..4725160156230d7efb89588c765fb5b63a7bbbe1 100644 --- a/src_code_for_reproducibility/utils/rollout_tree_stats.py +++ b/src_code_for_reproducibility/utils/rollout_tree_stats.py @@ -1,3 +1,8 @@ +""" +File: mllm/utils/rollout_tree_stats.py +Summary: Computes descriptive statistics from rollout tree collections. +""" + from typing import Any, Callable, List, Tuple from mllm.markov_games.rollout_tree import RolloutTreeRootNode @@ -39,7 +44,7 @@ def get_mean_rollout_tree_stats( metrics: List[Callable[[SimulationStepLog], Any]], ) -> StatPack: """Get the mean stats for a list of rollout trees.""" - # TODO complete this + # Compute per-rollout means first, then aggregate them across the entire batch. stat_tallies = [ get_rollout_tree_mean_stats(rollout_tree, metrics) for rollout_tree in rollout_trees diff --git a/src_code_for_reproducibility/utils/stat_pack.py b/src_code_for_reproducibility/utils/stat_pack.py index 46b397139a1a8a4149030a9cc33d2b3afb7b4a12..d4da475dafa8e3290ba9be10922be5687ac2c862 100644 --- a/src_code_for_reproducibility/utils/stat_pack.py +++ b/src_code_for_reproducibility/utils/stat_pack.py @@ -1,3 +1,8 @@ +""" +File: mllm/utils/stat_pack.py +Summary: Implements the StatPack container for incremental statistics. +""" + import csv import json import os @@ -10,11 +15,10 @@ from typing import Any, Dict, Iterator, List, Optional, Tuple, TypedDict import matplotlib.pyplot as plt import numpy as np -from torch.utils.tensorboard import SummaryWriter -plt.style.use( - "https://raw.githubusercontent.com/dereckpiche/DedeStyle/refs/heads/main/dedestyle.mplstyle" -) +style_path = os.environ.get("ADALIGN_MPLSTYLE") +if style_path: + plt.style.use(style_path) import wandb @@ -65,7 +69,7 @@ class StatPack: mean_st = StatPack() for key in self.keys(): if isinstance(self[key], list): - # TODO: exclude None values + # Ignore None entries so missing measurements do not bias the mean. non_none_values = [v for v in self[key] if v is not None] if non_none_values: mean_st[key] = np.mean(np.array(non_none_values)) diff --git a/src_code_for_reproducibility/utils/wandb_utils.py b/src_code_for_reproducibility/utils/wandb_utils.py index 3e5d83ed1a5304f208288f78457582b0acdb58c4..46289bfdbb48b72fa3fe3b531d150447cfc1eb01 100644 --- a/src_code_for_reproducibility/utils/wandb_utils.py +++ b/src_code_for_reproducibility/utils/wandb_utils.py @@ -1,7 +1,11 @@ +""" +File: mllm/utils/wandb_utils.py +Summary: Shared Weights & Biases helper functions. +""" + import os from typing import Any, Dict, Optional - _WANDB_AVAILABLE = False _WANDB_RUN = None @@ -125,7 +129,9 @@ def _summarize_value(value: Any) -> Dict[str, Any]: return {"text": str(value)} -def log_tally(array_tally: Dict[str, Any], prefix: str = "", step: Optional[int] = None) -> None: +def log_tally( + array_tally: Dict[str, Any], prefix: str = "", step: Optional[int] = None +) -> None: """ Flatten and summarize Tally.array_tally and log to WandB. Each leaf list/array is summarized with mean/min/max/last/size. @@ -153,12 +159,12 @@ def log_tally(array_tally: Dict[str, Any], prefix: str = "", step: Optional[int] log(summarized, step=step) -def log_flat_stats(stats: Dict[str, Any], prefix: str = "", step: Optional[int] = None) -> None: +def log_flat_stats( + stats: Dict[str, Any], prefix: str = "", step: Optional[int] = None +) -> None: if not _WANDB_AVAILABLE or _WANDB_RUN is None: return flat: Dict[str, Any] = {} _flatten(prefix, stats, flat) if flat: log(flat, step=step) - -